Chapter 4idei.fr/sites/default/files/medias/doc/by/crampes/cramp... · 2014. 6. 12. · Box 4-2 for...

29
Chapter 4 COST RECOVERY AND SHORT-RUN EFFICIENCY Claude Crampes Gremaq and Idei, University of Toulouse, France When designing a tariff for the transport of electricity, the main difficulty is that the transport industry apparently incurs high fixed costs and no real variable cost. In effect, when the infrastructure is installed and when the operators are at their workplace, the only input which is necessary to deliver electricity at a given withdrawal node is electricity at some injection node since electricity is flowing by itself. Consequently, at first sight the problem is just to allocate fixed costs, mainly infrastructure maintenance costs, wages and financial charges, among the different types of users of the grid. As a matter of fact, the transport of electricity creates two significant variable costs. 1 One is an internal cost, that is, a cost in terms of electricity: a fraction of the energy which is injected into the grid will be lost during transport. It results that the consumption of 1 MWh of electricity requires the generation of (1+L) MWh, and this extra L MWh is a real cost for which producers must be compensated. Additionally, because the lines and nodes used for transport have a limited capacity, the optimal allocation of production and consumption is not as efficient as it would be, absent any grid constraint. And pricing must include this economic cost due to congestion. When transport prices are computed using the marginal values of these two costs (the so-called “nodal prices”), they provide a revenue larger than the losses of energy. This surplus can be used to pay for fixed costs of transport but, in most cases, it is not large enough to balance the budget of the operator. This explains why transport tariffs must be either second-best linear prices (Ramsey prices) or non linear prices. The chapter presents the normative principles of the economic analysis of pricing in electricity transport. It is exclusively dedicated to short-term

Transcript of Chapter 4idei.fr/sites/default/files/medias/doc/by/crampes/cramp... · 2014. 6. 12. · Box 4-2 for...

Page 1: Chapter 4idei.fr/sites/default/files/medias/doc/by/crampes/cramp... · 2014. 6. 12. · Box 4-2 for some illustrations of network topologies). This explains why, in the optimal dispatch,

Cha

pter

4

CO

ST R

EC

OV

ER

Y A

ND

SH

OR

T-R

UN

E

FFIC

IEN

CY

C

laud

e C

ram

pes

Gre

maq

and

Idei

, Uni

vers

ity o

f Tou

lous

e, F

ranc

e

Whe

n de

sign

ing

a ta

riff f

or th

e tra

nspo

rt of

ele

ctric

ity, t

he m

ain

diff

icul

ty

is th

at th

e tra

nspo

rt in

dust

ry a

ppar

ently

incu

rs h

igh

fixed

cos

ts a

nd n

o re

al

varia

ble

cost

. In

eff

ect,

whe

n th

e in

fras

truct

ure

is i

nsta

lled

and

whe

n th

e op

erat

ors

are

at th

eir w

orkp

lace

, the

onl

y in

put w

hich

is n

eces

sary

to d

eliv

er

elec

trici

ty a

t a

give

n w

ithdr

awal

nod

e is

ele

ctric

ity a

t som

e in

ject

ion

node

si

nce

elec

trici

ty is

flow

ing

by it

self.

Con

sequ

ently

, at f

irst s

ight

the

prob

lem

is

just

to a

lloca

te fi

xed

cost

s, m

ainl

y in

fras

truct

ure

mai

nten

ance

cos

ts, w

ages

an

d fin

anci

al c

harg

es, a

mon

g th

e di

ffer

ent t

ypes

of u

sers

of t

he g

rid.

A

s a

mat

ter

of f

act,

the

trans

port

of e

lect

ricity

cre

ates

tw

o si

gnifi

cant

va

riabl

e co

sts.1 O

ne is

an

inte

rnal

cos

t, th

at is

, a c

ost i

n te

rms

of e

lect

ricity

: a

frac

tion

of t

he e

nerg

y w

hich

is

inje

cted

int

o th

e gr

id w

ill b

e lo

st d

urin

g tra

nspo

rt.

It re

sults

that

the

cons

umpt

ion

of 1

MW

h of

ele

ctric

ity r

equi

res

the

gene

ratio

n of

(1+L

) MW

h, a

nd th

is e

xtra

L M

Wh

is a

real

cos

t for

whi

ch

prod

ucer

s m

ust b

e co

mpe

nsat

ed.

Add

ition

ally

, bec

ause

the

lines

and

nod

es

used

fo

r tra

nspo

rt ha

ve

a lim

ited

capa

city

, th

e op

timal

al

loca

tion

of

prod

uctio

n an

d co

nsum

ptio

n is

not

as

effic

ient

as

it w

ould

be,

abs

ent

any

grid

con

stra

int.

And

pric

ing

mus

t in

clud

e th

is e

cono

mic

cos

t du

e to

co

nges

tion.

W

hen

trans

port

pric

es a

re c

ompu

ted

usin

g th

e m

argi

nal v

alue

s of

thes

e tw

o co

sts

(the

so-c

alle

d “n

odal

pric

es”)

, the

y pr

ovid

e a

reve

nue

larg

er th

an

the

loss

es o

f en

ergy

. T

his

surp

lus

can

be u

sed

to p

ay f

or f

ixed

cos

ts o

f tra

nspo

rt bu

t, in

mos

t cas

es, i

t is

not l

arge

eno

ugh

to b

alan

ce th

e bu

dget

of

the

oper

ator

. Th

is e

xpla

ins

why

tran

spor

t tar

iffs

mus

t be

eith

er s

econ

d-be

st

linea

r pric

es (R

amse

y pr

ices

) or n

on li

near

pric

es.

The

chap

ter p

rese

nts t

he n

orm

ativ

e pr

inci

ples

of t

he e

cono

mic

ana

lysi

s of

pric

ing

in e

lect

ricity

tra

nspo

rt.

It is

exc

lusi

vely

ded

icat

ed t

o sh

ort-t

erm

Page 2: Chapter 4idei.fr/sites/default/files/medias/doc/by/crampes/cramp... · 2014. 6. 12. · Box 4-2 for some illustrations of network topologies). This explains why, in the optimal dispatch,

106

Tran

spor

t pri

cing

of e

lect

rici

ty n

etw

orks

anal

ysis

, th

at i

s to

the

ope

ratio

n of

a g

iven

ele

ctric

ity n

etw

ork.

Th

e de

velo

pmen

t of t

he tr

ansp

ort i

nfra

stru

ctur

e is

ana

lyse

d in

Cha

pter

5.

In

Sec

tion

1, w

e de

fine

the

varia

ble

cost

of

trans

port

as th

e su

m o

f th

e co

st o

f con

gest

ion

and

the

cost

of o

hmic

loss

es.

In S

ectio

n 2,

we

show

how

no

dal

pric

es c

an b

e us

ed t

o de

sign

tar

iffs

for

trans

port.

Se

ctio

n 3

is

dedi

cate

d to

the

prob

lem

of f

und

risin

g in

ord

er to

bal

ance

the

budg

et o

f the

tra

nspo

rt op

erat

or.

Firs

tly,

we

focu

s on

Ram

sey

pric

es,

whi

ch a

re t

he

seco

nd-b

est

linea

r pr

ices

w

hen

one

tries

to

re

ach

effic

ienc

y w

ithou

t im

pairi

ng t

he b

udge

t eq

uilib

rium

of

the

trans

port

firm

. S

econ

dly,

we

cons

ider

a s

peci

al c

lass

of

non-

linea

r pr

ices

, nam

ely,

tw

o-pa

rt ta

riffs

. W

e co

nclu

de in

Sec

tion

4.

We

do n

ot d

iscu

ss h

ow t

o de

fine

trans

mis

sion

rig

hts

on t

he t

rans

port

infr

astru

ctur

e an

d th

e co

nflic

t bet

wee

n th

e su

ppor

ters

of p

hysi

cal r

ight

s an

d th

e su

ppor

ters

of

finan

cial

rig

hts.2

Nei

ther

do

we

cons

ider

the

ow

ners

hip

and

the

gove

rnan

ce o

f th

e tra

nspo

rt fir

m.

We

supp

ose

that

use

rs f

ace

no

barr

ier

to g

ain

acce

ss to

the

grid

. Th

e re

gula

tion

of th

e tra

nspo

rt op

erat

or

unde

r alte

rnat

ive

hypo

thes

es c

once

rnin

g ve

rtica

l int

egra

tion

and

com

petit

ion

betw

een

grid

use

rs is

scru

tinis

ed in

Cha

pter

6.

Whi

le i

n th

e fo

llow

ing

sect

ions

w

e ha

ve a

dopt

ed a

non

-tech

nica

l pr

esen

tatio

n, th

e re

ader

can

fin

d in

the

appe

ndix

a f

orm

al m

odel

ling

of th

e m

ain

resu

lts.

1.

FIR

ST-B

EST

DIS

PAT

CH

IN A

N E

LE

CT

RIC

ITY

N

ET

WO

RK

1.1

The

tran

spor

t of e

lect

rici

ty

In e

cono

mic

term

s, a

good

is d

efin

ed b

y: (i

) som

e in

trins

ic c

hara

cter

istic

s (w

eigh

t, si

ze, q

ualit

y, e

tc.);

(ii)

the

loca

tion;

(iii

) the

dat

e; a

nd (i

v) th

e st

ate

of n

atur

e w

here

it is

ava

ilabl

e. T

rans

port

is th

e ac

tivity

that

mai

nly

cons

ists

in

mod

ifyin

g at

tribu

te (

ii), e

ven

if, a

s si

de e

ffec

ts, t

he th

ree

othe

r at

tribu

tes

are

also

mod

ified

in

mos

t ca

ses.

It

resu

lts t

hat,

to a

naly

se t

he u

tility

of

trans

porti

ng a

spe

cific

goo

d, w

e ne

ed to

ana

lyse

the

utili

ty a

nd th

e co

st o

f th

at g

ood

at th

e de

partu

re a

nd a

rriv

al lo

catio

ns.

The

diff

eren

ce b

etw

een

the

net u

tility

of t

he g

ood

at th

e ar

rival

loca

tion

and

at th

e de

partu

re lo

catio

n is

th

e gr

oss

utili

ty fr

om tr

ansp

ortin

g it.

Thi

s di

ffer

ence

is to

be

com

pare

d w

ith

the

cost

of t

rans

port

in o

rder

to d

ecid

e if

the

good

is to

be

disp

lace

d or

if it

sh

ould

rem

ain

at th

e in

itial

loca

tion.

Th

is i

s th

e no

rmat

ive

prin

cipl

e th

at w

e ha

ve t

o ap

ply

whe

n an

alys

ing

elec

trici

ty t

rans

port.

Th

e st

artin

g po

int

is t

o de

term

ine

the

quan

titie

s to

Page 3: Chapter 4idei.fr/sites/default/files/medias/doc/by/crampes/cramp... · 2014. 6. 12. · Box 4-2 for some illustrations of network topologies). This explains why, in the optimal dispatch,

Cos

t rec

over

y an

d sh

ort-r

un e

ffici

ency

10

7

gene

rate

and

to c

onsu

me

in o

rder

to m

axim

ise

the

wel

fare

of

all t

he a

gent

s th

at u

se t

he t

rans

port

infr

astru

ctur

e.

In t

he s

hort

run,

the

equ

ipm

ent

for

gene

ratio

n, t

rans

porta

tion

and

dist

ribut

ion

is f

ixed

. T

he p

refe

renc

es o

f co

nsum

ers

also

are

fixe

d. T

he o

ptim

al a

lloca

tion

is li

mite

d to

dec

idin

g ho

w

muc

h to

gen

erat

e at

eac

h no

de a

nd h

ow m

uch

to c

onsu

me

at e

ach

node

, gi

ven

the

rest

rain

ts im

pose

d by

the

topo

logi

cal c

hara

cter

istic

s of t

he n

etw

ork

and

the

tech

nica

l ca

pabi

lity

of e

ach

piec

e of

equ

ipm

ent

(see

Box

4-1

for

de

tails

on

the

obje

ctiv

e to

max

imis

e w

elfa

re u

nder

alte

rnat

ive

sets

of

cons

train

ts).

Box

4-1

: Fir

st-b

est,

seco

nd-b

est,

and

cons

trai

nts t

o m

axim

ise

wel

fare

In

the

shor

t run

, the

who

le g

ener

atio

n an

d tra

nspo

rtatio

n eq

uipm

ents

are

fixe

d, a

s w

ell a

s all

the

need

s.

Th

e “g

rid-f

ree”

firs

t be

st a

lloca

tion

is t

he s

et o

f qu

antit

ies

of e

lect

ricity

ge

nera

ted

and

cons

umed

at e

ach

node

that

max

imis

es th

e ne

t wel

fare

, tha

t is,

the

sum

of

the

diff

eren

ce b

etw

een

the

utili

ty o

f el

ectri

city

for

con

sum

ers

and

the

cost

to g

ener

ate

it, in

a fi

ctiti

ous

situ

atio

n w

here

ene

rgy

can

flow

from

one

nod

e to

oth

ers w

ithou

t any

con

stra

int a

nd w

ithou

t any

loss

.

Th

e “g

rid-c

onst

rain

ed”

first

bes

t allo

catio

n al

so m

axim

ises

the

net w

elfa

re, b

ut

taki

ng in

to a

ccou

nt th

e ph

ysic

al c

hara

cter

istic

s of t

he g

rid.

In th

is c

ase:

(i) s

ome

ener

gy i

s lo

st d

urin

g tra

nspo

rt; a

nd (

ii) b

ecau

se s

ome

lines

and

int

erm

edia

ry

node

s ha

ve l

imite

d ca

paci

ty,

the

“grid

-fre

e” o

ptim

al f

low

s ar

e no

lon

ger

feas

ible

, whi

ch c

reat

es a

“co

nges

tion

cost

”.

In

eco

nom

ic t

heor

y, “

seco

nd-b

est”

mai

nly

refe

rs t

o a

situ

atio

n w

here

the

be

nevo

lent

pla

nner

has

to

bala

nce

the

budg

et o

f th

e pr

oduc

ers

he s

uper

vise

s.

Act

ually

, th

is e

xpre

ssio

n ca

n be

use

d in

any

situ

atio

n w

here

a c

onst

rain

t is

ad

ded

to a

n in

itial

allo

catio

n pr

oble

m.

In t

hat

resp

ect,

the

“grid

-con

stra

ined

” fir

st-b

est i

s a se

cond

-bes

t with

resp

ect t

o th

e “g

rid-f

ree”

firs

t bes

t allo

catio

n.

W

hen

a co

nstra

int

is a

dded

to

a gi

ven

optim

isat

ion

prob

lem

, ei

ther

tha

t co

nstra

int

is n

ot b

indi

ng a

nd t

he i

nitia

l al

loca

tion

does

not

cha

nge,

or

it is

bi

ndin

g an

d it

resu

lts i

n a

decr

ease

of

the

initi

al p

erfo

rman

ce.

The

new

al

loca

tion

can

neve

r gi

ve a

hig

her

perf

orm

ance

sin

ce,

if fe

asib

le n

ow,

it w

as

feas

ible

bef

ore

the

new

con

stra

int i

s ad

ded

and

it w

ould

hav

e be

en c

hose

n. T

he

diff

eren

ce b

etw

een

the

perf

orm

ance

with

out

and

the

perf

orm

ance

with

the

co

nstra

int

is t

he e

cono

mic

cos

t of

the

con

stra

int.

For

exa

mpl

e, t

he e

cono

mic

co

st o

f th

e tra

nspo

rt gr

id is

the

diff

eren

ce b

etw

een

the

grid

-fre

e so

cial

wel

fare

an

d th

e gr

id-c

onst

rain

ed so

cial

wel

fare

.

In

the

sam

e w

ay, o

ne c

an m

easu

re th

e co

st o

f add

ition

al c

onst

rain

ts, s

uch

as:

– th

e ob

ligat

ion

to b

alan

ce b

udge

t;

Page 4: Chapter 4idei.fr/sites/default/files/medias/doc/by/crampes/cramp... · 2014. 6. 12. · Box 4-2 for some illustrations of network topologies). This explains why, in the optimal dispatch,

108

Tran

spor

t pri

cing

of e

lect

rici

ty n

etw

orks

– th

e pr

ohib

ition

to d

iscr

imin

ate

on p

rices

; –

rest

rictio

ns o

n ta

riff c

lass

es (l

inea

r, tw

o-pa

rt, e

tc.);

the

inab

ility

of t

he o

pera

tor t

o co

llect

info

rmat

ion

on p

refe

renc

es a

nd c

osts

; –

the

univ

ersa

l ser

vice

obl

igat

ion;

etc.

Sh

ort-r

un d

ecis

ions

are

con

stra

ined

by

the

inca

paci

ty t

o ad

apt

the

trans

port

infr

astru

ctur

e an

d th

e ge

nera

tion

plan

ts.

