Skip to content The Open University

Workshop on the Mathematics of Demand Side Management and Energy Storage

Monday 1st and Tuesday 2nd June 2015, The Open University, Walton Hall,  Milton Keynes.

Where available, slides for the presentations may be obtained from the presentation title.  Posters are listed at the end; where available, the poster may be obtained from the poster title.


Monday, 1 June

1.30 Welcome and Session 1:  Storage in distribution and transmission

1.30 Welcome by William Nuttall (Chair, Open Energy)

1.40 Lewis Dale (National Grid) System balancing and role of storage

2.20 Phil Taylor (University of Newcastle) Integration and control of energy storage in distribution networks

3.00 Ben Godfrey (Low Carbon & Innovation Engineer, Western Power) Project FALCON - energy storage for a DNO

3.20 Tea and Coffee

3.40 Session 2:  Electricity markets, energy economics and the incentivisation of storage

3.40 Goran Strbac (Imperial College London) Quantifying the value of energy storage in supporting integration of renewable generation

4.20 John Moriarty (University of Manchester) Bayesian inference on a jump-diffusion Ornstein-Uhlenbeck process for modelling electricity spot prices, and applications to the economically optimal control of CHP and heat storage

5.00 Jan Palczewski (University of Leeds) American contracts for power system balancing

5.20 Danica Vukadinovic Greetham  (University of Reading) A probabilistic framework for forecasting household energy demand profiles

5.40 Lisa Flatley (University of Warwick) Competing stores in an arbitrage market

6.00 Close

7.30 Dinner  

Tuesday, 2 June

9.00 Welcome to second day and Session 2 continued

9.00 James Cruise (Heriot-Watt University) Evaluating the value of storage facilities for buffering and arbitrage

9.40 David Angeli (Imperial College London)  Mean-field game formulations for distributed storage management in dynamic electricity markets

10.00 Session 3:   Smart grids, storage and demand side management

10.00 Ana Bušić (École Normale Supérieure - Paris) Ancillary service to the grid using intelligent deferrable loads

10.40 Nicolas Gast  (INRIA - Grenoble) Stochastic analysis of real and virtual storage in the smart grid

11.20 Tea and Coffee

11.40 Session 4:   Challenges from Industry Chair: Chris Dent (Durham Energy Institute)

Graham Oakes  (Upside Energy) Upside Energy's Open Innovation Platform

Andrew Haslett (Energy Technologies Institute)

Panagiotis Papadopoulos (UK Power Networks)
Smarter Network Storage Project

Clive Tomlinson  (Swanbarton)  Local energy markets versus storage in the network

1.10 Lunch including poster session

2.00 Session 5:   Storage methods and their effect on system issues

2.00 David Howey (University of Oxford) Behaviour and modelling of batteries and supercapacitors

2.40 Zofia Lukszo (Delft University of Technology)  Electric mobility in future energy systems. Car as power plant?

3.20 Tea and Coffee

3.40 Simon Tindemans (Imperial College London) Controlling a large population of smart refrigerators as a leaky storage unit  

4.00 Peter Boait (De Montfort University) Demand Shaper - an approach to domestic demand response that enhances system stability and use of renewables through biased randomisation of controlled appliances

4.20 Session 6:   Storage research directions 

4.20 Stan Zachary (Heriot-Watt and Open University) Future storage research directions

4.50 Jacqueline Edge (Energy Storage Research Network)  The ESRN and SUPERGEN Hub and future storage events

4.55 Closing remarks

5.00  Close

Speakers, titles and abstracts

David Angeli (Imperial College London)  Mean-field game formulations for distributed storage management in dynamic electricity markets

Introduction of flexible demand and distributed storage within dynamic electricity markets yields the possibility of shifting demand in order to buy energy when it is cheaper. When a multitude of such agents play this game it is no longer possible to optimize individual power absorption schedules by disregarding the population level effects that this will have on price dynamics. Mean-field games and games with a continuum of players are a way to model such scenarios. We will present some preliminary results in this respect, as well as numerical schemes for finding Nash equilibria and optimal schedules under simple rules of electricity price dynamics.

