Demand Response From Utilities to Smart CitiesDemand Response: From Utilities to Smart Cities...
Transcript of Demand Response From Utilities to Smart CitiesDemand Response: From Utilities to Smart Cities...
Demand Response: From Utilities to Smart Cities
Introduction
1Demand response (DR) is an important tool deployed in
electricity markets to overcome temporary real-time mismatches
in supply and demand. In a typical DR event, a utility suppresses
short-term demand by offering incentives to users to avoid more
expensive supply expenditure. In other words, the utility elicits a
response (reduced consumption) from the demand side using a
signal, economic or otherwise (for instance, an incentive per unit
of energy reduced).
A similar behavior is seen with cab aggregators that use surge
pricing to both reduce demand and potentially increase supply.
Another example is the use of different charges at different times
of the day by hyper-local last-mile services delivering food on
demand. Generally, DR is most useful when increasing supply is
not viable and demand needs to be managed instead.
Can we apply the DR paradigm to other scenarios? Let us
consider the example of smart cities. Smart city programs
typically aim for excellence in citizen experience while availing city
services. But even the best IoT-enabled infrastructure in a smart
city can come under stress during short periods of peak demand.
An example would be the strain on public transportation during
periods of peak citizen engagement, such as a fireworks display.
One might argue that smart services during normal operations
increase the expectations of citizens, who then feel let down by
poor experiences during peak demand periods.
Can DR be of help here? DR can be thought of as a ‘horizontal’
smart city service that cuts across multiple ‘vertical’ citizen
services to ensure the best city-government service, even during
peak demand. For example, a smart city can move citizens to and
from mass public events in a controlled way by incentivizing them
to choose alternative paths or stagger their travel. This would
improve the citizen experience and help the city avoid
infrastructure investment that is justified only during peak
WHITE PAPER
WHITE PAPER
demand. The city-government thus acts as an aggregator that
can handle peak demand for its offered services.
This white paper explores the role of DR in improving citizen
services in a smart city. We will present a simplified DR
architecture and explore possible use cases; we will also illustrate
this with a hypothetical situation. We will present a sequence of a
smart city operation that includes DR and conclude with some
suggested approaches in realizing DR for smart cities.
Architecture for DR
An aggregator connects the supply side to the demand side,
operating on two timescales: short term and long term. In short-
term operations, the aggregator collates the real-time supply and
the constraints at a given time and over the entire duration of the
operations. For example, a cab aggregator knows the number of
vehicles available across a city. The aggregator would also know
or forecast the demand at any given point.
Depending on the demand and supply, the aggregator can
present offers at ‘normal’ prices or try to reduce the demand by
increasing real-time prices (for a taxi) or offering incentives (to
electricity consumers). The guiding principle in this optimization is
the business objective: revenue/profit and customer satisfaction.
The algorithm should factor in demand requests and supply
availability.
In long-term operations, the aggregator can try to predict the
typical demand as a function of environmental parameters (for
example, time of week, weather and road conditions) and could
also learn the sensitivity of the demand side to price or incentive
signals. These longer-term learned models can be used by the
aggregator to optimize the supply and demand in real time.
In addition, regulatory requirements and preferences might need
to be reflected in the aggregator’s decisions, even if it is in
conflict with the profit goal. For example, a cab aggregator would
like to increase supply of cabs in a congested area with high
demand while the regulator (city-government) might want exactly
the opposite. Therefore, a DR architecture should consider
regulatory constraints that affect decisions.
WHITE PAPER
Use Cases in a Smart City
Figure 1 illustrates the use of DR in multiple domains in a smart
city. For each domain, we provide examples of the components
proposed. We observe that the possible application of DR
constitutes a diverse spectrum of domains, with the goods and
services being both discrete in time and space, as well as
continuous, like in the case of electricity.
Domain Electricity Parking Healthcare – Epidemic Management
Mobility
Supply side Utility provider Private/Public parking provider/owner
Healthcare professionals
Cab operators, public transport providers
Aggregator Ancillary service providers
City authority or a private company
City administration/government
City transport authority
Regulator Utility/Electricity regulator
City administration/government
City administration/City government
City administration/government
Demand side Consumers Drivers Citizens Commuter
Goods/Service Uninterrupted supply
Adequate parking Vaccination for all Multimodal journey/trips
Figure 1: Examples of DR in a Smart City
Let us take a specific use case, focusing on mobility as a primary
service and exploring how the demand response paradigm can be
applied here. For this example, let us assume that the city offers
all smart services through a platform. Citizens can access services
in the healthcare, utilities, transport and e-government domains
by logging on to the smart city platform.