For

this

rea

son,

the

shor

t-run

cos

t, in

w

hich

the

re a

re s

igni

fican

t fix

ed c

osts

, is

hig

her

than

the

lon

g-ru

n co

st ex

clus

ivel

y m

ade

of o

ptim

ally

cho

sen

varia

ble

inpu

ts.

To u

nder

stan

d w

hy th

e op

timal

allo

catio

n ca

n be

def

ined

in th

at s

impl

e w

ay,

it is

int

eres

ting

to s

tress

som

e im

porta

nt d

iffer

ence

s be

twee

n th

e tra

nspo

rt of

ele

ctric

ity a

nd o

ther

tran

spor

t act

iviti

es, f

or e

xam

ple

frei

ght o

r pa

ssen

gers

tran

spor

t. F

irstly

, ele

ctric

ity is

hig

hly

stan

dard

ised

, whi

ch m

eans

cl

ose

subs

titut

abili

ty b

etw

een

gene

ratio

n no

des

for

a gi

ven

need

and

, sy

mm

etric

ally

, clo

se s

ubst

ituta

bilit

y be

twee

n co

nsum

ptio

n no

des

for a

giv

en

prod

uctio

n. T

his h

omog

enei

ty p

rope

rty a

llow

s to

pool

qua

ntiti

es.

Seco

ndly

, fo

r el

ectri

city

, the

tim

e at

tribu

te o

f th

e go

od i

s no

t m

odifi

ed b

y tra

nspo

rt.

Pow

er f

low

s in

stan

tane

ousl

y th

roug

h lin

es a

nd i

nter

med

iary

nod

es.

(Se

e B

ox 4

-2 fo

r som

e ill

ustra

tions

of n

etw

ork

topo

logi

es).

Thi

s exp

lain

s why

, in

the

optim

al d

ispa

tch,

inje

ctio

ns a

nd w

ithdr

awal

s ar

e co

ntem

pora

neou

s. F

or

the

sam

e re

ason

, the

dis

patc

her

can

know

for

sur

e th

e st

ate

of n

atur

e at

a

with

draw

al n

ode

whe

n in

ject

ing

pow

er a

t an

othe

r lo

catio

n, w

hich

gre

atly

re

duce

s th

e ra

ndom

ness

of n

et lo

catio

nal u

tiliti

es.

The

third

diff

eren

ce w

ith

mos

t tra

nspo

rt ne

twor

ks is

that

, for

ele

ctric

ity, t

he te

chno

logy

doe

s not

allo

w

to c

ontro

l the

phy

sica

l flo

w o

f en

ergy

thro

ugh

the

grid

(se

e B

ox 4

-2 f

or a

n ill

ustra

tion)

. C

onse

quen

tly, t

he a

ctua

l flo

w o

n ea

ch li

ne c

anno

t be

a co

ntro

l va

riabl

e.

Four

thly

, in

a g

iven

grid

, on

e ca

n pr

edic

t ve

ry p

reci

sely

the

am

ount

of t

rans

port

loss

es b

ecau

se th

ey fo

llow

wel

l kno

wn

phys

ical

law

s.

To s

um u

p, tr

ansp

ortin

g el

ectri

city

con

sist

s in

con

trolli

ng m

odifi

catio

ns

in it

s at

tribu

tes

(ii)

and

(iv)

with

out m

odify

ing

attri

bute

(iii

) an

d pr

ovok

ing

an u

ndes

irabl

e bu

t pr

edic

tabl

e ch

ange

in

attri

bute

(i).

Th

e ph

ysic

al p

ath

follo

wed

bec

ause

of t

he tr

ansf

orm

atio

n in

attr

ibut

e (ii

) can

not b

e co

ntro

lled;

it

resu

lts t

hat

trans

port

betw

een

two

node

s in

a m

eshe

d ne

twor

k cr

eate

s ex

tern

aliti

es o

n al

l lin

es a

nd n

odes

inte

ntio

nally

and

uni

nten

tiona

lly c

ross

ed

by th

e en

ergy

flow

. In

the

optim

al a

lloca

tion

defin

ed b

y a

bene

vole

nt s

ocia

l pla

nner

, at e

ach

node

, mar

gina

l ut

ility

is

equa

l to

mar

gina

l co

st.

If, a

t on

e no

de, m

argi

nal

cost

wer

e hi

gher

than

mar

gina

l util

ity, t

he la

st k

Wh

wou

ld b

e ge

nera

ted

at

loss

and

, sym

met

rical

ly, i

f m

argi

nal c

ost w

ere

less

than

mar

gina

l util

ity, i

t w

ould

mea

n th

at th

e en

tire

pote

ntia

l soc

ial s

urpl

us is

not

cre

ated

. A

t eac

h

Page 5: Chapter 4idei.fr/sites/default/files/medias/doc/by/crampes/cramp... · 2014. 6. 12. · Box 4-2 for some illustrations of network topologies). This explains why, in the optimal dispatch,

Cos

t rec

over

y an

d sh

ort-r

un e

ffici

ency

10

9

node

, th

is e

qual

ity o

f m

argi

nal

cost

and

mar

gina

l ut

ility

will

mos

t lik

ely

requ

ire a

tra

nsfe

r of

ene

rgy.

So

me

node

s w

ill h

ave

to e

xpor

t en

ergy

and

ot

hers

will

be

net i

mpo

rters

, dep

endi

ng o

n th

e co

st s

truct

ure

of g

ener

atio

n an

d gi

ven

the

cons

umer

s' pr

efer

ence

s fo

r el

ectri

city

. W

hen

no p

iece

of

infr

astru

ctur

e ex

hibi

ts c

onge

stio

n, a

nd w

hen

ther

e is

no

ohm

ic l

oss,

the

optim

al a

lloca

tion

is s

uch

that

ene

rgy

has

one

sing

le v

alue

thr

ough

out

the

who

le n

etw

ork,

whi

ch c

an b

e vi

ewed

as

a gi

ant u

niqu

e no

de o

r as

a p

late

. To

see

this

, obs

erve

that

if th

ere

rem

aine

d a

diff

eren

ce b

etw

een

two

node

s, it

wou

ld b

e ea

sy to

incr

ease

the

glob

al s

urpl

us b

y tra

nsfe

rrin

g so

me

kWh

from

th

e no

de w

ith th

e lo

wer

mar

gina

l val

uatio

n to

war

ds th

e no

de w

ith th

e hi

gher

va

luat

ion.

Box

4-2

: Of n

odes

and

line

s

An

elec

tric

grid

can

be

view

ed a

s a

set

of n

odes

, ei

ther

fin

al (

inje

ctio

n an

d w

ithdr

awal

no

des)

or

in

term

edia

ry

(tran

sfor

mer

s, m

eter

s, co

ntro

llers

, et

c.),

inte

rcon

nect

ed b

y lin

es.

The

sim

ples

t ne

twor

k is

mad

e of

one

sin

gle

line

conn

ectin

g tw

o fin

al n

odes

. Th

e “n

orth

-sou

th”

netw

ork

repr

esen

ted

belo

w (s

ee F

igur

e 4-

1) is

a u

sefu

l the

oret

ical

co

nfig

urat

ion

to u

nder

stan

d co

nges

tion

and

loss

es,

but

it al

so g

ives

a r

easo

nabl

y go

od p

ictu

re o

f the

grid

in s

ome

coun

tries

.3 Sin

ce th

ere

exis

ts a

sin

gle

line,

ther

e is

on

e un

ique

pos

sibl

e pa

th f

or t

rans

porti

ng e

lect

ricity

fro

m n

orth

to

sout

h or

fro

m

sout

h to

nor

th.

Abs

ent

any

ener

gy l

oss,

the

phys

ical

equ

ilibr

ium

of

the

elec

tric

indu

stry

impo

ses

gg

ww

ns

ns

qq

qq

, whe

re

g nq (r

espe

ctiv

ely

w nq) s

tand

s for

the

quan

tity

gene

rate

d (r

espe

ctiv

ely

cons

umed

) at t

he n

orth

nod

e an

d g sq (r

espe

ctiv

ely

w sq) s

tand

s fo

r the

qua

ntity

gen

erat

ed (r

espe

ctiv

ely

cons

umed

) at t

he so

uth

node

. C

onse

quen

tly,

the

quan

tity

of e

lect

ricity

flow

ing

on th

e lin

e is

g

ww

gn

ns

sq

qq

q.

But

in m

any

coun

tries

, par

ticul

arly

in c

ontin

enta

l Eur

ope,

ele

ctric

net

wor

ks a

re

mes

hed.

Th

e co

nseq

uenc

e is

that

ther

e is

not

one

uni

que

path

for

ele

ctric

ity to

go

from

one

nod

e to

ano

ther

. Th

is is

illu

stra

ted

in th

e th

ree-

node

net

wor

k he

reaf

ter (

see

Figu

re 4

-2).

Ene

rgy

flow

s fol

low

pat

hs o

f lea

st re

sist

ance

det

erm

ined

by

Kirc

hhof

f's

law

s. S

uppo

se a

gen

erat

ion

node

and

a c

onsu

mpt

ion

node

are

con

nect

ed b

y tw

o lin

es, o

ne w

ith a

res

ista

nce

twic

e th

e ot

her's

. W

hen

gene

rato

rs in

ject

qua

ntity

q a

t on

e no

de a

nd,

assu

min

g no

los

ses,

cons

umer

s w

ithdr

aw t

he s

ame

quan

tity

at t

he

othe

r no

de, t

he f

low

s on

the

low

res

ista

nce

line

and

on th

e hi

gh r

esis

tanc

e lin

e ar

e re

spec

tivel

y 2q

/3 a

nd q

/3.

In a

3-li

ne n

etw

ork

with

the

sam

e re

sist

ance

on

each

line

, lik

e in

Fig

ure

4-2,

if th

ere

are

two

gene

rato

rs in

stal

led

at n

odes

1 a

nd 2

resp

ectiv

ely,

an

d co

nsum

ers

are

loca

ted

at th

e th

ird n

ode,

the

sim

ulta

neou

s in

ject

ion

of

1g q a

nd

2g q w

ill g

ener

ate

a su

perp

ositi

on o

f flo

ws

on t

he l

ines

con

nect

ing

cons

umer

s to

ge

nera

tors

. Fo

r ex

ampl

e, th

e lin

e be

twee

n no

de 1

and

nod

e 3

trans

ports

two

third

s of

the

ener

gy in

ject

ed a

t nod

e 1

plus

one

third

of t

he e

nerg

y in

ject

ed a

t nod

e 2.

By

cont

rast

, on

ly o

ne t

hird

of

the

net

flow

circ

ulat

es o

n th

e lin

e be

twee

n th

e tw

o ge

nera

tors

.

Page 6: Chapter 4idei.fr/sites/default/files/medias/doc/by/crampes/cramp... · 2014. 6. 12. · Box 4-2 for some illustrations of network topologies). This explains why, in the optimal dispatch,

110

Tran

spor

t pri

cing

of e

lect

rici

ty n

etw

orks

nort

h

w sqg sq

w nq

sout

h

g nq

Figu

re 4

-1.

One

line

net

wor

k

12

3

gg

qq

node

2

inje

ctio

n 2g q

node

1

inje

ctio

n 1g q

12

23g

gq

q1

22 3

gg

qq

node

3

cons

umpt

ion

31

2w

gg

qq

q

Figu

re 4

-2.

Thre

e-lin

e ne

twor

k B

ut, a

s ex

plai

ned

belo

w, b

ecau

se o

f lo

sses

and

bec

ause

of

som

e sc

arce

ca

paci

ty in

tran

spor

t, tra

nsfe

rs fr

om o

ne n

ode

to a

noth

er c

anno

t be

done

for

Page 7: Chapter 4idei.fr/sites/default/files/medias/doc/by/crampes/cramp... · 2014. 6. 12. · Box 4-2 for some illustrations of network topologies). This explains why, in the optimal dispatch,

Cos

t rec

over

y an

d sh

ort-r

un e

ffici

ency

11

1

free

. It

mea

ns th

at th

e “v

ery-

first

-bes

t” a

lloca

tion

(or “

grid

-fre

e” a

lloca

tion)

is

not

feas

ible

. Th

e di

spat

cher

can

onl

y re

ach

the

“grid

-con

stra

ined

” op

timal

al

loca

tion

and,

con

sequ

ently

, ene

rgy

valu

atio

n re

sulti

ng f

rom

thi

s di

spat

ch

will

diff

er a

mon

g th

e no

des.

1.2

The

cos

t of c

onge

stio

n

Ass

ume

first

tha

t lo

sses

can

be

negl

ecte

d.

At

impo

rting

nod

es,

the

inco

min

g en

ergy

is li

mite

d by

the

capa

city

of

lines

and

tran

sfor

mer

s al

ong

the

phys

ical

pat

h fo

llow

ed b

y en

ergy

. A

s a

resu

lt, e

lect

ricity

is

rela

tivel

y sc

arce

and

it

is m

ore

valu

ed t

han

if th

ere

wer

e no

cap

acity

con

stra

int.

R

ecip

roca

lly,

at e

xpor

ting

node

s, en

ergy

is

rela

tivel

y in

exc

ess

and

low

va

lued

bec

ause

the

outg

oing

ene

rgy

is li

mite

d by

the

capa

city

of

lines

and

tra

nsfo

rmer

s. T

he m

ore

cong

este

d th

e ne

twor

k, th

e hi

gher

the

disc

repa

ncy

betw

een

noda

l val

uatio

ns.

The

limit

case

is th

e au

tark

y ca

se w

here

line

s ar

e cu

t sot

that

eac

h no

de is

isol

ated

from

the

othe

rs.

The

diff

eren

ce b

etw

een

two

noda

l val

uatio

ns o

f ene

rgy,

abs

ent a

ny lo

ss,

is a

n in

dex

of t

he t

ight

ness

of

the

trans

port

cons

train

t. I

t re

flect

s th

e in

capa

city

of

the

oper

ator

to

incr

ease

gen

erat

ion

at l

ow-c

ost

node

s an

d to

de

crea

se

it at

hi

gh-c

ost

node

s as

w

ell

as

its

inca

paci

ty

to

incr

ease

co

nsum

ptio

n at

hig

h-ut

ility

nod

es a

nd t

o de

crea

se i

t at

low

-util

ity n

odes

. Th

e “m

erit

orde

r” c

omm

ands

that

no

gene

rato

r sho

uld

be d

ispa

tche

d if

ther

e re

mai

ns s

ome

avai

labl

e ca

paci

ty w

ith a

low

er c

ost.

It

is n

o lo

nger

im

plem

enta

ble.

In

the

sim

ples

t ca

se o

f B

ox 4

-3,

with

one

sin

gle

line

conn

ectin

g ef

ficie

nt n

orth

ern

gene

rato

rs w

ith a

sou

th n

ode

whe

re th

ere

are

inef

ficie

nt g

ener

ator

s an

d th

e lo

ad, o

ne c

an e

asily

dra

w th

e gr

id-c

onst

rain

ed

optim

al q

uant

ities

and

mea

sure

how

they

dep

art f

rom

the

grid

-fre

e op

timal

qu

antit

ies.

The

diff

eren

ce in

nod

al v

alua

tions

exa

ctly

refle

cts

the

soci

al c

ost

of h

avin

g an

out

-of-

mer

it-or

der d

ispa

tch

beca

use

of th

e lim

ited

capa

city

for

trans

port

betw

een

north

and

sout

h. T

he d

iffer

ence

is th

e sh

adow

val

ue o

f the

tra

nspo

rt lin

e th

at s

igna

ls b

y ho

w m

uch

soci

al w

elfa

re w

ould

be

incr

ease

d if

the

cons

train

t cou

ld b

e re

laxe

d.

In m

eshe

d ne

twor

ks, e

nerg

y flo

ws

alon

g le

ast

resi

stan

ce p

aths

with

out

the

poss

ibili

ty to

con

trol t

hem

.4 As

a re

sult,

any

inje

ctio

n an

d w

ithdr

awal

of

a gi

ven

quan

tity

at t

wo

dist

inct

nod

es w

ill p

rovo

ke a

tra

nsit

of e

lect

ricity

th

roug

h al

l the

line

s of

the

netw

ork.

If

one

line

is c

onge

sted

, all

path

s w

ill

appe

ar c

onge

sted

. C

onse

quen

tly, i

n a

mes

hed

netw

ork,

con

gest

ion

on o

ne

line

is s

uffic

ient

for n

odal

val

ues

to d

iffer

thro

ugho

ut th

e ne

twor

k. B

ecau

se

of th

is “

cont

agio

n ef

fect

”, th

e du

al v

alue

of t

he c

onge

sted

line

is la

rger

than

th

e m

ere

diff

eren

ce b

etw

een

the

mar

gina

l va

lues

at

the

two

ends

of

the

cong

este

d lin

e.5

The

diff

eren

ce b

etw

een

thes

e tw

o va

lues

ref

lect

s th

e

Page 8: Chapter 4idei.fr/sites/default/files/medias/doc/by/crampes/cramp... · 2014. 6. 12. · Box 4-2 for some illustrations of network topologies). This explains why, in the optimal dispatch,

112

Tran

spor

t pri

cing

of e

lect

rici

ty n

etw

orks

nega

tive

exte

rnal

ities

in

el

ectri

city

tra

nspo

rt,

that

is

, th

e ad

ditio

nal

cong

estio

n co

st d

ue to

loop

-flo

ws.