Peter Boait
 (De Montfort University) Demand Shaper - an approach to domestic demand response that enhances system stability and use of renewables through biased randomisation of controlled appliances

The ability to influence electricity demand from domestic and small business consumers, so that it can be matched to intermittent renewable generation and distribution network constraints is a key capability of a smart grid. This involves signalling to consumers to indicate when electricity use is desirable or undesirable. However, simply signalling a time-dependent price does not always achieve the required demand response and can result in unstable system behaviour. The authors propose a demand response scheme, in which an aggregator mediates between the consumer and the market and provides a signal to a 'smart home' control unit that manages the consumer's appliances, using a novel method for reconciliation of the consumer's needs and preferences with the incentives supplied by the signal. This method involves random allocation of demand within timeslots acceptable to the consumer with a bias depending on the signal provided. By simulating a population of domestic consumers using heat pumps and electric vehicles with properties consistent with UK national statistics, the authors show the method allows total demand to be predicted and shaped in a way that can simultaneously match renewable generation and satisfy network constraints, leading to benefits from reduced use of peaking plant and avoided network reinforcement. The presentation will also describe an implementation and field trial of the concept now in progress. Further detail in a journal paper at: doi: 10.1049/iet-rpg.2012.0229

Ana Bušić (École Normale Supérieure - Paris) Ancillary service to the grid using intelligent deferrable loads

Renewable energy sources such as wind and solar power have a high degree of unpredictability and time-variation, which makes balancing demand and supply challenging. One possible way to address this challenge is to harness the inherent flexibility in demand of many types of loads. We propose a technique for decentralized control for automated demand response that can be used by grid operators as ancillary service for maintaining demand-supply balance. It is assumed that there is one-way communication from the grid operator to each load. The loads are either on or off - their power consumption is not continuously variable. A randomized control architecture is proposed, motivated by the need for decentralized decision making, and the need to avoid synchronization that can lead to large and detrimental spikes in demand. A mean-field limit is used to obtain an input-output model, where the output is the aggregate power consumption.

Joint work with S. Meyn, P. Barooah, Y. Chen, and J. Ehren.

James Cruise (Heriot-Watt University) Evaluating the value of storage facilities for buffering and arbitrage

The two problems of buffering and arbitrage in general lead to very different management regimes of storage facilities. Arbitrage leads to the store emptying and filling repeatedly in line with cost fluctuations, while minimisation of shortfall events leads to trying to keep the store as full as possible. In this talk we consider these two problems in a joint framework which enables us explore the economic trade-off between utilizing a storage facility to provide both these services. As part of the associated analysis we use Lagrangian methods to qualitative understand the behaviour of the associated optimal management strategies.

Lewis Dale (National Grid)  System balancing and role of storage

Lisa Flatley (University of Warwick) Competing stores in an arbitrage market

Large-scale energy storage has the potential to play an important role in managing our future energy systems. As such, an understanding of the underlying economics is crucial. Moreover, since a major component of a store's revenue is likely to be earned through arbitrage, a consequence of storage is usually that it reduces the variation of prices. Thus, a store has market impact which in turn serves to reduce its arbitrage opportunities (and therefore also its profit). When more than one store is connected to the electricity network, then each additional store reduces the price variation still further and, as such, the behaviour of each store alters. For investment decisions, this is an important aspect which should be understood and taken into account. In this talk, we investigate the behaviour of n competing stores and their resulting market impact at the Nash equilibrium, under what we believe to be a realistic model of competition. In the special case that each store is assumed to have only a small amount of market impact (so that the action of a store on prices can be approximated linearly), we further demonstrate that the profit of each store is proportional to 1/(n+1)^2. Hence, there is a quantifiable limit to the number of economically feasible stores which can be connected our electricity network. This is based on joint work with Dr Stan Zachary and Dr James Cruise.

Nicolas Gast  (INRIA - Grenoble) Stochastic analysis of real and virtual storage in the smart grid

The electrical grid of the future will require more storage to compensate for the intermittency of distributed generators (such as solar, wind, combined heat and power). Storage will be real (batteries, water reservoirs) or virtual (demand response). In this talk we analyze the impact of storage on electricity markets, using several stochastic models. Joint work with Jean-Yves Le Boudec, Alexandre Proutière and Dan Tomozei.

Ben Godfrey (Western Power) Project FALCON - Energy Storage for a DNO

Danica Vukadinovic Greetham
  (University of Reading) A probabilistic framework for forecasting household energy demand profiles

Recent work introduced a novel time-permuting error measure for forecasts of household-level energy demand, designed to reward forecasts which predict extremes (spikes) in demand at approximately the right times, albeit perhaps slightly early or late. In many applications such as smart storage control, such forecasts are preferable to those that predict no spikes at all. Building on that idea, we make three contributions. Firstly, we introduce a probabilistic framework to estimate error distributions for actuals about forecasts in a time series, using the time-permuting error measure. The framework includes a variable discount for older, possibly less relevant data. Secondly, we employ this framework to derive conditions that need to be satisfied by the optimal forecast under the time-permuting error measure. In turn, this requires a mixture of discrete (non derivative) optimisation and calculus to condition forecasts on available historical observations. Finally, we verify the performance of our framework on forecasting the daily energy demand profiles for a large number of domestic energy customers. In particular, we demonstrate how such customers might be classified according to the relative predictability of their behaviour and the corresponding need for different amounts of history to achieve such forecasts.

