Possibilities & Limitations of
Extending the Wholesale / Bulk Power Transactive Techniques to
Retail Markets & Distribution Operations Ralph Masiello
Jessica Harrison
DNV GL Energy [email protected]
1
What is Markets 3.0?
2
U.S. wholesale electricity markets are
characterized by the following trends: • Need to manage new products
e.g., Demand Response, Variable Energy
Resources, Microgrids, Self-Optimizing Customers
& Energy Storage
• Penetration of & coupling with retail resources use of distributed generation and smart load
resources from the industrial, commercial and
residential sectors
Markets 1.0 • Wholesale day ahead
energy on hourly schedules • Ancillary services • Balancing and regulation • Transmission rights
Markets 2.0 • Co-optimized energy &
ancillary services • Congestion pricing • Nodal real time dispatch • Capacity markets for DR
Markets 3.0 • Dynamic retail pricing • DR for ancillary services • Capacity markets for firming & DR • Intra-hr scheduling of renewables • Storage as a resource
1995 - 2003
2001-2010
2011-2020
Real-time wholesale markets meet retail resources
Recent History FERC NOI & NOPR on VER Integration & Cost Allocation FERC Report on Demand Response & NOPR 745 on compensation FERC NOPR on Fast Regulation from Storage
Integrating Distributed Energy Resources
• Early euphoria being subdued by challenge realization! – Visibility – no telemetry (AMI is NOT the solution !)
– Control (Definitely NOT AMI; multiple technologies for each end use / resource
– Grid Security - Backfeed, fault ride through, frequency response
– Market Integration “estimated response” for settlements; estimating elasticity in market clearing
• DER Categories – Distributed Generation – PV, CHP, micro-wind
– Distributed Storage
– Dynamic Pricing – autonomous demand price elasticity
– Dispatchable Demand Response
3
Integrating Demand Response: Key Research Questions
• What are the potential impacts of greater DR integration into the wholesale market? – What are the effects on real time markets prices & supply dispatch
over time?
– What are the conditions for preserving market convergence?
4
• Dispatchable Demand Response (DDR): – planned changes in consumption in response to direction from
someone other than the customer
– modeled as a supply resource dispatched similarly to generation
• Dynamic Pricing (DP) response: – customer decides whether and when to reduce consumption
– modeled as a voluntary customer response to market prices DNV KEMA study with NYISO, Market Dynamics of Integrating Demand Response into Wholesale Energy Markets, The Electricity Journal, April 1, 2013.
• Distributed Energy Resources (DERs) include a variety of supply-side and demand-side resources. Those examined in this study include:
Categories of Distributed Energy Resources
5
SOC – Self Optimizing Customers
DR – Demand Response (Including Autonomous Price
Responsive Load (Dynamic Pricing)
DES – Distributed Energy Storage
DG – (PV – Distributed and “Behind-the-meter”
PhotoVoltaics; CHP – Combined Heat and Power)
PEV – Plug in Electric Vehicles
Re
lative F
ore
casting C
om
ple
xity
Primary Control
Secondary Control
Tertiary Control
Time Control
Spinning Reserve
Non-Spinning Reserve
Load Following
InertiaGovernor Response
Regulation
Economic Dispatch
Supply Stack
Seconds
Minutes
Minutes
Hours
ContingencyReserve
Forecast ErrorMinutes
10 minutes
30 minutes
6
Time Domains for Flexibility
Microgrid Resource Configuration
7 Source: Quanta
8
8/6/2014
DER
Profile*
High
DER (Max MW)
Mid
DER (Max MW)
Low
DER (Max
MW)
Penetration
Assumptions Variability Drivers
PV 7812 4757 1747 Scaled according to ISO scenarios for distributed PV
Clearness index and PV Technology. Based upon forecast errors calculated in LTPP High Load case.
CHP 4468 3092 1732 Based upon CEUS Prices, temperature, conforming load
SOC 1277 806 337 Based upon CEUS Prices, temperature, conforming load
PEV -882 -662 -625 Based upon research by NREL
Commute time and traffic congestion
DES -2808 -1920 -1033 Based upon CEUS PV smoothing requirements and prices
DR -2466 -1926 -1390 Based upon existing utility programs
Prices, load and temperature
High DER Penetration leads to forecast uncertainty and increased
production costs.