For the sake of illustration, let us take the case of a citizen who
subscribes to mobility services — which could include notifications,
route planning and offers — through this platform and wants to 2
plan a trip from point A to B. The mobility services section of the
smart city demand response platform will perform the following
actions:
1. Accept the citizen’s trip demand, which consists of the origin
and destination, along with preferences such as pick-up time
2. Retrieve supply side info from transport aggregators
3. Calculate the supply levels using own data and data from
transport aggregators
4. Match the supply and demand resources
5. Learn from preferences based on its info on the citizen
WHITE PAPER
6. Process the demand and suggest travel plans to the citizen
with additional travel choices – offering discounts/coupons to
manage demand surge
7. Use the data for future learning
Demand Response Module in Mobility
Services
With the above details, let us do a deep dive using an illustration
of the experiences of two New Yorkers – Tom Peters and Greg
Allen. Let us say they want to go from Times Square (point A) to
Wall Street (point B), and access the mobility app on a Sunday
requesting options for 8 am on Monday. Once the request is
placed, the sequence of events that the demand response module
will start performing is illustrated in Figure 2.
Smart citizen
(Greg and Tom)Ride
requestDemand
store
Processingdemand
Load historical trip data (Tom
and Greg)
Load user-specictransport model
Aggregate current tripdemands
Predict transport
mode for Greg and Allen
Fetch supply options
Transportoperators
Matchdemand and
supplydata
Determine transport on live
dataDemand response recommendation algorithm
Demand > Supply YESNO
Tom and Greg selection
Offers (example, discount price at different time)
Predict nal transport mode and price
Appendhistoricaltrip data
1 2 3
4
5a5b 6a 6b 6c
6d5c
7
88a8b 9
10
11
12
Figure 2: Demand Response Module for Mobility Services
The key steps are listed below:
1. The details of the requests by Tom and Greg in terms of
source, destination, timing and trip preferences are captured
2. Historical requests are retrieved for the two citizens. The
objective is to learn their travel patterns based on historical
data
WHITE PAPER
3. The prediction model built inside the platform takes the
historical demand requests by Tom and Greg into account to
come up with a possible prediction in terms of transport
preferences. This prediction is certainly a candidate for deep
learning neural models or machine learning classification
models
4. The prediction output of the model is saved for further
consideration
5. The platform will also now analyze requests from other users
in terms of pick-up time, destination and location, along with
various other factors such as airport trip, rush hour, day of the
week, weather patterns, user’s preferred travel mode
(bus/train/taxi) and travel preferences in terms of time or price
6. It gathers supply information from its own inventory or looks
to transport aggregators for the best possible options for a trip
between Time Square and Wall Street
7. For further mobilization of available resources other than
inventory, during the period of demand spikes, the platform
also makes requests to suppliers or master aggregators.
Master aggregators in this context would be the bus
companies, train systems or taxi operators, from whom trip-
specific information such as bus numbers, train numbers, taxi
options and estimated time of travel will be sourced
8. The transport aggregators will provide information on the
availability of supply
9. The platform will now match the demand and supply resources
and decide the possible mode and price of the transport
options
10.Based on user preferences and supply/demand resources, it
will finally come up with recommendations in terms of
coupons/discounts and provide this information to the citizen
to manage the demand surge. Also, it will propose a premium
price for the route given the demand surge
After executing the above options, the demand module could
present Greg and Tom with the following options.
n Bus T2A from Times Square to Wall Street; estimated travel
time 30 minutes; $2.75
n Subway from Times Square to Wall Street; estimated travel
time 20 minutes; $2.75
WHITE PAPER
n Taxi operator 1 from Times Square to Wall Street; estimated
travel time 40 minutes; $8 at 8 AM
n Taxi operator 2 from Times Square to Wall Street; estimated
travel time 20 minutes; $4 at 7.30 AM, including a 50%
discount (recommended)
The module not only presents options at the preferred pick-up
time, but also highlights better deals around that time after
considering available offers and the user persona.
Tom and Greg may accept the discounted offer for a different
time or pay a premium to travel at 7.30 am. If Tom accepts the
offer, the DR module will apply specific discounts as per available
contracts with suppliers. In case Greg does not accept the offer,
he will pay a premium price for his travel at 8 am. Once Tom and
Greg make their selections, their requests will be appended to
their travel histories for further learning.
At a high level, the DR module aggregates user demands and
supplier choices, applies contractual discounts on the supplier
choices and finally, based on user preferences, provides multiple
choices to those who subscribe to the mobility app. The DR
module is able to provide such recommendations by solving an
optimization problem – matching supply and demand in real time
with demand and supply constraints. Machine learning techniques
can play a significant role in this optimization process, especially
in predicting real-time series demands. In the prediction of time
series demands, the deployment of recurrent neural networks like
LSTM can bring significant results in terms of the predictive power
of what future demand will look like.
Applying DR to Real-world Challenges
DR is an important tool to handle scale and efficiently manage
real-time supply-demand mismatches. This approach is common
in the energy domain, and we believe it is applicable for smart
city services as well. To implement DR in real-world systems, the
models must learn from past data or realistic simulations so that
the alpha/beta phases of deployment can be faster.