Box

4-3

: Out

-of-

mer

it-or

der

optim

al d

ispa

tch

Con

side

r th

e no

rth-s

outh

net

wor

k in

Box

4-2

(see

Fig

ure

4-1)

. A

ssum

e th

at a

ll co

nsum

ers

are

in ‘

sout

h’ a

nd t

heir

mar

gina

l ut

ility

fro

m w

ithdr

awin

g th

e qu

antit

y w sq

of

elec

trici

ty i

s gi

ven

by t

he d

ecre

asin

g fu

nctio

n '(

)10

ww

ss

Uq

q (

see

Figu

re

4-3)

. In

‘nor

th’,

elec

trici

ty c

an b

e ge

nera

ted

at a

con

stan

t mar

gina

l cos

t equ

al to

and,

in ‘s

outh

’, th

ere

exis

t pla

nts w

ith a

con

stan

t mar

gina

l cos

t equ

al to

4¢.

The

re is

no

con

stra

int o

f cap

acity

for g

ener

atio

n an

d th

ere

is n

o en

ergy

loss

on

the

line.

Let

K

den

ote

the

capa

city

of t

he li

ne.

12

3

gg

qq

node

2

inje

ctio

n 2g q

node

1

inje

ctio

n 1g q

12

23g

gq

q1

22 3

gg

qq

node

3

cons

umpt

ion

31

2w

gg

qq

q

Figu

re 4

-3.

Out

-of-

mer

it-or

der o

ptim

al d

ispa

tch

If K

is v

ery

larg

e, th

e op

timal

dis

patc

h co

nsis

ts in

pro

duci

ng n

othi

ng in

‘so

uth’

w

here

gen

erat

ion

is v

ery

cost

ly.

The

who

le e

nerg

y co

mes

fro

m ‘

north

’. T

he

optim

al c

onsu

mpt

ion

is s

uch

that

'(

)1

Uq

¢, th

at is

the

grid

-fre

e op

timal

qua

ntity

w sq

=9.

If

K <

9, t

he c

apac

ity o

f the

line

doe

s no

t allo

w to

impo

rt th

is q

uant

ity o

f ene

rgy

from

‘no

rth’.

The

dis

patc

h fir

st c

onsi

sts

in s

atur

atin

g th

e lin

e to

tran

sfer

as

muc

h en

ergy

as

poss

ible

fro

m ‘

north

’, w

hich

is K

. A

fter t

hat,

ther

e ar

e tw

o po

ssib

ilitie

s:

(i) if

the

mar

gina

l util

ity fr

om c

onsu

min

g K

is s

till h

ighe

r tha

n th

e m

argi

nal c

ost o

f ge

nera

tion

in ‘s

outh

’, us

e th

e so

uth

plan

t up

to th

e po

int w

here

the

mar

gina

l util

ity

of e

lect

ricity

is

equa

l to

the

mar

gina

l co

st o

f ge

nera

tion,

tha

t is

'(

)4

g sU

Kq

¢.

Oth

erw

ise,

do

not d

ispa

tch

the

cost

ly p

lant

. In

the

first

cas

e, th

e gr

id-c

onst

rain

ed

optim

al

allo

catio

n is

g nq

K,

6g sq

K

(as

long

as

6

K)

and

the

tota

l co

nsum

ptio

n is

6

w sq.

Page 9: Chapter 4idei.fr/sites/default/files/medias/doc/by/crampes/cramp... · 2014. 6. 12. · Box 4-2 for some illustrations of network topologies). This explains why, in the optimal dispatch,

Cos

t rec

over

y an

d sh

ort-r

un e

ffici

ency

11

3

In t

he s

econ

d ca

se,

that

is

for

K b

etw

een

6 an

d 9,

the

tot

al o

utpu

t is

the

co

nstra

ined

flo

w c

omin

g fr

om ‘

north

’, g nq

K,

0g sq

. T

hese

qua

ntiti

es a

re

grap

hed

in th

e m

iddl

e pa

nel o

f Fig

ure

4-3

as fu

nctio

ns o

f the

cap

acity

of t

he li

ne K

.

Bec

ause

of

the

inef

ficie

nt d

ispa

tch

crea

ted

by t

he l

imite

d ca

paci

ty o

f th

e lin

e,

wel

fare

can

not b

e as

hig

h as

it w

ould

be

if lo

w-c

ost g

ener

ator

s w

ere

loca

ted

at th

e so

uth

node

with

con

sum

ers.

The

low

er p

anel

of F

igur

e 4-

3 re

pres

ents

the

mar

gina

l va

lue

of th

e ca

paci

ty c

onst

rain

t, w

hich

rep

rese

nts

the

mar

gina

l co

st o

f co

nges

tion

. W

hen

K i

s sm

alle

r th

an 6

, on

e ad

ditio

nal

unit

of c

apac

ity w

ould

allo

w t

o su

bstit

ute

one

north

kW

h to

one

sou

th k

Wh,

that

is to

sav

e =

- 1¢

= 3¢

for a

n un

chan

ged

tota

l out

put.

Whe

n K

is b

etw

een

6 an

d 9,

one

add

ition

al u

nit o

f cap

acity

al

low

s to

incr

ease

the

cons

umpt

ion

by m

eans

of m

ore

impo

rts fr

om ‘n

orth

’, so

that

its

val

ue is

the

net

mar

gina

l util

ity o

f el

ectri

city

; =

U'(K

) - 1

¢ =

9 - K

. Fi

nally

, w

hen

K is

larg

er th

an 9

, any

dev

elop

men

t wou

ld b

e us

eles

s, w

hich

is s

igna

lled

by

= 0

. A

s sh

own

in C

hapt

er 5

, the

sha

dow

pric

e of

the

lines

is to

be

com

pare

d w

ith th

e re

al p

rice

of o

ne u

nit o

f eq

uipm

ent t

o kn

ow w

heth

er th

e tra

nspo

rt ca

paci

ty i

s to

be

incr

ease

d or

dec

reas

ed.

Whe

n th

e du

al v

alue

of

the

cons

train

t is

hig

her

than

the

cos

t of

one

add

ition

al u

nit

of e

quip

men

t, th

e tra

nspo

rt lin

e sh

ould

be

deve

lope

d.

And

it

shou

ld b

e do

wns

ized

in

the

oppo

site

cas

e. I

n ac

tual

net

wor

ks, t

he in

fras

truct

ure

is a

lmos

t alw

ays

larg

er

than

its

optim

al s

ize.

The

con

sequ

ence

is th

at th

e sh

adow

pric

e of

line

s an

d tra

nsfo

rmer

s is

less

than

thei

r un

it co

st.

And

it is

eve

n eq

ual t

o ze

ro w

hen

ther

e is

no

cong

estio

n at

all.

1.3

The

cos

t of l

osse

s

Supp

ose

now

tha

t th

ere

is n

o co

nges

tion.

Th

e m

ain

cost

of

deliv

erin

g 1

MW

h at

one

nod

e st

artin

g fr

om a

spe

cific

inje

ctio

n no

de r

esul

ts f

rom

the

fact

that

a q

uant

ity L

of t

he e

nerg

y in

ject

ed w

ill b

e lo

st in

tran

spor

t. It

mea

ns

that

1+ L

MW

h ar

e to

be

gene

rate

d. A

s a

resu

lt, in

an

optim

ised

net

wor

k th

e m

argi

nal v

alua

tion

of 1

MW

h w

ill d

iffer

fro

m o

ne n

ode

to th

e ot

her

by th

e va

lue

of th

e lo

st e

nerg

y. N

ote

that

it is

not

a c

ost i

ncur

red

by th

e tra

nspo

rt gr

id it

self.

It i

s a

cost

due

to th

e di

stan

ce b

etw

een

inje

ctio

n an

d w

ithdr

awal

no

des.

It d

irect

ly c

once

rns g

ener

ator

s and

con

sum

ers.

In e

lect

ric n

etw

orks

, los

ses

incr

ease

pro

porti

onal

ly t

o th

e sq

uare

of

the

ener

gy i

njec

ted.

Th

e co

nseq

uenc

e of

thi

s pr

ecis

e fu

nctio

nal

form

is

that

m

argi

nal l

osse

s are

twic

e hi

gher

than

ave

rage

loss

es.6

Page 10: Chapter 4idei.fr/sites/default/files/medias/doc/by/crampes/cramp... · 2014. 6. 12. · Box 4-2 for some illustrations of network topologies). This explains why, in the optimal dispatch,

114

Tran

spor

t pri

cing

of e

lect

rici

ty n

etw

orks

1.4

The

shor

t-ru

n m

argi

nal c

ost o

f tra

nspo

rt

To s

um u

p, b

ecau

se m

aint

aini

ng a

nd d

evel

opin

g th

e in

fras

truct

ure

is

cost

ly, i

t is

optim

al to

kee

p so

me

cong

estio

n in

mos

t pie

ces

of th

e tra

nspo

rt gr

id.

As

a re

sult,

the

dis

patc

h th

at w

ould

max

imis

e ne

t so

cial

wel

fare

w

ithou

t any

refe

renc

e to

the

grid

(lik

e if

all g

ener

ator

s an

d co

nsum

ers

wer

e lo

cate

d at

the

sam

e pl

ace)

is n

ot fe

asib

le.

The

actu

al d

ispa

tch

is s

ub-o

ptim

al

as c

ompa

red

with

the

fic

titio

us o

ne-n

ode

indu

stry

. T

he c

ost

due

to

cong

estio

n is

equ

al to

the

diff

eren

ce b

etw

een

the

max

imum

wel

fare

obt

aine

d w

ithou

t tra

nspo

rt co

nstra

ints

and

the

wel

fare

tha

t re

sults

fro

m t

he a

ctua

l di

spat

ch.

In

addi

tion

to c

onge

stio

n co

sts,

only

a f

ract

ion

of t

he q

uant

ity

inje

cted

can

be

cons

umed

. Th

e cu

mul

ativ

e ef

fect

s of

thes

e el

emen

ts is

that

in

a

netw

ork

built

fo

r el

ectri

city

tra

nspo

rt,

the

optim

al

allo

catio

n of

qu

antit

ies

to g

ener

ate

and

to w

ithdr

aw a

t ea

ch n

ode

is s

uch

that

mar

gina

l va

luat

ions

will

be

diff

eren

t fro

m o

ne n

ode

to o

ther

s. T

he n

odal

val

uatio

n of

en

ergy

is a

nat

ural

by-

prod

uct o

f the

opt

imis

atio

n al

gorit

hms

used

by

syst

em

oper

ator

s. W

hen

an I

ndep

ende

nt S

yste

m O

pera

tor

com

pute

s th

e fe

asib

ility

of

a g

iven

dis

patc

h on

the

grid

he

cont

rols

, the

val

ue o

f ene

rgy

at e

ach

node

ca

n be

pub

lishe

d in

stan

tane

ousl

y. T

he d

iffer

ence

bet

wee

n no

dal v

alua

tions

th

at i

nclu

des

a re

al c

ost

(the

valu

e of

los

t en

ergy

) an

d a

shad

ow c

ost

(the

valu

e of

lost

eff

icie

ncy)

, is

to b

e vi

ewed

as

the

shor

t-run

mar

gina

l cos

t of

trans

port.

Y

et,

note

tha

t no

ne o

f th

e tw

o co

mpo

nent

s of

the

sho

rt-ru

n m

argi

nal

cost

ca

n be

di

rect

ly

rela

ted

to

an

econ

omic

or

ac

coun

ting

expe

nditu

re in

curr

ed b

y th

e op

erat

or o

f the

tran

spor

t inf

rast

ruct

ure.

1.5

Tim

e va

riat

ion

The

need

s fo

r el

ectri

city

are

stro

ngly

var

iabl

e in

tim

e.

They

are

bot

h cy

clic

al (

acco

rdin

g to

wel

l kno

wn

daily

, wee

kly

and

year

ly v

aria

tions

) an

d ra

ndom

(for

exa

mpl

e be

caus

e of

cha

nges

in te

mpe

ratu

re).

The

re o

ccur

oth

er

type

s of

tim

e va

riabi

lity

on th

e ge

nera

tion

side

, due

to th

e sc

arci

ty o

f hyd

ro

reso

urce

s, th

e flu

ctua

tions

of

fuel

pric

es a

nd t

he a

vaila

bilit

y of

pla

nts

(mai

nten

ance

and

rep

airin

g).

In

cont

rast

, th

e tra

nspo

rt in

fras

truct

ure

is

alm

ost

fixed

for

the

med

ium

run

. Th

e op

timis

atio

n of

a s

trong

ly v

aria

ble

wel

fare

func

tion

unde

r inv

aria

ble

trans

port

cons

train

ts o

bvio

usly

resu

lts in

a

cont

inuo

usly

var

ying

opt

imal

allo

catio

n an

d, c

onse

quen

tly, a

con

tinuo

usly

va

ryin

g m

argi

nal v

alue

of e

nerg

y at

eac

h no

de.

Dur

ing

low

act

ivity

per

iods

, th

e ca

paci

ty o

f th

e lin

es a

nd t

rans

form

ers

is n

ot b

indi

ng.

The

nod

al

valu

atio

ns d

iffer

onl

y by

the

mar

gina

l val

ue o

f lo

sses

, whi

ch a

re r

athe

r lo

w

sinc

e th

e le

vel o

f con

sum

ptio

n is

low

. In

con

trast

, dur

ing

peak

per

iods

, the

di

ffer

ence

bet

wee

n no

dal v

alua

tions

is v

ery

high

. N

ote

that

the

last

ass

ertio

n is

not

alw

ays

true.

In

som

e ci

rcum

stan

ces,

the

diff

eren

ce b

etw

een

peak

load

Page 11: Chapter 4idei.fr/sites/default/files/medias/doc/by/crampes/cramp... · 2014. 6. 12. · Box 4-2 for some illustrations of network topologies). This explains why, in the optimal dispatch,

Cos

t rec

over

y an

d sh

ort-r

un e

ffici

ency

11

5

and

off-

peak

loa

d ca

n be

larg

er a

t an

exp

ortin

g no

de t

han

at a

n im

porti

ng

node

. In

that

cas

e th

e el

ectri

city

flow

ing

on th

e lin

e at

pea

k pe

riods

can

be

smal

ler

than

at

off-

peak

per

iods

and

the

nod

al d

iffer

ence

in

elec

trici

ty

valu

atio

n ca

n be

hig

her a

t off

-pea

k pe

riods

.

2.

NO

DA

L PR

ICE

S

2.1

Ene

rgy

pric

es

The

first

-bes

t allo

catio

n ca

n be

dec

entra

lised

by

mea

ns o

f pr

ices

whi

ch

refle

ct t

he m

argi

nal

valu

e of

ele

ctric

ity.

It

mea

ns t

hat

the

first

-bes

t ge

nera

tion

leve

ls a

nd c

onsu

mpt

ion

leve

ls w

ould

be

free

ly c

hose

n by

in

divi

dual

gen

erat

ors

and

indi

vidu

al c

onsu

mer

s if

they

wer

e fa

cing

pric

es

equa

l to

the

mar

gina

l va

lue

of e

lect

ricity

in

the

optim

al a

lloca

tion.

A

nd

sinc

e, a

s sh

own

form

erly

, mar

gina

l val

uatio

ns c

hang

e fr

om o

ne n

ode

to th

e ot

her

(and

fro

m ti

me

to ti

me)

, the

dec

entra

lisat

ion

of f

irst-b

est n

eces

sita

tes

noda

l pr

ices

. I

f pe

rfec

t co

mpe

titio

n m

echa

nism

s pr

evai

l at

eac

h no

de

(rou

ghly

sai

d, if

ther

e ex

ists

a la

rge

num

ber o

f sm

all b

uyer

s an

d su

pplie

rs o

f el

ectri

city

at

each

nod

e w

ho b

ehav

e as

pric

e-ta

kers

), th

e eq

uilib

rium

will

na

tura

lly d

eter

min

e en

ergy

pric

es e

qual

to

the

mar

gina

l va

lues

obt

aine

d fo

rmer

ly.

Con

sequ

ently

, or

gani

zing

com

petit

ion

is a

goo

d w

ay t

o re

ach

effic

ienc

y.