David Howey (University of Oxford)  Behaviour and modelling of batteries and supercapacitors

Batteries and supercapacitors promise flexible and responsive energy storage that is useful for voltage support, load management and other grid services, and costs are decreasing. However, challenges remain in modelling aspects of their dynamic behaviour and degradation. In this talk I will explore these and other performance considerations that may impact the design and control of these systems

Zofia Lukszo (Delft University of Technology)  Electric mobility in future energy systems. Car as power plant?

Electric vehicles, including plug-in EVs and fuel cell electric vehicles (FCEVs), have a huge potential to play an important role in future energy systems to facilitate the integration of renewable energy sources. As the variable energy sources drive the need for flexibility to restore a system’s energy balance, the flexibility sources, i.e. dispatchable power plants, demand side, storage and interconnection, respond to restore that balance. FCEVs and EVs can be used to discharge electricity to the grid, and when aggregating the power of a large number of vehicles, they can function as dispatchable power plants. Plug-in EVs can adapt their charging behaviour to the needs of the power system operator, and similarly they can act as storage by charging their batteries for example, when there is a surplus of renewable energy.

Fuel cell cars, while parked, can produce electricity more efficiently than the present electricity system and with useful ‘waste’ products heat and fresh water. The Car as Power Plant system has the potential to replace all electricity production power plants worldwide, creating an integrated, efficient, reliable, flexible, clean and smart energy and transport system. In terms of technology, the energy production system can be envisaged as a fleet of fuel cell vehicles, where cars while parked (over 90% of the time) can produce with the fuel cell electricity, heat and fresh water, which will be feed into the respective grids. From a social perspective the stakeholders directly and indirectly involved in the design, building and operation of such a system, are car park operators, the local power, heat and water distribution companies, gas suppliers, H2 producers, the equipment, system and software manufacturers but also municipalities, regulators, policy makers and not to forget the car owners/users. The presentation will conclude with an analysis of the feasibility of Car as Power Plant system as a detachable decentralized multi-modal energy system.

John Moriarty (University of Manchester)  Bayesian inference on a jump-diffusion Ornstein-Uhlenbeck process for modelling electricity spot prices, and applications to the economically optimal control of CHP and heat storage

The class of superposed Ornstein-Uhlenbeck (OU) processes has become established in modelling electricity spot prices. In addition to modelling the mean reversion common in commodity prices, such models are able to capture both short- and long-term price variations via diffusion, jumps or a mixture of the two, depending on the choices of driving noise. I will describe a fully Bayesian methodology for the calibration of superposed OU models using a Markov Chain Monte Carlo (MCMC) procedure. We take care to develop a consistent and unified method making minimal assumptions so that the characteristics of the market itself are revealed more fully. In particular we examine how many OU processes should be superposed, with what choices of driving noise, and how these price components should be interpreted, by studying both the APXUK and EEX spot markets. Since electricity spot price models are used in the operational modelling of flexible energy systems under price uncertainty, we conclude by drawing out the implications of our work for the economic optimisation of CHP with heat storage. This is joint work with Jan Palczewski (Leeds) and Jhonny Gonzalez (Manchester).

Graham Oakes  (Upside Energy) Upside Energy's Open Innovation Platform

Upside Energy aims to enable households and small businesses to participate in payments for demand response. We're building a cloud platform that aggregates energy stored in uninterruptible power supplies, solar PV systems, electric vehicles, domestic heating systems, etc. We'll then use that energy to provide balancing services to the grid. Upside was created for a challenge prize run by Nesta and the National Grid. We've recently won funding from Innovate UK and DECC to build a pilot of our systems. A key part of that project will be building an 'open innovation' platform to enable researchers to access our data and compute facilities and hence to develop, evolve and monetise algorithms for grid prediction, storage portfolio optimisation, etc. This presentation will describe the platform we're building and our plans to engage with the research community through our DECC funding.

Jan Palczewski  (University of Leeds) American contracts for power system balancing

We study utilisation of storage for balancing of power systems. This is formulated as repeated issuance of American-type real options on physical delivery/consumption of power. Using methods of optimal stopping we derive analytically optimal strategies for management of the installed storage: optimal conditions to trade power in the spot market and optimal time to offer a service to a network operator. These results enable assessment of the profitability of the storage for power system balancing for the battery operator and for the network operator. Joint work with J. Moriarty and D. Szabo.