Some suggested approaches to implementing a DR system for
smart cities:
1. Set up and build a simulation-based environment of a domain.
For example, in the use case mentioned in this paper, a
mobility environment that can be the basis of any ML or RL
approaches for designing the controller. The simulation will
WHITE PAPER
involve logging the multiple requests at different times – and
even future day and times – by citizens wishing to travel from
place A to B. The simulation should be a faithful model in
terms of the underlying distributions of the data generated.
2. Calibrate or design the behavioural response models on the
demand side. For example, in the use case given above, trip
requests would be the behavioral response. This would require
some research on current demand trends.
3. Build scalable algorithms for matching supply and demand in
real time with efficient management of constraints. This is the
optimization problem that the DR system is intended to solve.
In the example given above, the optimized route does not
necessarily have to ensure only shorter travel time; a travel
mode with a discount could be just as relevant. The algorithms
need to maintain optimal performance even when faced with
hundreds of thousands of trip requests. The supplier network
must be robust.
4. We have approached DR from a citizen-centric approach.
However, the approach must balance the benefits to all the
stakeholders. In the case of third-party suppliers, the
aggregator must demonstrate an increase in revenue to make
it attractive for them to sign up to the platform.
5. The approach should incorporate trade-offs between system
efficiency and data privacy.
6. Provide adequate incentives to citizens to manage demand
constraints.
References
1. IEEE; Demand Response in Electricity Markets: An Overview; June 2007; accessed
Sep. 2019; https://ieeexplore.ieee.org/abstract/document/4275494
2. arXiv.org; Managing travel demand: Location recommendation for system efficiency
based on mobile phone data; Oct. 2016; accessed Sep. 2019;
https://arxiv.org/pdf/1610.06825.pdf
All content / information present here is the exclusive property of Tata Consultancy Services Limited (TCS). The content / information contained here is correct at the time of publishing. No material from here may be copied, modified, reproduced, republished, uploaded, transmitted, posted or distributed in any form without prior written permission from TCS. Unauthorized use of the content / information appearing here may violate copyright, trademark and other applicable laws, and could result in criminal or civil penalties. Copyright © 2019 Tata Consultancy Services Limited
About Tata Consultancy Services Ltd (TCS)
Tata Consultancy Services is an IT services, consulting and business solutions
organization that delivers real results to global business, ensuring a level of
certainty no other firm can match. TCS offers a consulting-led, integrated portfolio
of IT and IT-enabled, infrastructure, engineering and assurance services. This is TMdelivered through its unique Global Network Delivery Model , recognized as the
benchmark of excellence in software development. A part of the Tata Group,
India’s largest industrial conglomerate, TCS has a global footprint and is listed on
the National Stock Exchange and Bombay Stock Exchange in India.
For more information, visit us at www.tcs.com
TCS
Des
ign
Serv
ices
M
09
19
II
I
WHITE PAPER
Contact
Visit the page on Research and Innovation www.tcs.com
Email: [email protected]
Blog: #Research and Innovation
Subscribe to TCS White Papers
TCS.com RSS: http://www.tcs.com/rss_feeds/Pages/feed.aspx?f=w
Feedburner: http://feeds2.feedburner.com/tcswhitepapers
About The Authors
Ramesh Balaji
Ramesh Balaji is a senior data
scientist, with a focus on smart
cities (understanding the interesting
patterns of smart people through
machine learning techniques) and in
assisted living (understanding the
activity patterns of the elderly
through data and image recognition
patterns). He holds a master's
degree in computer engineering
from California National University
for Advanced Studies, USA.
Arun Vasan
Arun Vasan is a principal scientist
with TCS R&I, where he works on
cyber-physical systems. He obtained
a BTech degree in computer science
and engineering from IIT Madras,
India, and an MS and PhD in
computer science from the
University of Maryland, College Park,
USA. He has served as visiting
faculty at IIT Madras and as a
reviewer for several conferences and
journals. His current research
interests are in analytics for
intelligent infrastructure, with a
focus on energy and mobility.
Srinivasan Raghavan
Venkatachari
Raghavan is the head of the digital
citizen program within the TCS
Research and Innovation (R&I) unit,
with a focus on creating research-
led business offerings for citizen-
centric smart cities. He holds a
bachelor’s degree in computer
science and engineering from
IIT Delhi, India, and a master’s
degree in aerospace engineering
from IIT Madras, India. He is a
senior member of the ACM.
Arun Vijayakumar
Arun Vijayakumar is an enterprise
architect with TCS R&I. With over
20 years of IT experience in
enterprise architecture and smart
city consulting, he presently works
on the digital citizen and connected
social systems research programs,
handling the architecture of various
solutions and research projects
within these programs. Arun holds a
master’s degree in applied statistics
from the University Campus,
University of Kerala, India, with
computer application and operations
research as his specializations, and
an MBA in operations management
from the University of Madras, India.