How

ever

, eve

n w

hen

the

num

ber

of g

ener

ator

s an

d th

e nu

mbe

r co

nsum

ers

are

reas

onab

ly l

arge

, th

e de

cent

ralis

atio

n of

dec

isio

ns r

equi

res

som

e pu

blic

inte

rven

tion,

for

exa

mpl

e to

org

aniz

e th

e m

atch

ing

of d

eman

d an

d su

pply

in w

hole

sale

mar

kets

. In

con

trast

, if

ther

e is

onl

y a

smal

l nu

mbe

r of

lar

ge s

uppl

iers

and

/or

buye

rs,

the

equi

libriu

m p

rice

wou

ld r

efle

ct t

he m

arke

t po

wer

of

thes

e ag

ents

.7 To

impl

emen

t the

firs

t-bes

t, no

dal p

rices

are

to b

e re

gula

ted

by th

e go

vern

men

t or

, m

ore

prec

isel

y, b

y a

regu

latio

n en

tity

or a

n an

titru

st

auth

ority

.

In a

ny c

ase,

whe

n co

nsum

ers

at n

ode

i can

buy

ele

ctric

ity a

nyw

here

and

pa

y el

ectri

city

com

ing

from

any

nod

e j a

t a p

rice

ip e

qual

to

the

optim

al

mar

gina

l val

uatio

n of

nod

e i,

they

con

sum

e th

e op

timal

qua

ntity

. A

nd w

hen

gene

rato

rs a

t nod

e i a

re a

utho

rised

to s

ell e

lect

ricity

any

whe

re a

nd re

ceiv

e p i

fo

r any

kW

h so

ld to

any

nod

e j,

they

pro

duce

the

optim

al q

uant

ity.8

2.2

Tra

nspo

rt p

rice

s

For

a gi

ven

trans

actio

n, w

hen

the

buye

r an

d th

e se

ller

are

loca

ted

at th

e sa

me

node

i, th

ey tr

ansa

ct a

t the

sam

e pr

ice

p i th

at c

over

s on

ly g

ener

atio

n

Page 12: Chapter 4idei.fr/sites/default/files/medias/doc/by/crampes/cramp... · 2014. 6. 12. · Box 4-2 for some illustrations of network topologies). This explains why, in the optimal dispatch,

116

Tran

spor

t pri

cing

of e

lect

rici

ty n

etw

orks

cost

s. If

the

selle

r is

at n

ode

i and

the

buye

r at n

ode

j, th

e fo

rmer

rece

ives

pi

and

the

latte

r pa

ys p

j. T

he p

robl

em i

s to

kno

w w

hat

to d

o w

ith t

he

diff

eren

ce.

To g

ive

an a

nsw

er, n

ote

that

ano

ther

way

to d

ecen

tralis

e th

e fir

st-b

est i

s to

dis

tingu

ish

the

pric

e of

ene

rgy

and

the

pric

e of

tran

spor

t. W

hen

ther

e ar

e tw

o se

para

te b

ills,

the

pric

e fo

r tra

nspo

rting

1 M

Wh

from

nod

e i t

o no

de j

is

to b

e t ij=

pj -

pi.

As

a m

atte

r of

fac

t, th

e co

nsum

er a

t no

de j

mus

t be

in

diff

eren

t bet

wee

n bu

ying

its

ener

gy a

t nod

e j a

t pric

e p j

on

the

one

hand

an

d, o

n th

e ot

her h

and,

buy

ing

it at

nod

e i a

t pric

e p i

and

then

pay

ing

t ij fo

r its

tra

nspo

rt to

nod

e j.

Sim

ilarly

, the

gen

erat

or a

t no

de i

mus

t be

ind

iffer

ent

betw

een

selli

ng it

s en

ergy

at n

ode

i at p

rice

p i o

n on

e ha

nd a

nd, o

n th

e ot

her

hand

, sel

ling

it at

nod

e j a

t pric

e p j

whi

le p

ayin

g t ij

for i

ts tr

ansp

ort t

o no

de j.

C

onse

quen

tly, w

e ca

n co

nclu

de th

at, f

rom

the

poin

t of

view

of

cons

umer

s an

d pr

oduc

ers,

the

diff

eren

ce t i

j= p

j - p

i re

ally

app

ears

as

a tra

nspo

rt fe

e. F

or

this

reas

on, i

t is n

atur

al to

pay

it to

the

oper

ator

of t

he tr

ansp

ort i

nfra

stru

ctur

e ev

en if

the

cost

s cov

ered

are

on

the

user

s' si

de.

N

ote

that

if th

e ne

t flo

w o

f ene

rgy

is fr

om i

(the

expo

rting

nod

e) to

j (th

e im

porti

ng n

ode)

, any

indi

vidu

al tr

ansa

ctio

n fr

om i

to j

incr

ease

s co

nges

tion

and

mus

t be

char

ged

t ij >

0. I

n co

ntra

st, a

ssum

e p j

< p

i so

that

nod

e j i

s a

net

expo

rter

to n

ode

i an

d th

ere

occu

rs a

n in

divi

dual

tra

nsac

tion

betw

een

a ge

nera

tor

loca

ted

at i

and

a co

nsum

er lo

cate

d at

j.

How

can

it b

e po

ssib

le

sinc

e th

e ge

nera

tor

rece

ives

mor

e th

an w

hat

is p

aid

by t

he c

onsu

mer

? B

ecau

se t

he i

ndiv

idua

l tra

nsac

tion

crea

tes

a co

unte

r-flo

w,

it al

levi

ates

the

co

nges

tion

on t

rans

port

lines

. C

onse

quen

tly,

the

trans

port

pric

ing

syst

em

shou

ld p

rom

ote

this

type

of

trans

actio

n by

rew

ardi

ng th

e pa

rties

inst

ead

of

char

ging

them

for t

rans

port.

And

this

act

ually

occ

urs w

ith n

odal

pric

es si

nce

the

trans

port

of 1

MW

h fr

om i

to j

wou

ld th

en b

e ch

arge

d t ij=

pj -

pi <

0. I

n ot

her w

ords

, the

cou

nter

-flo

w tr

ansa

ctio

ns w

ould

be

enco

urag

ed.

The

tota

l val

ue o

f th

e flo

ws

of e

nerg

y us

ing

the

trans

port

pric

es d

eriv

ed

from

nod

al p

rices

is th

e “m

erch

andi

zing

sur

plus

”. E

ven

if co

unte

r-flo

ws

are

to b

e re

war

ded,

the

mer

chan

dizi

ng s

urpl

us i

s ob

viou

sly

posi

tive9 f

or t

he

follo

win

g re

ason

. R

ecal

l th

at n

odal

pric

es r

efle

ct o

ptim

al v

alua

tions

of

ener

gy.

Con

sequ

ently

, the

mer

chan

dizi

ng s

urpl

us is

mad

e of

two

elem

ents

. Th

e fir

st p

iece

is

a ne

t su

rplu

s du

e to

the

spe

cific

fun

ctio

nal

form

of

the

loss

es o

f ene

rgy

and

thei

r cov

erag

e by

a m

argi

nal p

ricin

g ru

le.

Act

ually

, the

ge

nera

tor

mus

t be

paid

for

the

elec

trici

ty h

e pr

oduc

es, e

ven

if it

cann

ot b

e co

nsum

ed.

But

bec

ause

los

ses

incr

ease

with

the

squ

are

of t

he l

oad,

the

ir m

argi

nal

valu

e is

tw

ice

thei

r av

erag

e va

lue.

Con

sequ

ently

, af

ter

com

pens

atin

g th

e ge

nera

tor f

or it

s los

ses,

it re

mai

ns a

n eq

ual s

um th

at c

an b

e us

ed to

pay

for t

rans

port.

The

sec

ond

part

is th

e co

st o

f con

gest

ion,

and

its

over

all v

alue

can

not b

e ne

gativ

e. E

ither

som

e lin

es o

r no

des

are

cong

este

d an

d th

en t

he s

urpl

us i

s po

sitiv

e, o

r th

ere

is n

o co

nges

tion

and

then

the

Page 13: Chapter 4idei.fr/sites/default/files/medias/doc/by/crampes/cramp... · 2014. 6. 12. · Box 4-2 for some illustrations of network topologies). This explains why, in the optimal dispatch,

Cos

t rec

over

y an

d sh

ort-r

un e

ffici

ency

11

7

surp

lus

is z

ero.

B

y its

ver

y de

finiti

on, t

he m

erch

andi

sing

sur

plus

doe

s no

t co

rres

pond

to th

e pa

ymen

t of c

osts

incu

rred

by

the

oper

ator

. Th

e tra

nspo

rt pr

ices

cal

cula

ted

usin

g no

dal p

rices

var

y w

ith th

e da

te, t

he

initi

al n

ode

and

the

term

inal

nod

e of

the

tra

nsac

tion.

A

ll th

e us

ers

who

tra

nsac

t at t

he s

ame

date

, the

sam

e in

itial

nod

e an

d th

e sa

me

term

inal

nod

e sh

ould

pay

(or

sho

uld

be p

aid

for

coun

ter-

flow

s) t

he s

ame

unit

pric

e.

Ther

efor

e, w

e ca

n as

sert

that

the

nod

al p

rice

syst

em i

s ob

ject

ivel

y no

n-di

scrim

inat

ory

(see

Cha

pter

1).

3.

CO

NST

RA

INE

D A

ND

UN

CO

NST

RA

INE

D

PRIC

ING

3.1

Eff

icie

ncy

conc

ern

vs.

fund

-rai

sing

con

cern

Firs

t-bes

t pric

ing

allo

ws

payi

ng fo

r all

the

cost

s of

an

indu

stry

whe

n th

e eq

uipm

ent i

s op

timal

ly d

esig

ned

and

ther

e ar

e no

incr

easi

ng re

turn

s to

sca

le

in th

e lo

ng ru

n. I

n th

is h

ypot

hetic

al s

ituat

ion,

on

the

one

hand

the

shor

t-run

m

argi

nal

cost

and

the

lon

g-ru

n m

argi

nal

cost

are

equ

al a

nd,

on t

he o

ther

ha

nd,

the

long

-run

mar

gina

l co

st i

s at

lea

st e

qual

to

the

long

-run

ave

rage

co

st.

Con

sequ

ently

, a p

rice

equa

l to

mar

gina

l cos

t can

not b

e be

low

the

long

-ru

n av

erag

e co

st.

B

ut th

ese

optim

al c

ondi

tions

are

pra

ctic

ally

impo

ssib

le to

mee

t in

actu

al

netw

orks

, bec

ause

of t

he fo

llow

ing

reas

ons:10

(i)

dis

crep

ancy

bet

wee

n lo

ng-r

un a

nd sh

ort-r

un n

eeds

Th

e op

timal

dyn

amic

exp

ansi

on o

f th

e tra

nspo

rt in

fras

truct

ure

does

not

am

ount

to

the

mer

e co

nnec

tion

of m

any

stat

ic p

lans

. M

ost

netw

ork

faci

litie

s ar

e bu

ilt fo

r 20

or 3

0 ye

ars

and

it is

unl

ikel

y fo

r any

one

of t

he

faci

litie

s to

hav

e ex

actly

the

opt

imal

sta

tic c

apac

ity f

or a

giv

en y

ear.

A

dditi

onal

ly, d

evia

tions

fro

m t

he o

ptim

al p

lan

are

due

to e

rror

s in

the

fo

reca

st o

f dem

and

and

gene

ratio

n co

sts.

(ii) t

echn

ical

non

con

vexi

ties

For

trans

port

faci

litie

s, de

cisi

ons

are

disc

rete

op

tions

ra

ther

th

an

cont

inuo

us v

aria

bles

. Fo

r a g

iven

rein

forc

emen

t, th

e fe

asib

le s

et is

ver

y sm

all,

e.g.

, lin

es o

f 220

kV

or 4

00 k

V.

Con

sequ

ently

, as

com

pare

d w

ith

the

“mar

gina

l opt

imum

”, th

e re

sult

of p

rogr

amm

ing

with

inte

ger n

umbe

r w

ill b

e an

app

aren

tly e

ither

ove

rsiz

ed o

r un

ders

ized

net

wor

k, w

ith a

na

tura

l te

nden

cy

tow

ards

ov

ersi

zing

w

hen

dem

and

is

expe

cted

to

in

crea

se.

Page 14: Chapter 4idei.fr/sites/default/files/medias/doc/by/crampes/cramp... · 2014. 6. 12. · Box 4-2 for some illustrations of network topologies). This explains why, in the optimal dispatch,

118

Tran

spor

t pri

cing

of e

lect

rici

ty n

etw

orks

The

tend

ency

tow

ards

ove

r in

vest

men

t is

int

ensi

fied

beca

use

of s

cale

ec

onom

ies:

the

lar

ger

the

capa

city

of

the

inve

stm

ent

the

smal

ler

the

capi

tal c

ost p

er u

nit o

f cap

acity

. (ii

i) ad

ditio

nal c

onst

rain

ts

Net

wor

k ex

pans

ion

plan

ning

is

subj

ect

to r

elia

bilit

y co

nstra

ints

whi

ch

crea

te th

e ne

ed f

or e

xtra

cap

acity

with

res

pect

to th

e op

timal

pla

n un

der

stric

t ec

onom

ic t

erm

s. F

inan

cial

, en

viro

nmen

tal,

tech

nica

l or

eve

n po

litic

al re

stric

tions

are

als

o im

pose

d on

the

expa

nsio

n of

the

netw

ork

in

gene

ral o

r of a

giv

en c

orrid

or.

W

e co

nclu

de t

hat,

beca

use

the

infr

astru

ctur

e is

ove

rsiz

ed a

s co

mpa

red

with

sho

rt-ru

n ne

eds

and

beca

use

ther

e ex

ist

long

-run

eco

nom

ies

of s

cale

, co

nges

tion

rent

s ca

lcul

ated

on

the

basi

s of

sho

rt-ru

n m

argi

nal c

osts

will

be

rath

er lo

w, a

nd th

e re

nts

from

loss

es w

ill n

ot b

e hi

gh e

noug

h to

cov

er a

ll th

e fix

ed c

osts

incu

rred

by

the

grid

ope

rato

r.

3.2

Bud

get c

onst

rain

t

Seve

ral s

olut

ions

are

ava

ilabl

e to

bal

ance

the

budg

et o

f the

tran

spor

t firm

. If

pub

lic s

ubsi

dies

are

allo

wed

, th

ey c

an b

e us

ed t

o pa

y th

e di

ffer

ence

be

twee

n th

e m

erch

andi

sing

sur

plus

and

the

fixed

cos

ts o

f the

infr

astru

ctur

e.

But

the

taxe

s lev

ied

to fu

nd th

e su

bsid

ies c

reat

e ec

onom

ic d

isto

rtion

s and

the

resu

lting

allo

catio

n of

con

sum

ptio

n an

d pr

oduc

tion

cann

ot b

e th

e fir

st-b

est.

B

ecau

se o

f th

e co

st o

f pu

blic

fun

ds,

we

can

only

rea

ch a

sec

ond

best

. A

dditi

onal

ly,

with

in t

he E

urop

ean

liber

alis

ed f

ram

ewor

k, t

he S

tate

-aid

so

lutio

n is

not

aut

horis

ed.

Ther

efor

e, p

rices

are

to b

e ad

just

ed to

cov

er a

ll th

e co

sts.

Whe

n co

nstra

ined

to

bala

nce

the

budg

et o

f th

e tra

nspo

rt op

erat

or,

the

optim

al ta

riffs

will

dep

art f

rom

firs

t-bes

t if s

peci

fic re

stric

tions

are

impo

sed

in th

eir c

ompu

tatio

n. W

hen

any

func

tiona

l for

m c

an b

e im

plem

ente

d, th

at is

w

hen

the

bill

is n

ot n

eces

saril

y pr

opor

tiona

l to

the

quan

tity

of e

nerg

y th

at is

tra

nspo

rted

(non

lin

ear

pric

es),

the

first

-bes

t al

loca

tion

rem

ains

fea

sibl

e.11

A

mon

g no

n-lin

ear

tarif

fs w

e w

ill o

nly

disc

uss

two-

part

pric

es.

On

the

cont

rary

, if p

rices

are

rest

ricte

d to

be

linea

r, th

e be

st o

nes a

re R

amse

y pr

ices

, an

d th

e re

sulti

ng a

lloca

tion

of g

ener

atio

n an

d co

nsum

ptio

n w

ill o

nly

be a

se

cond

bes

t. W

e fir

st c

onsi

der t

he c

ase

whe

re o

nly

linea

r pric

es a

re fe

asib

le.

3.3

Seco

nd-b

est l

inea

r pr

ices

Ram

sey

pric

es a

re th

e be

st li

near

pric

es w

hen

the

bala

ncin

g of

bud

get i

s m

anda

tory

. Th

ey c

over

mar

gina

l cos

t (co

nges

tion

and

loss

es) p

lus

a m

argi

n co

mpu

ted

to p

ay f

or t

he f

ixed

cos

ts.