Goran Strbac (Imperial College London) Quantifying the value of energy storage in supporting integration of renewable generation

Phil Taylor (University of Newcastle) Integration and control of energy storage in distribution networks

Simon Tindemans (Imperial College London) Controlling a large population of smart refrigerators as a leaky storage unit   

Refrigerators and other thermostatically controlled loads (TCLs) have power consumption requirements that are not tightly coupled to usage patterns. Instead, these appliances need only to satisfy averaged power consumption constraints in order to maintain their target temperatures. As a result, TCLs such as refrigerators and air-conditioners collectively represent a large flexible demand resource, but it has proven challenging to deploy this resource in an optimal way due to issues with duty cycle synchronisation and practical communication and computation constraints. I will present an end-to-end solution to this challenge that enables a population of TCLs to act as an electric energy storage unit, without onerous communication requirements. This solution consists of (1) a decentralised controller that accurately controls the aggregate TCL power consumption and (2) a corresponding system-level representation as a 'leaky storage unit'. The storage representation allows for convenient embedding in system-level dispatch algorithms, which can optimally allocate the flexible resource represented by the TCLs. This leaky storage model is both sufficient and exact in the limit of large appliance numbers, even for a heterogeneous population. This implies that a central dispatcher (e.g. flexible demand aggregator) is able to determine a relative power consumption patterns that is guaranteed to be implementable, without knowledge of the states of individual appliances. In turn, the appliances implement a stochastic controller that tracks the desired power consumption in expectation, even though individual devices can only be on or off. This way, the diversity of appliances is maintained without the need for a real-time communication and control infrastructure. This is joint work with Vincenzo Trovato and Goran Strbac.

Clive Tomlinson  (Swanbarton)  Local energy markets versus storage in the network

Energy storage in distribution networks as a means of enabling distributed renewable generation has been widely discussed. However, little attention has been given to how such storage may be made economically viable. Present energy market models are inimical to deployment of storage in the network. As an alternative, this paper introduces the mechanics of local energy markets in which energy prices are set freely between participants, in real time. It reports modelling, simulation and laboratory trial of such markets. It shows that a local energy market can enable financial return for energy storage when deployed as a commercial instrument. It incidentally demonstrates that where a local energy market coincides with a low-voltage feeder, commercial storage management tends to reduce peak flows at the constrained interface between the low-voltage feeder and the distribution network. Increased densities of distributed generation are, therefore, made possible. Trading intervention by network operators may influence storage behaviour in ways that reduce the peak flows further

Poster presenters, titles and abstracts

Where available the poster may be obtained from the poster title.
John Brooke (University of Manchester) Modelling the behaviour of energy systems using computational steering

In terms of understanding the response of distributed energy systems (represented as large dynamical systems) we can utilise observations of the system via sensors to steer the trajectory of the simulation of the energy system. Such steering raises questions about the stability and accuracy of models steered in this way, in terms of how often the updating of the simulation is performed and also to ensure that partial coverage of the system by sensors (since we cannot fully instrument the system for cost reasons) does not distort the trajectory of the dynamical system so that it becomes less representative of the real state of the system, as opposed to the desired goal of steering the simulation so that it tracks the evolution of the physical system it is designed to represent. Such issues are important in real-time modelling of energy systems.

Qiong Cai  (University of Surrey) Optimal control strategies for integrating electrolysers with intermittent renewable energies

The penetration of intermittent renewable energies demands the development of energy storage technologies. High temperature electrolysis using solid oxide electrolyser cells (SOECs), as a potential energy storage technology, provides the prospect of a cost-effective and energy efficient route to clean hydrogen production. This paper presents a study of optimal control strategies for hydrogen production based on SOEC technology, with the aim of offering efficient large-scale system operation when coupling an SOEC system with renewable energy sources. The system model used includes a 1D dynamic SOEC stack model and an air compressor model, to examine hydrogen production in relation to energy consumption. Control strategies considered include maximizing hydrogen production, minimizing SOEC energy consumption and minimizing compressor energy consumption. Optimal control trajectories of the operating variables over a given period of time show feasible control for the chosen situations. Temperature control of the SOEC stack is ensured via constraints on the overall temperature difference across the cell and the local temperature gradient within the SOEC stack, to link materials properties with system performance; these constraints are successfully managed. The relative merits of the optimal control strategies are revealed.