Impacts by DER Type & Penetration
9
8/6/2014
No Visibility Case
Ma
x L
oa
d
Follo
win
g D
ow
n
Ma
x L
oa
d
Follo
win
g U
p
Ma
x
Re
gu
latio
n
Up
Ma
x
Regu
lation
Dow
n
5,079 MW
5,683 MW
1,084 MW
760 MW
Ma
x L
oa
d
Fo
llow
ing
Do
wn
M
ax L
oa
d
Follo
win
g U
p
Ma
x
Re
gu
latio
n
Up
Ma
x
Re
gu
latio
n
Dow
n
4,652 MW
4,753 MW
1,083 MW
749 MW
Visibility Case
Visibility provides a large reduction in the 95th percentile of Load
Following requirements. Minimal Impact on Regulation.
Estimated Load Following & Regulation Requirements by Visibility Scenario
10
Density (units/square mile)
Rate of DER
State Change
PVgrid PVBehindMeter
CHPPrice Taker
CHPDynamic
StorageUtility StorageBehindMeter
DP
Re
al-T
ime
DD
RD
isp
atc
h
DP1-HourAhead
EVSmart
EVPassive
1 min
5 min
1 hour
SOC
Device density and rate of change are the drivers
for communications technology and costs
Information Requirements
B. Technical requirements for monitoring and control to achieve market and operational benefits
Communication Architectures Various Stakeholders play in own time and density domain
11
Coverage / Availability Density
System Polling
Time
Utility DA
Private Network
Customer
Internet
Other/3rd-Party Private Network
1 min
5 min
1 hour
Public Carrier Wireless
Utility AMI
Private Network Broadcast Semi-Control
Only
8/6/2014
Communications Architectures
12
Ownership /
Timeline
Present SCADA AMI Mesh
Networks
Broadcast
Radio
Cellular GPRS
SMSWi-Fi
Internet POP /
Ethernet/WiFi
BAS
Networks -
larger
commercial
EmergingDistribution
Automation
AMI Mesh
Networks700 MHz Cellular LTE
Wi -Fi public
hot spots
pervasive in
C&I and most
residences
EV GPRS/
Wireless
BAS
penetration
and Open
ADR
DER
maintenance
via cellular /
internet
2020SCADA / DA
on fiber / 700
MHz
not EOL for
current AMI
systems yet
migrated to
other
spectrum?
Adopted for
DA and
mobile
apps/ AMI?
Not Availablenext
generation?next evolution? pervasive
EV on next
generation
BAS
ubiquitous
in C&I
DER
maintenance /
ops via next
generation
Pros
Low Latency
NERC CIPS
inherent for
utility DER
assets
Ubiquitous
and Low Cost
Modems
Available
Very Low Cost
potential
spectrum re-
allocation to
utility use.
ubiquitous and
low costs
already used
for PV
ubiquitous
high
performance
new cellular
standard
low modem
costs and nil
data cost
ubiquitous and
low modem /
nil data cost
rely on auto
industry
directions
and
capabilities
support of
Open ADR
likely no
incremental
cost for DER
low
incremental
cost for DER
monitoring
Consexpensive /
proprietary
/not on LV
Utility owned
and
controlled
provisions for
3rd party
access
Ubiquitous but
with spots of
non-access
only one way
Utility
owned and
controlled
provisions
for 3rd party
access
obsolete and
carriers will
abandon 3-5
years
higher
modem costs
and higher
service
impacts/cost
s for DER
data
not ubiquitous;
security
authentication
and validation
required
possibily
encryption
proprietary
and closed
proprietary
and closed
proprietary
and closed
DER
Applications
Utilty scale
PV and utility
storage
Rooftop PV;
Residential
HVAC;