In

met

hods

for

pric

ing

appl

ied

by

prac

titio

ners

, fix

ed c

osts

are

allo

cate

d ac

cord

ing

to so

me

prop

ortio

nalit

y ru

le

Page 15: Chapter 4idei.fr/sites/default/files/medias/doc/by/crampes/cramp... · 2014. 6. 12. · Box 4-2 for some illustrations of network topologies). This explains why, in the optimal dispatch,

Cos

t rec

over

y an

d sh

ort-r

un e

ffici

ency

11

9

(for

exa

mpl

e th

e co

st s

hare

pai

d by

age

nt i

whe

n co

nsum

ing

the

quan

tity

iq

is:

/i

jj

qq

, or

()/

()

ii

jj

jC

qC

q,

whe

re

()

ii

Cq

is th

e di

rect

cos

t to

prod

uce

iq, o

r an

y ot

her

ratio

12).

In

Ram

sey

pric

es, o

ne ta

kes i

nto

acco

unt t

he re

actio

n of

the

infr

astru

ctur

e us

ers

whe

n th

ey a

re b

illed

or

paid

. P

ricin

g ab

ove

mar

gina

l co

st i

s in

effic

ient

be

caus

e it

prov

okes

a d

ecre

ase

in c

onsu

mpt

ion.

Sim

ilarly

, whe

n a

prod

ucer

is

pai

d le

ss th

an m

argi

nal u

tility

, he

decr

ease

s his

pro

duct

ion.

Con

sequ

ently

, th

e ai

m o

f sec

ond-

best

pric

ing

is to

lim

it th

is b

ias

in q

uant

ities

pro

duce

d an

d co

nsum

ed.

It is

eas

y to

und

erst

and

that

, if

som

e co

nsum

ers

reac

t to

pric

e in

crea

ses

less

than

oth

ers,

it is

opt

imal

to m

ake

them

pay

mor

e th

an m

ore

reac

tive

cons

umer

s. A

nd s

ymm

etric

ally

, if

som

e ge

nera

tors

dec

reas

e th

eir

outp

ut m

ore

than

oth

er p

rodu

cers

whe

n th

e se

lling

pric

e is

dec

reas

ed, i

t is

optim

al to

mod

ify th

eir s

ellin

g pr

ice

less

. Th

is e

xpla

ins

that

Ram

sey

pric

es

are

inve

rsel

y re

late

d to

the

ela

stic

ity o

f de

man

d an

d to

the

ela

stic

ity o

f su

pply

. A

s a

cons

eque

nce,

the

y ar

e di

scrim

inat

ory:

eac

h se

gmen

t of

use

rs t

hat

can

be is

olat

ed f

rom

the

othe

rs w

ill p

ay o

r w

ill b

e pa

id a

diff

eren

t Ram

sey

noda

l pric

e. I

n th

e si

mpl

e ca

se w

here

, at e

ach

node

, one

can

not d

istin

guis

h be

twee

n di

ffer

ent

type

s of

con

sum

ers

and

betw

een

diff

eren

t ty

pes

of

prod

ucer

s, it

rem

ains

true

that

one

can

dis

tingu

ish

a gr

oup

of c

onsu

mer

s on

on

e si

de a

nd a

gro

up o

f pr

oduc

ers

one

the

othe

r si

de.

Con

sequ

ently

the

R

amse

y de

man

d pr

ice

w ip w

ill b

e di

ffer

ent

from

the

Ram

sey

supp

ly p

rice

g ip:

beca

use

of t

he n

eed

to l

evy

fund

s to

cov

er f

ixed

cos

ts,

we

lose

the

pr

oper

ty o

f pric

e un

ique

ness

at o

ne n

ode.

W

hen

ener

gy a

nd tr

ansp

ort o

f ene

rgy

from

i to

j ar

e bi

lled

sepa

rate

ly, t

he

equi

libriu

m c

ondi

tion

is t

hat

payi

ng

w jp t

o co

nsum

e at

nod

e j

is t

o be

eq

uiva

lent

to b

uyin

g at

pric

e g ip a

t nod

e i a

nd tr

ansp

ortin

g to

war

ds j,

that

is

gw j

iij

pp

t.

It re

sults

that

the

Ram

sey

pric

e fo

r tra

nspo

rt g

wij

ji

tp

p is

th

e w

eigh

ted

sum

of t

wo

elem

ents

: (i)

the

cost

in te

rms

of e

ffic

ienc

y du

e to

en

ergy

loss

es a

nd c

onge

stio

n w

e al

read

y ha

d in

Sec

tion

1; a

nd (i

i) th

e co

st in

te

rms

of e

ffic

ienc

y du

e to

the

bud

get

cons

train

t. T

he f

irst

part

can

be

posi

tive

or n

egat

ive

depe

ndin

g on

the

dire

ctio

n of

indi

vidu

al o

pera

tions

as

com

pare

d w

ith th

e ne

t flo

w b

etw

een

i and

j. T

he s

econ

d pa

rt is

nec

essa

rily

posi

tive:

it

depe

nds

on t

he e

last

icity

of

the

ener

gy s

uppl

y fu

nctio

n at

the

in

ject

ion

node

and

on

the

elas

ticity

of

the

ener

gy d

eman

d fu

nctio

n at

the

w

ithdr

awal

nod

e. T

he s

mal

ler t

hese

ela

stic

ities

, the

hig

her t

he R

amse

y pr

ice

for

trans

port

betw

een

i an

d j,

wha

teve

r th

e co

sts

of o

hmic

los

ses

and

cong

estio

n.

Page 16: Chapter 4idei.fr/sites/default/files/medias/doc/by/crampes/cramp... · 2014. 6. 12. · Box 4-2 for some illustrations of network topologies). This explains why, in the optimal dispatch,

120

Tran

spor

t pri

cing

of e

lect

rici

ty n

etw

orks

Not

e th

at R

amse

y pr

icin

g co

mm

ands

a g

loba

l al

loca

tion

of f

ixed

cos

ts

base

d on

dem

and

elas

ticity

, but

tran

spar

ency

and

pol

itica

l arg

umen

ts im

pose

so

me

frag

men

tatio

n of

fix

ed c

osts

. A

ctua

lly, t

he o

blig

atio

n to

bal

ance

the

budg

et c

an b

e so

lved

in m

any

diff

eren

t way

s. O

ne p

ossi

bilit

y is

to im

pose

a

glob

al b

udge

t co

nstra

int

to t

he t

rans

port

oper

ator

. I

n th

is c

ase,

Ram

sey

pric

es a

lloca

te a

ll th

e fix

ed c

osts

of

infr

astru

ctur

e to

all

user

s. L

ines

and

tra

nsfo

rmer

s in

low

vol

tage

are

pai

d by

all

user

s, in

clud

ing

larg

e co

nsum

ers

and

gene

rato

rs w

ho u

se o

nly

high

or

med

ium

vol

tage

equ

ipm

ent.

Ano

ther

so

lutio

n is

to e

xem

pt la

rge

user

s fr

om p

ayin

g fo

r th

e fix

ed c

ost o

f th

e lo

w

volta

ge s

yste

m.

In

this

cas

e, t

he o

ptim

al a

lloca

tion

of c

onsu

mpt

ion

and

gene

ratio

n am

ong

node

s an

d th

e re

sulti

ng f

low

s on

line

s is

to b

e ca

lcul

ated

su

bjec

t to

seve

ral b

udge

t con

stra

ints

. A

s exp

lain

ed in

Box

4-1

, the

larg

er th

e nu

mbe

r of

con

stra

ints

, the

hig

her

the

effic

ienc

y lo

ss.

It m

eans

tha

t, w

hen

pric

e di

scrim

inat

ion

is a

llow

ed, d

isco

mpo

sing

the

cos

t re

cove

ry c

onst

rain

t in

to

seve

ral

user

-targ

eted

co

nstra

ints

ca

nnot

be

ju

stifi

ed

in

term

s of

ef

ficie

ncy.

3.4

Tw

o-pa

rt ta

riff

s

The

draw

back

s of

Ram

sey

pric

es a

re o

bvio

us:

(i) a

lar

ge q

uant

ity o

f in

form

atio

n is

nec

essa

ry t

o co

mpu

te t

he e

last

iciti

es;

(ii) d

iscr

imin

atio

n is

fo

rbid

den

by l

aw, (

iii) t

he r

esul

ting

quan

titie

s de

viat

e fr

om f

irst

best

. Fo

r th

ese

reas

ons,

mul

ti-pa

rt pr

ices

can

be

pref

erre

d.

In th

e si

mpl

e ca

se o

f tw

o-pa

rt pr

ices

, the

con

sum

er p

ays

a fix

ed a

mou

nt

of m

oney

inde

pend

ent o

f the

qua

ntity

of t

rans

port

he w

ill re

quire

and

, the

n,

he p

ays

a “m

argi

nal

pric

e” f

or e

ach

unit

he w

ants

to

trans

port.

Si

nce

the

mar

gina

l pric

e ca

n be

set e

qual

to th

e m

argi

nal c

ost o

f con

gest

ion

and

loss

es,

this

tar

iff a

llow

s im

plem

entin

g th

e fir

st-b

est

allo

catio

n, p

rovi

ded

that

the

fix

ed p

art o

f the

tarif

f doe

s no

t exc

lude

any

use

r with

a m

argi

nal w

illin

gnes

s to

pay

hig

her t

han

the

mar

gina

l cos

t. In

eff

ect,

the

risk

with

one

uni

que

two-

part

pric

e is

tha

t th

e fix

ed f

ee m

ust

be h

igh

enou

gh t

o co

ver

all

the

fixed

co

sts,

so th

at u

sers

with

low

inco

me

will

not

be

able

to p

ay f

or it

. If

thes

e lo

w-in

com

e us

ers

can

be e

asily

ide

ntifi

ed, a

sol

utio

n is

to

prop

ose

them

a

spec

ific

two-

part

pric

e, fo

r ins

tanc

e w

ith a

zer

o fix

ed p

art.

The

flaw

is th

at

we

are

back

to d

iscr

imin

atio

n. E

ven

a si

ngle

two-

part

pric

e is

dis

crim

inat

ory

sinc

e th

e un

it pr

ice

paid

is d

ecre

asin

g w

ith th

e qu

antit

y co

nsum

ed.

Inso

far

as tw

o-pa

rt pr

ices

are

acc

epta

ble

on le

gal g

roun

ds, i

t is

effic

ient

to p

ropo

se

to u

sers

not

one

but

a w

hole

set

of t

wo-

part

pric

es, l

ettin

g ea

ch u

ser c

hoos

e w

ithin

the

men

u th

e ta

riff

he p

refe

rs.

It is

a v

ery

effic

ient

way

to

colle

ct

fund

s fo

r fix

ed c

osts

cov

erin

g.

It is

als

o a

good

way

to o

rgan

ize

“sec

ond

degr

ee p

rice

disc

rimin

atio

n” w

hen

the

pric

e m

aker

lack

s in

form

atio

n ab

out

the

will

ingn

ess t

o pa

y of

use

rs.13

Page 17: Chapter 4idei.fr/sites/default/files/medias/doc/by/crampes/cramp... · 2014. 6. 12. · Box 4-2 for some illustrations of network topologies). This explains why, in the optimal dispatch,

Cos

t rec

over

y an

d sh

ort-r

un e

ffici

ency

12

1

App

lyin

g th

ese

prin

cipl

es to

the

trans

port

of e

nerg

y, th

e ta

riff t

o tra

nspo

rt a

give

n qu

antit

y fo

rm n

ode

i to

node

j sh

ould

hav

e a

first

par

t pro

porti

onal

to

the

quan

tity

(whe

re t

he p

ropo

rtion

ality

coe

ffic

ient

is

the

mar

gina

l co

st o

f co

nges

tion

and

loss

es)

and

a se

cond

par

t, fix

ed,

com

pute

d af

ter

the

fixed

pa

rts o

f th

e en

ergy

tarif

fs f

or c

onsu

mer

s at

nod

e j a

nd g

ener

ator

s at

nod

e i.

Th

e fix

ed p

art o

f the

tarif

fs fo

r ene

rgy

mus

t be

adju

sted

not

onl

y to

pay

for

the

fixed

cos

ts o

f ge

nera

tors

but

als

o to

pay

for

the

fixed

cos

ts o

f th

e gr

id

oper

ator

. B

y de

finiti

on, t

he fi

xed

part

of tw

o-pa

rt ta

riffs

doe

s not

dis

tort

effic

ienc

y.

Ther

efor

e, th

e ta

riff

mak

er h

as s

ome

degr

ees

of f

reed

om to

cal

cula

te th

em.

He

mai

nly

has

to a

void

the

exc

lusi

on o

f so

me

agen

ts,

indi

vidu

ally

or

colle

ctiv

ely.

In

elec

trici

ty

trans

port,

th

e m

ain

conc

ern

is

that

la

rge

cons

umer

s ha

ve a

ltern

ativ

e op

portu

nitie

s to

the

use

of th

e gr

id, f

or e

xam

ple

by i

nsta

lling

gen

erat

ion

plan

ts a

t th

eir

loca

tion.

To

pre

vent

thi

s by

-pas

s, se

vera

l tw

o-pa

rt ta

riffs

mus

t be

prop

osed

. Ea

ch it

em in

the

men

u is

tailo

red

so th

at n

o on

e is

exc

lude

d (p

artic

ipat

ion

cons

train

t) an

d ea

ch a

gent

cho

oses

th

e ta

riff

calc

ulat

ed f

or h

im (

self-

sele

ctio

n co

nstra

int).

W

hen

also

tak

ing

into

acc

ount

the

poss

ibili

ty o

f co

aliti

on b

etw

een

seve

ral p

rodu

cers

or

larg

e co

nsum

ers,

or b

etw

een

a pr

oduc

er a

nd o

ne o

f its

clie

nts

in o

rder

to b

ypas

s th

e tra

nspo

rt in

fras

truct

ure,

it is

nec

essa

ry to

hav

e re

cour

se to

coo

pera

tive-

gam

es th

eory

to o

ptim

ally

des

ign

the

fixed

par

t of t

he ta

riff.14

Fi

nally

, no

te t

hat

devi

ces

able

to

cont

inuo

usly

met

er t

he f

low

of

elec

trici

ty o

n a

line

are

very

exp

ensi

ve a

nd,

in m

any

coun

tries

, th

ey a

re

inst

alle

d on

ly o

n th

e hi

gh v

olta

ge g

rid.

Add

ition

ally

, w

hen

avai

labl

e, i

t w

ould

be

very

cos

tly b

oth

at t

he o

pera

tor

leve

l an

d at

the

use

r le

vel

to

proc

ess

all

the

info

rmat

ion

they

col

lect

. A

s a

resu

lt, i

nste

ad o

f se

ndin

g in

voic

es b

ased

on

the

time

prof

ile o

f th

e de

man

d fo

r tra

nspo

rt, i

n m

ost

coun

tries

one

use

s to

cal

cula

te t

he a

ggre

gate

qua

ntity

tra

nspo

rted

durin

g a

give

n pe

riod

(for

exa

mpl

e, o

ne y

ear)

and

to c

harg

e a

unifo

rm m

argi

nal p

rice

of tr

ansp

ort a

ll th

e ye

ar lo

ng.

The

draw

back

is th

at th

e sa

me

tota

l qua

ntity

of

ener

gy c

an b

e tra

nsfe

rred

with

stro

ng t

ime

regu

larit

y or

with

a h

igh

irreg

ular

ity.

And

the

irreg

ular

load

pro

file

is m

uch

mor

e de

man

ding

in te

rms

of tr

ansp

ort c

apac

ity th

an th

e re

gula

r one

. O

ne s

olut

ion

is to

dis

tingu

ish

an

ener

gy c

ompo

nent

and

a c

apac

ity c

ompo

nent

in

the

dem

and

for

trans

port,

ea

ch w

ith a

spe

cific

con

stan

t mar

gina

l pric

e.

But

it r

emai

ns tr

ue th

at th

ese

two

elem

ents

are

to b

e de

sign

ed to

sen

d ac

cura

te s

igna

ls o

f sca

rcity

to u

sers

. A

fixe

d pa

rt re

mai

ns n

eces

sary

to c

over

the

fixed

cos

ts o

f the

infr

astru

ctur

e w

ithou

t dis

torti

ng th

e de

cisi

on o

f the

use

rs.

Page 18: Chapter 4idei.fr/sites/default/files/medias/doc/by/crampes/cramp... · 2014. 6. 12. · Box 4-2 for some illustrations of network topologies). This explains why, in the optimal dispatch,

122

Tran

spor

t pri

cing

of e

lect

rici

ty n

etw

orks

4.