David Greenwood (Newcastle University) Smarter network storage

Energy Storage (EES) has many applications within power networks. However, installing EES for a single application or to solve a single problem will rarely justify the required investment. Consequently, large scale EES is more likely to succeed if it engages in a variety of network and commercial services. The challenges include: determining the appropriate size of the EES in terms of power and energy, prioritizing which services to engage in, selecting the most profitable commercial services and managing the battery's power and energy resources. The solutions presented demonstrate that this deployment of EES is realizable, and provides a route to wide-scale adoption.

EngTseng Lau  (Brunel University) Carbon savings in demand side response programmes

We quantify carbon savings (tonnesCO2) in the demand side response (DSR) programmes. We consider DSR programmes such as Short Term Operating Reserve (STOR), Triad, Fast Reserve and Smart Metering, with various types of smart interventions involved (using diesel generators, hydro-pumped generation and use of tariffs) addressing the substitution of business-as-usual (BAU) energy consumption in energy networks. We model each of the DSR programmes with configurations and assumptions appropriate for the energy industry. This enables us to compare carbon emissions between the BAU solutions and the smart intervention applied, thus deriving the carbon savings. Whether such DSR produces positive CO2 savings or not depends on the used technologies, as well as the scale of the interventions.

Vanaja Rao (University of Central Lancashire) Improving the efficiency of the flywheel energy storage system with wind power generation by using neural network based controller

This PhD project investigates and introduces a power control strategy of a Flywheel Energy Storage System (FESS) based on an Artificial Neural Network (ANN). These types of storage systems are useful in grid connected applications such as grid frequency support and control, power conditioning and provision of back-up power. Due to the proliferation of non-linear loads, the utility becomes more vulnerable to disturbances such as voltage sags, unbalanced power flow and frequency fluctuations. This is also the case with the source of power being a renewable power generation such as wind power generation. It gives rise to high instability across the power network. Therefore, energy storage systems have become an essential part of electrical power utilities as they provide a higher level of power quality and stability. Flywheels as energy storage devices exhibit high performance with grid connected applications such as power conditioning, frequency regulation and voltage sag compensation. This is due to their capability of storing energy in the form of kinetic energy depending on the rotating speed and their moment of inertia. A good control system is therefore required to regulate the flywheel operation so that there is enough capacity to absorb a possible upcoming regeneration peak. Similarly, there should be enough reserve energy available to feed a demand peak. The proposed system is a simple FESS based on a permanent magnet pancake motor. The controller is designed to avoid machine overloading while the flywheel is charged or discharged. Additionally, it avoids using the required outer power loop or a hysteresis power controller, hence, simplifies the overall control algorithm. The validity of the developed control system is investigated via computer simulations using Simulink. The proposed system is also compared with conventional power control strategy with an additional outer power control loop to highlight their importance. Development of a fully parametric overall simulation model, to optimize a Flywheel is the main focus of this PhD research. An intelligent control system using the Fully Connected Cascade (FCC) Neural Network based learning algorithm for this model is demonstrated here. The entire system level integration of a FESS based on FCC neural network gives an adaptive controller. The empirical results obtained will therefore establish the objective of this research, which contributes to the increase in efficiency of the FESS and provide better control in its operation across the power network.

Renaldi Renaldi  (University of Edinburgh) Optimisation of energy systems with thermal energy storage

Thermal energy storage (TES) can support the decarbonisation of heating system, for example by alleviating the intermittency and supply-demand mismatch issue inherent in renewable-based heating system. However, the integration of TES significantly increases the complexity of design and operational optimisation of energy systems. At present, there are no widely accepted rules regarding the level of detail of the energy systems optimisation model required to properly reflect the operational characteristics of TES while managing acceptable problem complexity. This project involves the development of an optimisation framework for energy systems with TES. Models are developed starting from a single dwelling and moving towards a district energy system. The single dwelling optimisation work illustrates the design and operational optimisation of a domestic heating system consisting of an air source heat pump coupled with TES. The optimisation results show that the integration of TES and a time-of-use tariff reduce the operational costs of the heat pump system. Various type of TES, both short- and long-term storage, are currently being modelled with different complexities. Furthermore, an optimisation framework capable of handling multiple time scales is under development. The single dwelling model will then be integrated in this framework to build a district energy optimisation model.

Andrew Smith (University College London) Quick and dirty: a web-hosted storage model 

Sometimes a very crude model is enough to illustrate the challenges in scenarios with high penetration of exogenously-variable renewables. The particular crude model presented for the first time at this workshop is based on the National Grid half-hourly data for PV, wind and demand.