Distributed
Storage
small DR
assets
residential hot
water and AC
Unknown
adoption in
CA
Distributed PV
Residential AC
distributed
storage
GPRS targets
And DER near
an Internet POP
with WiFi
access indoor
esp
Any C&I
facility DER
and most
residential
EV smart
charging
commercial
DER, all
SOC, most
CHP
distributed PV
and
distributed
storage
UTILITY COMMON CARRIER 3rd Party
13
CAISO Benefits in millions $ for 2020
Generation
Cost & Start/
Stop
CO2
Emissions
CAISO Production
Cost With Visibility
Savings/year
$7,541/yr
$7,932/yr
$391/yr for
Monitoring
$307/yr
$84/yr
Less Large
Plant
Generation &
Fewer start/stop
costs on plants
Lower
Emissions
CAISO Production
Cost With No Visibility
Benefits of monitoring and control are significant compared to the
communication, monitoring, and forecasting infrastructure costs
C. What are the CAISO costs and expected benefits to increased DER visibility and control?
Price Elastic Load = Sequential Markets 1. Market Measures / Forecasts Load
2. Market Clears the Supply Side Bids and Sets Prices
3. Load Reacts to Price
4. Repeat
• Anecdote – UK in the 80’s (courtesy of Richard Tabors)
– First UK Markets had industrial customers exposed to market prices
– Customers would react to prices once set
– Some price oscillations observed
• Market did not take elasticity / behavior into account
• Anecdote – A 2013 Swedish study had similar findings (Sweco Energy Markets)
• Economists are familiar with the iterative interaction between elastic supply & elastic demand:
The “Cobweb Theorem”
14
- Relative elasticities dictate convergence or divergence
However, the Cobweb Theorem does not consider time dynamics
Key to Understanding Behavior:
• Price is a control signal • Market clearing and the
establishment of supply and demand curves are dynamic processes
Simulation
Process Control Model
Simple market model based on control theory - captures generation & demand time dynamics • Supply-demand imbalance is input to clearing
function which adjusts price according to supply & demand elasticities.
• Feedback gain is inverse of sum of supply & demand elasticities. Delay equals periodicity of market clearing function.
• Critical parameters: price elasticity ratios, time delay ratios ; demand elasticity error
System Dynamics Model
Detailed dynamic model of a market operation using system dynamics • Non-linear supply curves representative
of a real market and non-linear demand curves based on published demand elasticity research.
• Integrates day ahead, hour ahead, and real time energy market processes.
• Includes residential and commercial end-uses (HVAC, lighting, water heating, refrigeration).
• Does not predict price but captures market dynamics
15
Imbalance
Demand
Supply
Price
price
-1Z
delay-demand
z
1
z
1
-K-Supply time delay
-K-
Supply elasticity
Step
-K-
Market Gain
1Market
Demand/Supply
-K-Demand time delay
-K-
Demand elasticity
-1Z
1 time step delay-supply
-1Z
delay-priceforsupply
-1Z
delay-price
RT Dispatch
RT Imbalance
RT Total Demand
RT DDR
ResponseRT Supply Curve
RT Generation
Response
RT Total Supply
-+
+
+
+
+RT Price
RT DP Response+
+
-
RT Total
Responsive Demand
RT Unresponsive
Demand
+
-
Forecast DemandForecast Supply
RT Commitment-
+
++
RTC DDR
Response+
RTC Generator
Response+
+
+
RT DP ResponseRT Supply Response RT DP Response
RT DDR Response
RTC DDR ResponseRTC Supply Response
Theory
Market Models
Control Theory Modeling Results
16
Under some scenarios of DR integration, the markets can become unstable.
A simple example considers how the dynamic response of generation, demand, and market operations affect market stability over ranges of relative supply and demand elasticity.
In this case, the market misestimates demand elasticity (i.e., 100% error)
Where generation is less elastic than demand, the system goes unstable.
-1 -0.8 -0.6 -0.4 -0.2 0 0.2 0.4 0.6 0.8 1-10
-8
-6
-4
-2
0
2
4
6
8
10
Real part
Ima
gin
ary
pa
rt
error = 100%
Real Part
Imagin
ary
Part
Stable region
Routh Hurwitz
Criterion
Solving Analytically for System Poles
(Root-Locus)
There are scenarios for which the overall system will not be stable when the market misestimating demand elasticity (i.e., 100% error). Misestimating elasticity is akin to operating the market as it is operated today.
The Real World is MUCH MORE Complex
• Multiple Markets – Day Ahead, Hour Ahead, Real Time
• Multiple Supply Resources with Different Time Dynamics
• Multiple Load Side Elements
• More Complex Load Side Behaviors
• Non-linear / Time Varying Elasticities
First Key Observation
18
At scaled up penetration, DP response
becomes unstable as shown when the
duration is 60 min. Added to information
latencies, this means the market is clearing
for load that responded to the prior period
price but is not aware of that effect.