CO

NC

LU

DIN

G C

OM

ME

NT

S

To c

alcu

late

the

tar

iffs

to t

rans

port

elec

trici

ty,

the

trade

-off

bet

wee

n ef

ficie

ncy,

bud

get r

equi

rem

ents

, leg

al c

onst

rain

ts a

nd p

ract

icab

ility

, res

ults

in

a n

on li

near

tarif

f ba

sed

on th

e no

dal p

rices

of

ener

gy a

nd o

n th

e fix

ed

cost

s of t

he in

fras

truct

ure.

The

sim

ples

t is t

wo-

part.

The

tota

l bill

to b

e pa

id

to th

e tra

nspo

rt fir

m is

mad

e of

an

“eff

icie

ncy-

orie

nted

” pa

rt, v

aria

ble

with

th

e qu

antit

y tra

nspo

rted,

and

a “

cost

-rec

over

y” p

art,

inde

pend

ent

of t

he

quan

tity

trans

porte

d.

The

first

par

t is

aim

ed a

t si

gnal

ling

the

cost

of

phys

ical

los

ses

and

the

cost

of

cong

estio

n.

It sh

ould

var

y w

ith t

ime

and

loca

tion

and

shou

ld b

e in

crea

sing

with

the

quan

tity

of e

nerg

y tra

nspo

rted.

15 P

ract

ical

ly, i

t sho

uld

be

a fu

nctio

n of

the

inje

ctio

n an

d w

ithdr

awal

nod

es (n

ot o

n th

e ph

ysic

al p

ath

of

ener

gy s

ince

it

actu

ally

can

not

be c

ontro

lled)

and

on

the

dire

ctio

n of

the

flo

w.

For

prac

tical

rea

sons

, thi

s fir

st p

art o

f th

e ta

riff

can

be d

isco

mpo

sed

into

an

ener

gy te

rm a

nd a

cap

acity

term

, bot

h lin

ear f

unct

ions

, pro

vide

d th

at

the

scar

city

sig

nals

the

y tra

nsm

itted

are

sim

ilar

to t

he o

nes

that

allo

w a

n ef

ficie

nt u

se o

f the

grid

. Th

e fix

ed p

art o

f the

tarif

f is

aim

ed a

t pay

ing

for t

he in

fras

truct

ure

cost

s.

Bec

ause

its

pur

pose

is

pure

ly b

udge

tary

, it

shou

ld n

ot i

nter

fere

with

the

sc

arci

ty s

igna

ls s

ent b

y th

e va

riabl

e pa

rt(s)

. In

par

ticul

ar, w

hen

the

trans

port

tarif

f dis

tingu

ishe

s bet

wee

n an

ene

rgy

com

pone

nt a

nd a

cap

acity

com

pone

nt,

the

latte

r sho

uld

not b

e vi

ewed

as t

he fi

xed

part

of th

e ta

riff.

In m

ost c

ount

ries,

actu

al tr

ansp

ort t

ariff

s ar

e tw

o-pa

rt bu

t the

way

they

ba

lanc

e th

e va

riabl

e pa

rt(s)

and

the

fixed

par

t is v

ery

hete

roge

neou

s.16

APP

EN

DIX

MO

DEL

SET

TIN

G

Two

node

s, n

(for

nor

th) a

nd s

(for

sout

h), a

re c

onne

cted

by

a lin

e of

cap

acity

K.

The

unit

cost

of

capa

city

is r

. Th

e qu

antit

y of

ele

ctric

ity tr

ansp

orte

d th

roug

h th

e lin

e is

mea

sure

d in

th

e sa

me

unit

as K

. W

e no

te

ii

Uq

the

utili

ty d

eriv

ed fr

om th

e co

nsum

ptio

n of

a q

uant

ity

iq o

f ele

ctric

ity a

t nod

e i a

nd

ii

Cq

the

cost

of g

ener

atin

g iq a

t nod

e i

(,

)i

ns

.

Let

,,

,,

wg

wg

wg

wg

nn

ss

nn

nn

ss

ss

Wq

qq

qK

Uq

Cq

Uq

Cq

rK s

tand

for t

he

wel

fare

func

tion

of th

is in

dust

ry, w

here

sup

ersc

ript w

sta

nds

for “

with

draw

al”

and

supe

rscr

ipt

g st

ands

for

“ge

nera

tion”

. W

e co

nsid

er o

nly

shor

t-run

dec

isio

ns, t

hat i

s, K

is n

ot a

con

trol

varia

ble.

The

opt

imal

allo

catio

n is

the

solu

tion

to:

Page 19: Chapter 4idei.fr/sites/default/files/medias/doc/by/crampes/cramp... · 2014. 6. 12. · Box 4-2 for some illustrations of network topologies). This explains why, in the optimal dispatch,

Cos

t rec

over

y an

d sh

ort-r

un e

ffici

ency

12

3

,

,,

max

,,

,,

wg

wg

nn

ss

gg

ww

nn

ss

qq

qq

Wq

qq

qK

(1

)

s.t.

g

ww

gn

ns

sq

qq

q

(2)

de

f

gw

nsn

nz

qq

K

(3)

Prov

isio

nally

, we

assu

me

ther

e is

no

loss

, whi

ch is

refle

cted

by

(2).

To

sim

plify

not

atio

ns,

we

assu

me

that

pre

fere

nces

and

cos

ts a

re s

uch

that

the

north

nod

e w

ill a

lway

s ex

port

a ne

t flo

w o

f ele

ctric

ity

nszto

war

ds so

uth.

For

this

reas

on,

0nsz

.

THE

SOC

IAL

CO

ST O

F C

ON

GES

TIO

N

Let

den

ote

the

dual

var

iabl

e as

soci

ated

to c

onst

rain

t (3)

. Th

e fir

st-o

rder

con

ditio

ns th

at

char

acte

rise

the

solu

tion

(,

,,

)go

gow

ow

on

sn

sq

qq

qto

the

abov

e pr

oble

m a

re:

'w

ogo

ss

ss

Uq

Cq

(4

) '

wo

gon

nn

nU

qC

q

(5)

ogo

gos

sn

nC

qC

q

(6)

gogo

wo

wo

ns

ns

qq

qq

(7

)

0,,

.0

ogo

wo

ogo

wo

nn

nn

Kq

qK

qq

. (8

)

By

(4)

and

(5),

at e

ach

node

mar

gina

l ut

ility

and

mar

gina

l co

st a

re t

o be

equ

al.

If t

he

capa

city

of t

he li

ne is

ver

y la

rge,

we

obta

in th

e “g

rid-f

ree”

allo

catio

n. I

n th

is c

ase,

o nsz

K

so t

hat

0o

by

the

com

plem

enta

ry s

lack

ness

con

ditio

n (8

). T

hen,

by

(6),

the

mar

gina

l va

luat

ion

of e

lect

ricity

is th

e sa

me

at b

oth

node

s.

The

serie

s of

eq

ualit

ies

''

wo

gow

ogo

ss

ss

nn

nn

Uq

Cq

Uq

Cq

, to

geth

er

with

co

nditi

on (7

), fu

lly d

escr

ibe

the

optim

al a

lloca

tion.

(se

e Fi

gure

4-4

).

impo

rts=

o nsz

expo

rts=

o nsz

' nU

nC

wo

nq

go nq

nq

go sqw

osq

sq

' sU

sC

Nor

th

Sout

h

Figu

re 4

-4.

No

capa

city

con

stra

int,

sam

e va

luat

ion

at b

oth

node

s

Page 20: Chapter 4idei.fr/sites/default/files/medias/doc/by/crampes/cramp... · 2014. 6. 12. · Box 4-2 for some illustrations of network topologies). This explains why, in the optimal dispatch,

124

Tran

spor

t pri

cing

of e

lect

rici

ty n

etw

orks

Con

vers

ely,

if K

is s

mal

l, th

e co

nstra

int (

3) is

bin

ding

and

we

obta

in th

e gr

id-c

onst

rain

ed

allo

catio

n.

The

optim

al p

rodu

ctio

n in

‘no

rth’

is g

iven

by

'go

gon

nn

nU

qK

Cq

and

, in

‘sou

th’,

by

'go

gos

ss

sU

qK

Cq

. N

ow,

the

shad

ow

pric

e '

'o

gogo

gogo

ss

nn

ss

nn

Uq

KU

qK

Cq

Cq

is

st

rictly

pos

itive

. It

sign

als

that

a l

arge

r K

wou

ld a

llow

to

gene

rate

mor

e en

ergy

in ‘

north

’ w

here

the

cost

of o

ne a

dditi

onal

uni

t is s

mal

ler t

han

in ‘s

outh

’ and

to c

onsu

me

mor

e in

‘sou

th’

whe

re th

e ut

ility

of t

his a

dditi

onal

uni

t is l

arge

r tha

n in

‘nor

th’ (

see

Figu

re 4

-5).

o

K K

' nU

nC

wo

nq

go nq

nq

go sqw

osq

sq

' sU

sC

Nor

th

Sout

h

Figu

re 4

-5.

Bin

ding

tran

spor

t cap

acity

, nod

al m

argi

nal v

alua

tions

of e

nerg

y di

verg

e

THE

SOC

IAL

CO

ST O

F LO

SSES

A

ssum

e no

w t

hat

a fr

actio

n of

the

ele

ctric

ity i

njec

ted

at t

he n

orth

nod

e is

los

t be

fore

ar

rivin

g at

the

sout

h no

de.

In th

e pr

ogra

m to

obt

ain

the

first

-bes

t allo

catio

n, w

e m

ust r

epla

ce

(2) w

ith:

wg

gw

wg

ss

nn

ss

qq

qq

Lq

q

(9)

sinc

e tra

nsm

issi

on lo

sses

.

L in

crea

se w

ith th

e po

wer

con

sum

ed b

y th

e lo

ad.

The

first

-or

der c

ondi

tions

rem

ain

(4),

(5) a

nd (8

). B

ut (7

) is

repl

aced

by

(9) w

ith th

e co

nseq

uenc

e th

at

(6) i

s rep

lace

d by

:

1o

gogo

ss

nn

Cq

Cq

L

(10)

whe

re L

stan

ds fo

r mar

gina

l los

ses.

In t

he n

o-co

nges

tion

case

her

eafte

r ill

ustra

ted

by F

igur

e 4-

6 (

0o

), w

e se

e th

at t

he

optim

al a

lloca

tion

sets

:

(1)

gogo

gos

sn

nn

Cq

Cq

LC

q.

This

refle

cts t

hat i

t tak

es (1

+L')

MW

h ge

nera

ted

in ‘n

orth

’ to

mat

ch th

e ne

ed fo

r 1 M

Wh

in

‘sou

th’.

As

com

pare

d w

ith t

he c

ase

whe

re

0,L

the

vol

ume

of t

rade

is

redu

ced

and

Page 21: Chapter 4idei.fr/sites/default/files/medias/doc/by/crampes/cramp... · 2014. 6. 12. · Box 4-2 for some illustrations of network topologies). This explains why, in the optimal dispatch,

Cos

t rec

over

y an

d sh

ort-r

un e

ffici

ency

12

5

cons

umpt

ion

in ‘

sout

h’ i

s sm

alle

r. B

ut g

ener

atio

n in

‘no

rth’

can

be l

arge

r in

ord

er t

o co

mpe

nsat

e fo

r the

loss

es.

'(1

)n

LU

(1)

gon

nC

qL

go

ss

Cq

nq

sq

' nU

sC

' sU

(1)

nL

C

nC

gon

nC

q

wo nq

go nq

go sqwo sq

Figu

re 4

-6.

Opt

imal

allo

catio

n w

ith e

nerg

y lo

sses

TIM

E V

AR

IATI

ON

Fo

r a g

iven

tran

spor

t inf

rast

ruct

ure,

the

optim

al d

ispa

tch

is ti

me

depe

nden

t, fo

llow

ing

the

chan

ges

in c

onsu

mer

s pr

efer

ence

s of

and

in

gene

ratio

n pl

ants

ava

ilabi

lity.

Th

e ob

viou

s co

nseq

uenc

e is

that

nod

al v

alua

tions

of e

nerg

y ar

e al

so ti

me-

depe

nden

t. A

ssum

e th

ere

are

no

loss

es a

nd th

e tra

nspo

rt lin

e of

cap

acity

K is

use

d du

ring

two

perio

ds o

f equ

al d

urat

ion:

a p

eak

perio

d (la

belle

d D

for d

ay) a

nd a

n of

f-pe

ak p

erio

d (la

belle

d N

for n

ight

). T

he p

robl

em is

to

max

imis

e:

,

,[

]w

g

ii

ii

DN

in

sU

qC

qrK

,

subj

ect t

o tw

o co

nstra

ints

of

type

(2)

and

tw

o co

nstra

ints

of

type

(3)

, one

pai

r fo

r ea

ch

perio

d.17

Fro

m th

e op

timal

allo

catio

n, w

e de

rive

the

shad

ow p

rice

of th

e lin

e du

ring

the

day

ogo

goD

sDsD

nDnD

Cq

Cq

an

d th

e sh

adow

pr

ice

of

the

line

durin

g th

e ni

ght

ogo

goN

sNsN

nNnN

Cq

Cq

. A

s the

nee

d fo

r tra

nspo

rtatio

n is

hig

her d

urin

g th

e da

y, w

e ha

ve

oo

DN

, w

here

the

equ

ality

hol

ds o

nly

in t

he c

ase

0o

oD

N,

that

is

whe

n th

e lin

e is

pe

rman

ently

non

-bin

ding

(w

hich

sho

uld

neve

r oc

cur

sinc

e th

e un

it co

st o

f ca

paci

ty i

s po

sitiv

e).

Dep

endi

ng o

n th

e di

scre

panc

y be

twee

n th

e ni

ght

and

day

char

acte

ristic

s of

co

nsum

ptio

n an

d ge

nera

tion,

w

e ca

n ha

ve

eith

er

a pe

rman

ently

bi

ndin

g ca

paci

ty

(0

oo

DN

) or c

onge

stio

n on

ly d

urin

g th

e pe

ak p

erio

d (

0o

oD

N).

In

any

cas

e, b

ecau

se th

e sa

me

line

is u

sed

to sa

tisfy

diff

eren

t nee

ds, i

t has

the

feat

ures

of a

“p

ublic

goo

d”:18

thi

s ex

plai

ns w

hy t

he t

est t

o kn

ow w

heth

er t

he c

apac

ity o

f th

e lin

e is

too

la

rge

or to

o sm

all c

onsi

sts n

ow in

com

parin

g th

e un

it co

st o

f cap

ital r

with

the

sum

(o

oD

N).

IMPL

EMEN

TATI

ON

Th

e fir

st-b

est s

olut

ion

can

be d

ecen

tralis

ed b

y m

eans

of

noda

l pric

es.

At t

he s

outh

nod

e an

y tra

nsac

tion

will

cos

t '

wo

gos

ss

ss

pU

qC

q p

er u

nit t

o a

cons

umer

and

will

retu

rn

sp

Page 22: Chapter 4idei.fr/sites/default/files/medias/doc/by/crampes/cramp... · 2014. 6. 12. · Box 4-2 for some illustrations of network topologies). This explains why, in the optimal dispatch,

126

Tran

spor

t pri

cing

of e

lect

rici

ty n

etw

orks

to a

pro

duce

r, w

hate

ver t

he lo

catio

n of

the

othe

r sid

e in

the

trans

actio

n. S

imila

rly, a

t the

nor

th

node

'

wo

gon

nn

nn

pU

qC

q w

ill b

e ch

arge

d fo

r any

uni

t of c

onsu

mpt

ion

and

will

be

paid

fo

r any

uni

t of p

rodu

ctio

n.

It is

eas

y to

che

ck th

at c

onsu

mer

s who

solv

e:

max

ii

ii

iq

Uq

pq

will

cho

ose

wo

iq, a

nd p

rodu

cers

who

solv

e:

max

ii

ii

iq

pq

Cq

will

cho

ose

go iq a

t ,

in

s.

Ther

efor

e no

dal

pric

es,

eith

er r

esul

ting

from

com

petit

ive

noda

l mar

kets

or i

mpl

emen

ted

by a

regu

latio

n en

tity,

allo

w to

reac

h th

e fir

st-b

est a

lloca

tion.

W

hen

ener

gy a

nd tr

ansp

ort a

re b

illed

sep

arat

ely,

the

equi

libriu

m p

rice

for

trans

port

from

no

rth to

sou

th m

ust b

e su

ch th

at

sn

nsp

pt

; the

n ns

sn

tp

p.

We

conc

lude

that

, whe

n lo

sses

can

be

negl

ecte

d, fr

om (6

) the

firs

t-bes

t tra

nspo

rt pr

ice

is:

oo

nst

(11)

In w

ords

, ab

sent

any

ohm

ic l

osse

s, tra

nspo

rt sh

ould

be

free

if

and

only

if

ther

e is

no

cong

estio

n. W

here

ther

e is

con

gest

ion,

a p

ositi

ve p

rice

is to

be

char

ged.

Sym

met

rical

ly, t

he

equi

libriu

m p

rice

for

trans

port

from

sou

th t

o no

rth m

ust

be s

uch

that

n

ssn

pp

t.

We

dedu

ce t

hat

reve

rse

flow

s sh

ould

be

char

ged

0o

osn

ns

tp

p.

Ind

eed,

as

long

as

0g

ww

gns

nn

ss

zq

qq

q, i

ndiv

idua

l tra

nsac

tions

from

sou

th to

nor

th s

houl

d be

rew

arde

d si

nce

they

redu

ce th

e co

nges

tion

of th

e lin

e.

Sim

ilarly

, ta

king

los

ses

into

acc

ount

, fr

om (1

0) w

e ob

tain

'

'0

oo

nsn

tL

C a

nd

'(

')

0o

osn

nt

LC

. Th

e la

tter i

s on

ce m

ore

just

ified

bec

ause

any

net

inje

ctio

n in

‘sou

th’

decr

ease

s th

e qu

antit

y of

ene

rgy

that

mus

t be

inje

cted

at t

he n

orth

nod

e an

d, c

onse

quen

tly,

decr

ease

s bot

h th

e lo

sses

of e

nerg

y an

d th

e co

nges

tion

on th

e lin

e.

BA

LAN

CIN

G T

HE

BU

DG

ET O

F TH

E TR

AN

SPO

RT

OPE

RA

TOR

Is

thi

s fir

st-b

est

sust

aina

ble

on f

inan

cial

gro

unds

? O

n th

e ge

nera

tor’

s si

de,

if th

ere

are

incr

easi

ng re

turn

s to

scal

e, m

argi

nal-c

ost p

ricin

g is

not

suff

icie

nt to

bre

ak e

ven.

But

sinc

e th

is

repo

rt is

ded

icat

ed to

tran

spor

t, w

e as

sum

e th

at g

ener

ator

s do

not l

ose

mon

ey w

hen

gene

ratio

n is

val

ued

at fi

rst b

est19

.

Wha

t abo

ut th

e re

sour

ces

of th

e gr

id o

pera

tor?

The

mar

ket e

quili

briu

m u

nder

nod

al p

rices

cr

eate

s a “

mer

chan

disi

ng su

rplu

s”:

wo

wo

gogo

ss

nn

ss

nn

MS

pq

pq

pq

pq

.o

n

LC

LK

K

if th

e co

nstra

int i

s bin

ding

'

.o

nns

o nsLC

Lz

z

othe

rwis

e.

Page 23: Chapter 4idei.fr/sites/default/files/medias/doc/by/crampes/cramp... · 2014. 6. 12. · Box 4-2 for some illustrations of network topologies). This explains why, in the optimal dispatch,

Cos

t rec

over

y an

d sh

ort-r

un e

ffici

ency

12

7

This

surp

lus i

s pos

itive

for t

wo

reas

ons:

be

caus

e of

the

cong

estio

n re

nt

oK

whe

n th

e lin

e is

con

gest

ed,

beca

use

loss

es a

re in

crea

sing

with

the

squa

re o

f the

ene

rgy

inje

cted

. It

resu

lts th

at:

2o nsL

Lz

,

so th

at th

e su

rplu

s due

to lo

sses

is

.n

CL

In

sum

mar

y, u

nder

nod

al p

ricin

g th

e di

ffer

ence

MS

betw

een

the

expe

nditu

res o

f cus

tom

ers

and

the

gain

s of

pro

duce

rs i

s po

sitiv

e an

d it

can

be u

sed

to c

over

the

cos

t of

the

tra

nspo

rt op

erat

or.

But

is it

true

that

MS

rK?

If t

he s

ize

of t

he l

ine

coul

d be

opt

imal

ly a

djus

ted

to t

rans

port

need

s, w

e w

ould

hav

e o

r a

nd th

e M

S w

ould

leav

e th

e op

erat

or w

ith a

net

pro

fit e

qual

to th

e va

lue

of lo

sses

(p

rovi

ded

ther

e ar

e no

-incr

easi

ng re

turn

s to

scal

e).

But

for t

echn

ical

, eco

nom

ical

and

pol

itica

l re

ason

s, in

mos

t tra

nspo

rt ne

twor

ks, t

here

is e

xces

s cap

acity

so th

at

or

, and

eve

n 0

o

for a

bsol

ute

exce

ss c

apac

ity.

In m

ost t

rans

port

infr

astru

ctur

e, it

resu

lts th

at th

e m

erch

andi

sing

su

rplu

s doe

s not

cov

er th

e fix

ed c

osts

.

We

succ

essi

vely

con

side

r sec

ond-

best

line

ar p

rices

and

non

line

ar p

rices

.

RA

MSE

Y N

OD

AL

PRIC

ES

To s

impl

ify o

ur a

naly

sis,

we

com

e ba

ck to

the

hypo

thes

is o

f no

loss

es.

Then

, the

pro

blem

is

to

max

imiz

e (1

) w

ith r

espe

ct t

o pr

ices

(,

,,

)w

wg

gn

sn

sp

pp

p s

ubje

ct t

o (2

) an

d (3

), pl

us t

he

addi

tiona

l obl

igat

ion

to b

alan

ce th

e bu

dget

of t

he tr

ansp

ort f

irm, t

hat i

s:

0w

ww

wg

gg

gn

ns

sn

ns

sp

qp

qp

qp

qrK

, (1

2)

know

ing

that

pric

e-ta

king

con

sum

ers

beha

ve o

ptim

ally

'

ww

ii

iU

qp

and

pric

e-ta

king

ge

nera

tors

beh

ave

optim

ally

g

gi

ii

Cq

p a

t eac

h no

de i=

n,s.

D

enot

ing

by

the

mul

tiplie

r as

soci

ated

with

the

bud

get

cons

train

t (1

2),

by

the

m

ultip

lier a

ssoc

iate

d w

ith th

e flo

w c

onst

rain

t (2)

and

by

the

mul

tiplie

r ass

ocia

ted

with

the

capa

city

con

stra

int

(3),

from

the

firs

t-ord

er c

ondi

tions

, it

is s

traig

htfo

rwar

d to

writ

e th

e se

cond

-bes

t pric

es a

s:20

11

1

wg

nn

wg

nn

wg

nn

pp

pp

in

‘nor

th’

11

1

wg

ss

wg

ss

wg

ss

pp

pp

in

‘sou

th’.

Whe

re:

Page 24: Chapter 4idei.fr/sites/default/files/medias/doc/by/crampes/cramp... · 2014. 6. 12. · Box 4-2 for some illustrations of network topologies). This explains why, in the optimal dispatch,

128

Tran

spor

t pri

cing

of e

lect

rici

ty n

etw

orks

/ /

def

ww

ii

w iw

wi

i

qq

pp

is th

e pr

ice

elas

ticity

of d

eman

d an

d

/ /

def

gg

ii

g ig

gi

i

qq

pp

is th

e pr

ice

elas

ticity

of s

uppl

y at

nod

e i.

Bot

h co

nsum

ers

and

gene

rato

rs a

t eac

h no

de h

ave

to c

ontri

bute

to c

over

the

fixed

cos

t of

trans

port.

At n

ode

i we

obse

rve

that

:

()

1

wg

ii

wg

ii

wg

ii

pp

pp

.

It is

no

long

er tr

ue th

at th

ere

exis

ts o

ne u

niqu

e pr

ice

at e

ach

node

. Th

e di

ffer

ence

bet

wee

n de

man

d pr

ice

and

supp

ly p

rice

incr

ease

s w

ith th

e va

lue

of th

e fix

ed c

ost o

f tra

nspo

rt to

be

paid

(w

hich

app

ears

in

wg

ii

pp

thr

ough

the

dua

l va

riabl

e )

and

it de

crea

ses

whe

n th

e de

man

d an

d su

pply

ela

stic

ities

incr

ease

.

RA

MSE

Y T

RA

NSP

OR

T PR

ICES

It

obvi

ousl

y re

mai

ns t

rue

that

the

equ

ilibr

ium

tra

nspo

rtatio

n ch

arge

s ap

plie

d to

the

do

min

ant f

low

(fr

om n

orth

to s

outh

) ar

e su

ch th

at

wg

sn

nsp

pt

. U

sing

the

Ram

sey

noda

l pr

ices

, we

obta

in:

wg

nss

nt

pp

11

wg

sn

wg

sn

pp

(1

3)

As c

ompa

red

with

(11)

, we

obse

rve

two

dist

ortio

ns:

first

, to

the

mar

gina

l co

st o

f co

nges

tion21

we

have

to

add

the

cont

ribut

ion

to t

he

reco

very

of t

he tr

ansp

ort f

ixed

cos

ts;

seco

nd, t

he m

argi

nal c

ost o

f con

gest

ion

is it

self

dist

orte

d.

An

incr

ease

in

K b

oth

incr

ease

s f

or f

inan

cial

rea

sons

and

dec

reas

es

for

tec

hnic

al

reas

ons.

It

conf

irms

that

the

pric

ing

of t

rans

port

mai

nly

appe

ars

as t

he c

over

age

of f

ixed

co

sts,

exce

pt w

hen

the

capa

city

is v

ery

smal

l.

Fina

lly, n

ote

that

for r

ever

se tr

ansa

ctio

ns (f

rom

sout

h to

nor

th) t

he R

amse

y pr

ice

shal

l be:

wg

snn

st

pp

11

wg

ns

wg

ns

pp

(1

4)

It m

eans

that

the

rew

ard

for a

llevi

atin

g lo

sses

and

con

gest

ion

will

pro

babl

y be

mor

e th

an

com

pens

ated

by

the

part

of t

he t

ariff

tha

t pa

rtici

pate

s to

the

cov

erag

e of

fix

ed c

osts

. I

f de

man

d is

muc

h m

ore

elas

tic i

n ‘s

outh

’ th

an i

n ‘n

orth

’ an

d su

pply

muc

h m

ore

elas

tic i

n ‘n

orth

’ th

an in

‘so

uth’

, it c

an e

ven

be th

e ca

se th

at

snns

tt

, des

pite

the

coun

ter-

flow

eff

ect.

Th

is is

bec

ause

fina

ncia

l con

cern

s app

ear a

s dom

inan

t as c

ompa

red

with

eff

icie

ncy

conc

erns

. In

any

cas

e, c

ompa

ring

(13)

and

(14

), w

e ob

serv

e th

at

nssn

tt

onl

y by

coi

ncid

ence

. A

ctua

lly, s

econ

d-be

st p

rices

are

dire

ctio

nal b

oth

for t

echn

ical

and

fina

ncia

l rea

sons

. A

lso,

it

Page 25: Chapter 4idei.fr/sites/default/files/medias/doc/by/crampes/cramp... · 2014. 6. 12. · Box 4-2 for some illustrations of network topologies). This explains why, in the optimal dispatch,

Cos

t rec

over

y an

d sh

ort-r

un e

ffici

ency

12

9

mus

t be

reca

lled

that

in (

13) a

nd (

14)

the

dual

var

iabl

es, t

he n

odal

pric

es a

nd th

e el

astic

ities

ar

e al

l tim

e de

pend

ent.

We

conc

lude

that

Ram

sey

trans

port

pric

es m

ust b

e tim

e de

pend

ent.

TWO

-PA

RT

TAR

IFFS

W

e co

nsid

er n

ow t

he c

ase

whe

re p

rices

are

not

con

stra

ined

to

be l

inea

r. W

e lim

it th

e an

alys

is to

the

two-

part

tarif

f w

here

(

)w

wi

iT

q =

w

ww

ii

ia

qb

is th

e ex

pend

iture

for

con

sum

ing

w iq a

t nod

e i a

nd

()

gg

ii

Tq

=

gg

gi

ii

aq

b is

the

reve

nue

whe

n ge

nera

ting

g iq a

t nod

e i.

The

ob

ligat

ion

to b

alan

ce th

e bu

dget

of t

he tr

ansp

orta

tion

firm

read

s:

()

0w

ww

wg

gg

gw

wg

gn

ns

sn

ns

sn

sn

sa

qa

qa

qa

qb

bb

brK

(1

5)

By

fixin

g '

'(

)(

)w

wo

ggo

ii

ii

ii

aU

qa

Cq

at e

ach

node

, we

obvi

ousl

y im

plem

ent t

he f

irst-

best

allo

catio

n of

con

sum

ptio

n an

d ge

nera

tion

at e

ach

node

. G

iven

thes

e pr

ices

and

giv

en th

e flo

w c

onst

rain

t (2

) (a

ssum

ing

L=0)

, (15

) re

ads

()

()

0w

wg

go

ns

ns

bb

bb

rK

. O

ne

obta

ins a

n in

finite

set o

f sol

utio

ns fo

r the

fixe

d pa

rt of

the

loca

tiona

l ene

rgy

tarif

fs, e

ven

whe

n ta

king

int

o ac

coun

t th

e pa

rtici

patio

n co

nstra

ints

of

cons

umer

s (

)(

)w

ww

wi

ii

ii

Uq

Tq

u a

nd

gene

rato

rs

()

()

gg

gg

ii

ii

iT

qC

qg

, w

here

iu (

resp

ectiv

ely

ig)

stan

ds f

or t

he r

eser

vatio

n va

lue

of c

onsu

mer

s (r

espe

ctiv

ely,

gen

erat

ors)

at n

ode

i. A

ny ru

le to

sha

re (

)o

rK

am

ong

the

agen

ts b

y m

eans

of

the

fixed

par

t of

the

tar

iffs

that

sat

isfie

s th

e ab

ove

parti

cipa

tion

cons

train

ts,

plus

add

ition

al c

onst

rain

ts d

ue t

o th

e in

form

atio

nal

disa

dvan

tage

of

the

pric

e m

aker

22, i

s per

mitt

ed.

Whe

n th

e bu

yer a

nd th

e se

ller o

f a v

olum

e q

are

loca

ted

at th

e sa

me

node

i, th

ey c

reat

e a

surp

lus:

()

()

wg

ww

gg

ii

ii

ii

Tq

Tq

aq

ba

qb

=

wg

ii

bb

(1

6)

To p

reve

nt a

ny a

rbitr

age

betw

een

the

two

node

s, th

e tra

de o

f a q

uant

ity q

of e

nerg

y fr

om

north

to so

uth

is to

be

bille

d (

)(

)(

)w

gw

wg

gns

sn

ss

nn

Tq

Tq

Tq

aq

ba

qb

. Th

eref

ore:

()

ow

gns

sn

Tq

qb

b

(17)

Sym

met

rical

ly, a

n in

divi

dual

reve

rse

trans

actio

n sh

ould

be

bille

d:

()

()

()

wg

ow

gsn

ns

ns

Tq

Tq

Tq

qb

b

(18)

For t

he sa

ke o

f tra

nspa

renc

y, o

ne c

an p

refe

r a u

nifo

rm

w b fo

r all

cons

umer

s and

a u

nifo

rm

g b f

or a

ll ge

nera

tors

wha

teve

r th

eir

loca

tion,

but

the

ris

k of

exc

lusi

on (

parti

cipa

tion

cons

train

ts th

at w

ould

not

be

satis

fied)

shou

ld n

ot b

e ne

glec

ted.

Fo

r th

e sa

me

reas

ons

as th

e on

es m

entio

ned

form

erly

, bot

h th

e m

argi

nal p

rice

a an

d th

e fix

ed fe

e b

shou

ld b

e tim

e de

pend

ent.

As l

ong

as p

artic

ipat

ion

cons

train

ts a

re n

ot b

indi

ng, t

he

varia

tion

of b

with

the

dat

e is

sup

erflu

ous.

In

cont

rast

, th

e va

riabl

e pa

rt of

the

tar

iff f

or

ener

gy a

and

, con

sequ

ently

, the

var

iabl

e pa

rt of

the

tar

iff f

or t

rans

port

o s

houl

d be

tim

e de

pend

ent i

n or

der t

o se

nd g

ood

scar

city

sign

als t

o us

ers.

MU

LTIP

RO

DU

CT

TWO

-PA

RT

TAR

IFFS

A

s al

read

y m

entio

ned,

ene

rgy

pric

es a

nd c

onse

quen

tly tr

ansp

ort p

rices

sho

uld

be v

aryi

ng

with

tim

e. F

or e

xam

ple,

the

hour

ly tw

o-pa

rt ta

riff d

efin

ed in

(17)

shou

ld re

ad:

Page 26: Chapter 4idei.fr/sites/default/files/medias/doc/by/crampes/cramp... · 2014. 6. 12. · Box 4-2 for some illustrations of network topologies). This explains why, in the optimal dispatch,

130

Tran

spor

t pri

cing

of e

lect

rici

ty n

etw

orks

(,

())

()

()

()

nsns

nsns

nsT

qA

qB

(1

8)

and

the

tota

l exp

endi

ture

s pa

id to

the

trans

port

oper

ator

by

an a

gent

tran

sfer

ring

ener

gy

from

nor

th to

sout

h, le

t us s

ay d

urin

g on

e ye

ar, w

ould

be:

8760 1

()

()

()

nsns

nsns

TA

qB

.