DP responding to an hourly price signal
with a 60 min duration affects RTD prices
but 20 min durations do not.
RTD Price
1,000
750
500
250
0
0 138 276 414 552 690 828 966 1104 1242 1380
Time (Minute)
$/M
W
RTD Price : 7BC
RTD Price : 7dumbdp_20
RTD Price : 7dumbd60
$/M
Wh
July - Base Case
July - DP, 20 min duration
July - DP,60 min duration
Market impacts depend on: penetration, timing of price signals, and relative duration of DP compared
to the frequency of the market dispatch & price publication.
Total Aggregate DP Response
4,000
2,000
0
-2,000
-4,000
0 138 276 414 552 690 828 966 1104 1242 1380
Time (Minute)
MW
Total Aggregate DP Response : 7BC
Total Aggregate DP Response : 7dumbdp_20
Total Aggregate DP Response : 7dumbd60
July - Base Case
July - DP, 20 min duration
July - DP,60 min duration
Second Key Observation
19
DP impacts are very sensitive to DP penetration, demand elasticity, and the accuracy of estimated
demand elasticity in the market clearing algorithms.
As the amount of responsive DP in the
market increases, price potentially
increases and can grow to be volatile
As the amounts of viable DP in the market
grow, load oscillations grow. Increased
“penetration” of DP in effect increases the
ratio of demand elasticity to supply
elasticity and increases instability.
Total Aggregate DP Response
20,000
10,000
0
-10,000
-20,000
0 138 276 414 552 690 828 966 1104 1242 1380
Time (Minute)
MW
Total Aggregate DP Response : 7 dumb dp 2-5 RTCH
Total Aggregate DP Response : 7 dumb dp 2 RTCH
Total Aggregate DP Response : 7 dumb dp 1 RTCH
July - DP, 2.5 x penetration
July - DP, 2 x penetration July - DP, 1 x penetration
Graph for RTD Price
1,000
750
500
250
0
0 138 276 414 552 690 828 966 1104 1242 1380
Time (Minute)
$/M
W
RTD Price : 7 dumb dp 2-5 RTCH
RTD Price : 7 dumb dp 2 RTCH
RTD Price : 7 dumb dp 1 RTCH
July - DP, 2.5 x penetration
July - DP, 2 x penetration
July - DP, 1 x penetration
$/M
Wh
RTD Price
Thinking about ISOs, DSOs & MGOs
20
Registration
Bidding
Market Clearing
Notification
Measurement & Validation
Settlements
Wholesale Markets
T
F
Suppliers
Demand Side Aggregators
Wholesale Takeout Point
DSO
Microgrids
The Devil is in the Details
21
Rules of the Game • Can one entity have multiple roles? - Direct access; DSO resource; MGO
Bi-lateral Transactions & “Open Access Distribution” (OADIS)
• Some microgrid operators will have multiple sites on different takeout points (e.g., DOD)
Settlements • What constitutes a revenue meter? - e.g., EV and chargers have meters and
comms; why duplicate?
Validation • The inevitable DR “what would it have been?” question
Market Co-ordination • Timing of bidding closure, market clearing, notification across layers
Gaps in Understanding
• Information Arbitrage
– Interaction of DSO and ISO markets in time and opportunity to influence pricing
– Stability of Market Behavior with layered clearing processes
• Interaction of gate closures, processing time, notification, participant decision making
• Business Models for New Resources in Markets
22
Business Models – Example - Storage
• Storage as a Generator
– Must separately bid discharging and charging and take risks of not clearing / duplicate clearing
• Storage co-optimized by market operator
– Basis of bidding? Paid clearing price like a generator?
– New asset class offering storage services?
• Hybrid: Storage as a (regulated) asset class and 3rd parties “own” energy in storage
23
And – Reliability Issues
• Today DG MUST disconnect on grid low voltage for safety reasons
• At high penetrations this can cause grid level event “magnification”
– Routine cleared line fault becomes loss of 000’s MW of PV
• So Fault Ride through, low voltage ride through standards needed
• And – rules on “pre-emptive disconnect”
24
Conclusions
• If We Want to Use Price as a Control Signal
– Better do the Control Systems Design
– Artificial Volatility is NOT a Good Thing
• The More Complex the Market Design – the More Opportunities for “Strategic Bidding” and Unexpected Outcomes
25
Top Related