For t

he re

ason

exp

lain

ed fo

rmer

ly, t

he fi

xed

part

of th

e ta

riff c

an b

e ch

osen

qui

te e

asily

23,

for

exam

ple

it ca

n be

a c

onst

ant.

The

diff

icul

ty is

that

(18

) re

quire

s to

met

er a

nd to

bill

the

flow

s fr

om n

orth

to s

outh

alm

ost c

ontin

uous

ly.

Bec

ause

the

trans

actio

n co

sts

wou

ld b

e ve

ry

high

, the

ope

rato

r w

ill in

stal

l sim

ple

met

erin

g an

d bi

lling

dev

ices

that

do

not d

istin

guis

h th

e da

te o

f ea

ch f

low

. T

his

mea

ns t

hat

the

mar

gina

l pr

ice

()

nsA w

ill n

ot b

e a

cont

inuo

us

func

tion

of ti

me.

In

mos

t cas

es, i

t will

be

a pi

ecew

ise

func

tion

with

two

seas

onal

(su

mm

er

and

win

ter)

and

two

daily

(da

y an

d ni

ght)

valu

es.

At w

orst

it w

ill b

e un

iform

all

the

year

lo

ng.

Ass

ume

the

latte

r.

In th

is c

ase

()

nsns

AA

so th

at:

8760 1

(,

())

()

()

ens

nsns

nsns

nsns

nsT

qA

qB

Aq

B,

whe

re

8760 1

()

e nsns

qq

is th

e to

tal e

nerg

y co

nsum

ed a

ll th

e ye

ar lo

ng.

The

obvi

ous

draw

back

of

this

tw

o-pa

rt pr

icin

g is

tha

t it

does

not

allo

w t

o di

scrim

inat

e be

twee

n re

gula

r use

rs w

ho c

an b

e sa

tisfie

d w

ith a

“m

ediu

m s

ize”

grid

on

one

side

and

, on

the

othe

r si

de, i

rreg

ular

use

rs w

ho n

eed

larg

e ca

paci

ty o

f tra

nspo

rt fo

r sh

ort p

erio

ds o

f tim

e.

A

solu

tion

is t

o co

mbi

ne t

he t

rans

port

dem

and

for

ener

gy

e nsq a

nd a

pro

xy f

or t

he t

rans

port

dem

and

for c

apac

ity, f

or e

xam

ple

max

()

K nsns

qq

, and

to b

ill th

em se

para

tely

in a

way

that

ap

prox

imat

es th

e op

timal

tim

e-de

pend

ent t

wo-

part

tarif

f. W

e ob

tain

a m

ulti-

prod

uct t

wo-

part

tarif

f:

ee

KK

nsns

nsns

nsns

TA

qA

qB

(1

9)

LOO

P FL

OW

S C

onsi

der

the

thre

e-no

de n

etw

ork

of B

ox 4

-2 a

nd s

uppo

se t

hat

the

line

betw

een

the

two

inje

ctio

n no

des

is t

he o

nly

one

with

a l

imiti

ng c

apac

ity K

. N

egle

ctin

g lo

sses

, the

opt

imal

di

spat

ch is

the

solu

tion

to:

Page 27: Chapter 4idei.fr/sites/default/files/medias/doc/by/crampes/cramp... · 2014. 6. 12. · Box 4-2 for some illustrations of network topologies). This explains why, in the optimal dispatch,

Cos

t rec

over

y an

d sh

ort-r

un e

ffici

ency

13

1

3

11

22

12

3,

,m

axw

gg

gg

wq

qq

Uq

Cq

Cq

rK

s.t.

3

12

wg

gq

qq

1

2

3

gg

qq

K.

Whe

n th

e ca

paci

ty c

onst

rain

t on

this

line

is b

indi

ng, e

ach

gene

rato

r cre

ates

a c

ount

er-f

low

th

at a

llevi

ates

the

load

pro

voke

d by

the

inje

ctio

n of

pow

er fr

om th

e ot

her p

lant

. B

ecau

se o

f th

is p

ositi

ve e

xter

nalit

y, t

he o

ptim

al d

ispa

tch

can

com

man

d th

at a

hig

h-co

st p

lant

sho

uld

gene

rate

pow

er d

espi

te th

e ex

iste

nce

of a

vaila

ble

gene

ratio

n ca

paci

ty a

t low

cos

t som

ewhe

re

else

in th

e ne

twor

k.

The

leve

l of

the

ther

mal

con

stra

int o

n th

e lin

e be

twee

n no

des

1 an

d 2

affe

cts

all

noda

l pr

ices

. If

pla

nt 1

is

mor

e ef

ficie

nt t

han

plan

t 2,

i.e

, if

''

12

Cq

Cq

, pr

ices

:

'2

22

3'

3

def

og

wp

Cq

Uq

>

33

'de

f

wp

Uq

>

'1

11

3'

3

def

og

wp

Cq

Uq

allo

w to

enc

oura

ge (d

isco

urag

e) g

ener

atio

n at

nod

e 2

(nod

e 1)

.

If e

nerg

y an

d tra

nspo

rt ha

ve t

o be

inv

oice

d se

para

tely

, fro

m t

he e

quili

briu

m c

ondi

tions

j

iij

pp

t, w

e ob

tain

the

set o

f lin

ear t

ariff

s for

tran

spor

t:

13/3

ot

, 23

/3o

t,

122

/3o

t.

They

exp

licitl

y en

cour

age

cons

umer

s to

tran

sact

with

exp

ensi

ve g

ener

ator

s (n

ode

2) a

nd

they

dis

suad

e th

em t

o tra

nsac

t w

ith l

ow-c

ost

gene

rato

rs (

node

1).

The

se p

rices

sen

d th

e ac

cura

te s

igna

ls to

pre

vent

jeop

ardi

sing

the

safe

ty o

f the

grid

, nam

ely

the

ohm

ic c

onst

rain

t on

the

line

betw

een

node

s 1 a

nd 2

. N

ote

that

, in

cont

rast

to th

e on

e-lin

e m

odel

, in

a m

eshe

d ne

twor

k th

e sh

adow

cos

t of

the

ther

mal

con

stra

int

on t

he l

ine

betw

een

two

node

s is

no

long

er e

qual

to

the

diff

eren

ce i

n m

argi

nal v

alua

tions

at t

he n

odes

. In

our

illu

stra

tion,

'

'2

21

13

2g

go

Cq

Cq

. Th

is is

be

caus

e al

l con

nect

ed li

nes a

re a

ffec

ted

by a

bila

tera

l tra

nsac

tion.

Con

sequ

ently

, the

mar

gina

l va

lue

of c

onge

stio

n re

flect

s al

l th

e ne

gativ

e ex

tern

aliti

es c

reat

ed t

hrou

gh t

he g

rid b

y th

e lim

ited

capa

city

of e

ach

piec

e of

equ

ipm

ent.

RE

FER

EN

CE

S

Cha

o, H

.P.

and

S.Pe

ck (

1996

), “A

Mar

ket

Mec

hani

sm f

or E

lect

ric P

ower

Tra

nsm

issi

on”,

Jo

urna

l of R

egul

ator

y Ec

onom

ics,

vol 1

0, p

p 25

-60.

C

hao,

H.P

., S.

Peck

, S.

Ore

n an

d R

.Wils

on (

2000

), “F

low

-Bas

ed T

rans

mis

sion

Rig

hts

and

Con

gest

ion

Man

agem

ent”

, Ele

ctric

ity J

ourn

al, O

ctob

er, p

p 38

-58.

C

ram

pes

C.

and

J.J.L

affo

nt (

2001

) “T

rans

port

Pric

ing

in t

he E

lect

ricity

Ind

ustry

” O

xfor

d Re

view

of E

cono

mic

Pol

icy,

Aut

umn,

vol

17,

no

3, p

p 31

3-32

8.

Cre

mer

, H

. an

d J.J

.Laf

font

(20

02),

“Com

petit

ion

in G

as M

arke

ts”,

Eur

opea

n Ec

onom

ic

Revi

ew, v

ol 4

6, n

o 4-

5, p

p 92

8-93

5.

Page 28: Chapter 4idei.fr/sites/default/files/medias/doc/by/crampes/cramp... · 2014. 6. 12. · Box 4-2 for some illustrations of network topologies). This explains why, in the optimal dispatch,

132

Tran

spor

t pri

cing

of e

lect

rici

ty n

etw

orks

Hog

an, W

.W. (

1992

) “C

ontra

ct N

etw

orks

for

Ele

ctric

ity P

ower

Tra

nspo

rt”,

Jour

nal

of

Regu

lato

ry E

cono

mic

s vol

4, p

p 21

1-24

2.

Hog

an, W

.W. (

1998

) “N

odes

an

d Zo

nes

in

Elec

trici

ty

Mar

kets

: Se

ekin

g Si

mpl

ified

C

onge

stio

n Pr

icin

g”, i

n H

ung-

Po C

hao

and

H.G

. Hun

tingt

on (E

ds.)

Des

igni

ng C

ompe

titiv

e El

ectr

icity

Mar

kets

, Klu

wer

Aca

dem

ic P

ublis

hers

, Lon

don,

pp

33-6

2.

Hsu

, M. (

1997

) “A

n In

trodu

ctio

n to

the

Pric

ing

of E

lect

ric P

ower

Tra

nsm

issi

on”,

Util

ities

Po

licy,

vol

6, n

o 3,

pp

257-

270.

Jo

skow

, P.L

. and

J.T

irole

(200

0) “

Tran

smis

sion

Rig

hts

and

Mar

ket P

ower

on

Elec

tric

Pow

er

Net

wor

ks”,

RAN

D J

ourn

al o

f Eco

nom

ics,

vol 3

1, n

o 3,

Aut

umn,

pp

450-

487.

rez-

Arr

iaga

I.J.

, F.J.

Rub

io, J

.F.P

uerta

, J.A

rcel

uz a

nd J

.Mar

ín (

1995

) “M

argi

nal p

ricin

g of

tra

nsm

issi

on s

ervi

ces:

an

anal

ysis

of

cost

rec

over

y”,

IEEE

Tra

nsac

tions

on

Pow

er

Syst

ems,

vol 1

0, n

o 1,

Feb

ruar

y, p

p 54

6-55

3.

Schw

eppe

F.,

M.C

aram

anis

, R.T

abor

s an

d R

.Boh

n (1

988)

Spo

t pric

ing

for

elec

tric

ty, K

luw

er

Aca

dem

ic P

ublis

hers

. St

oft S

. (2

002)

Pow

er S

yste

m E

cono

mic

s, IE

EE/W

iley.

W

u, F

., P.

Var

aiya

, P.

Spill

er a

nd S

.Ore

n (1

996)

, “F

olk

Theo

rem

s on

Tra

nsm

issi

on O

pen

Acc

ess:

Pro

ofs a

nd C

ount

er e

xam

ples

”, J

ourn

al o

f Reg

ulat

ory

Econ

omic

s, pp

5-2

3.

NO

TE

S

1 Th

is is

a g

ood

illus

tratio

n of

the

diff

icul

ty to

giv

e a

prec

ise,

obj

ectiv

e de

finiti

on o

f cos

ts,

as e

xpla

ined

in C

hapt

er 2

. 2

As

show

n in

Cha

o an

d Pe

ck (

1996

), ph

ysic

al r

ight

s an

d fin

anci

al r

ight

s ar

e eq

uiva

lent

w

hen

the

ener

gy m

arke

ts a

nd th

e rig

hts

mar

kets

are

per

fect

ly c

ompe

titiv

e.

Josk

ow a

nd

Tiro

le (2

000)

pro

vide

an

anal

ysis

of v

ario

us n

on-c

ompe

titiv

e co

nfig

urat

ions

. 3

See

Josk

ow a

nd T

irole

(200

0), p

. 45

2.

4 Fo

r an

illus

tratio

n se

e B

ox 4

-2.

For m

ore

deta

ils, s

ee fo

r exa

mpl

e H

su (1

997)

. 5

Schw

eppe

et a

l. (1

988)

. 6

See

for e

xam

ple

Stof

t (20

02),

p. 4

17.

7 St

udie

s of

mar

ket

pow

er a

re b

ased

on

calib

rate

d si

mul

atio

n m

odel

s. S

ee r

efer

ence

s in

C

ram

pes a

nd L

affo

nt (2

001)

. 8

We

do n

ot c

onsi

der t

he is

sue

of n

odal

-pric

e ra

ndom

ness

. To

hed

ge a

gain

st p

rice

vola

tility

, us

ers c

an si

gn fi

nanc

ial c

ontra

cts.

See

Hog

an (1

992)

and

Cha

pter

5 in

fra.

9

See

Ore

n et

al.

(199

6).

10

See

Pére

z A

rria

ga e

t al.

(199

5).

11

Prov

ided

the

auth

ority

in c

harg

e of

pric

ing

is n

ot c

onst

rain

ed b

y a

lack

of i

nfor

mat

ion

on

the

will

ingn

ess t

o pa

y of

the

user

s. F

or a

n ill

ustra

tion,

see

Cha

pter

6.

12

Se

e C

hapt

er 3

. 13

Fo

r det

ails

on

econ

omic

and

lega

l pric

e di

scrim

inat

ions

, the

read

er is

refe

rred

to C

hapt

er 1

. 14

O

n th

e pr

inci

ples

of c

oope

rativ

e-ga

mes

theo

ry, s

ee C

hapt

er 3

. 15

In

the

tw

o-pa

rt ca

se, t

he v

aria

ble

part

is i

ncre

asin

g lin

early

with

the

qua

ntity

of

ener

gy

trans

porte

d.

In m

ulti-

part

tarif

fs,

the

varia

ble

part

is a

pie

cew

ise

linea

r in

crea

sing

fu

nctio

n.

16

Det

ails

on

actu

al ta

riffs

are

pre

sent

ed in

Cha

pter

8.

17

We

do n

ot d

iscu

ss t

he a

dditi

onal

int

erte

mpo

ral

cons

train

ts d

ue t

o th

e m

anag

emen

t of

hy

drop

lant

s or t

o th

e ob

ligat

ion

to sa

tisfy

hea

ting

cond

ition

s in

ther

mal

pla

nts.

Page 29: Chapter 4idei.fr/sites/default/files/medias/doc/by/crampes/cramp... · 2014. 6. 12. · Box 4-2 for some illustrations of network topologies). This explains why, in the optimal dispatch,

Cos

t rec

over

y an

d sh

ort-r

un e

ffici

ency

13

3

18

A

ctua

lly, D

and

N a

re s

ucce

ssiv

e, b

ut th

e ar

gum

ent i

s th

e sa

me

as if

ther

e w

ere

seve

ral

non-

rival

sim

ulta

neou

s ne

eds

to s

atis

fy w

ith th

e sa

me

equi

pmen

t: it

is th

e w

illin

gnes

s to

pa

y of

all

user

s tha

t mus

t be

take

n in

to c

onsi

dera

tion.

19

Th

is i

s tru

e w

hen

the

long

run

mar

gina

l co

st o

f ge

nera

tion

is n

on d

ecre

asin

g an

d ge

nera

tors

hav

e op

timiz

ed th

e si

ze o

f the

ir pl

ants

. 20

Se

e C

rem

er a

nd L

affo

nt 2

001.

21

To

whi

ch th

e m

argi

nal c

ost o

f los

ses s

houl

d be

add

ed.

22

Self-

sele

ctio

n co

nstra

ints

are

to

be a

dded

to

desi

gn t

he m

enu

of t

ariff

s w

hen

the

pric

e m

aker

can

not o

bser

ve s

ome

indi

vidu

al c

hara

cter

istic

s of

the

pric

e ta

kers

, for

exa

mpl

e th

e co

nsum

ers'

will

ingn

ess

to p

ay o

r th

e ge

nera

tion

cost

of

prod

ucer

s. T

he p

robl

em i

s to

pr

even

t an

opp

ortu

nist

ic s

witc

h of

som

e us

ers

tow

ards

tar

iffs

that

are

des

igne

d fo

r so

meb

ody

else

. C

hapt

er 6

pro

vide

s an

illu

stra

tion

of th

e de

sign

of

cont

rol m

echa

nism

s un

der i

nfor

mat

ion

asym

met

ry.

23

Exce

pt w

hen

ther

e is

a se

rious

info

rmat

iona

l gap

that

impo

ses f

ine

tuni

ng to

resp

ect a

ll th

e pa

rtici

patio

n an

d se

lf se

lect

ion

cons

train

ts.