Search

The wholesale electricity markets in the United States operate under the two-settlement system, comprising the day-ahead (DA) market and the real-time (RT) market. The DA market clears bid-in supply against bid-in demand and determines DA physical schedules for generators, virtual awards, and DA locational marginal prices (LMPs), which are defined as the marginal costs of serving the next increment of demand at pricing nodes consistent with existing transmission constraints and performance characteristics of generation resources. The RT market procures "balancing" energy to meet the forecast RT grid energy demand and determines RT dispatch signals for resources and RT LMPs. The introduction of virtual bids to electricity markets is to mitigate the discrepancy between the DA market and the RT market in LMP spreads. In this dissertation, we focus on developing a machine learning-driven portfolio optimization framework for virtual bidding in electricity markets considering both risk constraint and price sensitivity. The algorithmic trading strategy is developed from the perspective of a proprietary trading firm to maximize profit. A recurrent neural network-based LMP spread forecast model is developed by leveraging the inter-hour dependencies of the market clearing algorithm. The LMP spread sensitivity with respect to net virtual bids is modeled as a monotonic function with the proposed constrained gradient boosting tree. We perform a comprehensive empirical analysis on PJM, ISO-NE, and CAISO with the proposed framework. We further introduce a similar framework to arbitrage congestion with virtual bids. A deep neural network is designed to estimate the difference between congestion spreads in DA and RT markets. A clustering algorithm is adopted to separate pricing nodes into a few groups, between which the congestion spreads can be exploited. We validate the proposed algorithmic trading strategy using publicly available data from CAISO.The electric power system is a major contributor to greenhouse gas (GHG) emissions. To reduce GHG emissions, accurate emission predictions are essential. The marginal emission factor (MEF) is a useful signal for distributed energy resource aggregators and end-use customers to mitigate GHG emissions by scheduling the flexible loads accordingly. We propose a hybrid machine learning framework to predict GHG emissions and locational MEF, which integrates feed-forward neural networks (FNNs) with spatio-temporal graph convolutional networks (STGCNs). A comprehensive case study in CAISO shows that the proposed approach outperforms the existing techniques in prediction accuracy. The proposed model provides short-term locational MEF predictions with high time granularity using only publicly available data.

Electricity generation is a significant source of global carbon emissions and a major driver of climate change. In the interest of mitigating climate change impacts, there is substantial interest in minimizing the carbon footprint associated with energy generation through demand-side shifts and supply-side transitions to cleaner power sources. Effective implementation of these measures requires a thorough spatiotemporal understanding of marginal emission factors (MEFs) that are associated with the generators that respond to changes in electricity demand. In this study, we present a hybrid model that estimates MEFs in real-time and forecasts hour-ahead values at a zonal level. The model proposed combines the capabilities of a dispatch model to capture the physical characteristics of the electricity grid with an advanced machine learning algorithm to project these values in the future. This analysis focuses on the NYS electricity grid, based on a publicly available digital twin of the NYISO. The proposed model provides insights into seasonal and diurnal patterns of zonal MEFs and the expected future behavior to support informed decision-making toward reducing carbon emissions.

Prosumers, with the ability to act both as a supplier and a consumer in a power market, have received considerable attention recently. Having distributed energy resources, their capability to operate in an isolated mode, separated from the main grid, has also been promoted as a vital option to enhance the power system's resilience. One emerging concern is the prosumer's ability to manipulate the power market as a buyer or as a seller. This thesis vets the outcomes of a power market in presence of strategic prosumers and formulates electricity markets in different frameworks depending on the prosumer's strategy to obtain the equilibria representing the market outcome.This thesis posits a situation in which a strategic prosumer owns a renewable unit with variant output and a dispatchable backup unit, and participates in a power market following a price-taker, quantity-based, or Stackelberg strategy. The prosumer is assumed to maximize its benefit by deciding amount of power to buy from or sell into the main grid, amount of renewable power to forego consumption, and amount of power to produce from backup unit. The interaction of prosumers and the main grid is modeled through shifting the residual supply curve in the wholesale market, thereby avoiding the possible numerical issues when aggregating their demand horizontally with consumers. The model is applied to a case study of the IEEE 24-bus RTS as an illustrative example. Prosumers in the market are also subject to uncertain renewable output and, as they participate in the day-ahead market, decide how much energy to consume, sell, or purchase, considering the potential imbalance due to uncertain output. The question of prosumer's risk preferences, taking into account the uncertainty in its renewable output, is investigated and the dynamics between prosumer's risk decisions and its implications for the prosumer's profit maximization are examined in a chance-constrained framework.This thesis illustrates power market outcomes in presence of strategic prosumers following price-taker, quantity-based, and Stackelberg strategies exercised by the prosumer and shows existence of market-clearing equilibria in the case of perfect and imperfect competition and demonstrate that the prosumer's position in the market does not depend on its strategy along with the fact that the prosumer is able to attain higher pay-off in the price-taker case compared to the quantity-based strategy. This thesis also shows the impact of prosumer's Stackelberg strategy on rent distribution and price formation in the marketplace in a chance-constrained framework to account for uncertainty of its renewable output. Finally, the thesis investigates prosumer risk attitudes and quantifies the impact of renewable output uncertainty and imbalance prices on prosumer's risk preferences.

Abstract: With continued growth in global carbon emissions, much of the developed world is curtailing its dependence on fossil fuels. Broad decarbonization will not come from a single source, but rather a combination of policy, technology, and innovation. Wind power has seen significant development in Canada, having over 6,700 operational turbines installed over the last decade. Increased investment into wind energy has lead to improvements in the technology, with wind turbines roughly doubling in height since the early 1990’s. In recent years, hydrogen has seen a resurgence in popularity as a low carbon energy carrier. Pairing existing wind farms with energy storage, such as hydrogen, could increase their electricity revenue through price arbitrage, or generate new revenue through selling the hydrogen. This work examines the opportunity to generate green hydrogen in Alberta using a wind-hydrogen hybrid plant. To do so, Alberta’s electricity market operations were forecast to 2030 using commercial market simulation software, Aurora, including hourly prices and unit dispatch. To better simulate price spikes, electricity price perturbations, achieved through time series decomposition of historic price signals, are administered as random shocks to the hourly price forecasts. Wind-hydrogen hybrid plant characteristics, capacity and operating schedule, are optimized using linear programming. Results of this work show some potential for wind-generated green hydrogen to be cost competitive with steam-methane reforming in the next decade, with some scenarios achieving levelized cost of green hydrogen below $2/kg. Regardless of market conditions, optimal electrolyser sizing is contingent on achieving a capacity factor between 40% and 60%. At current costs, energy arbitrage through hydrogen fuel cells is not financially viable.

The sheer growth of electricity demand and the rising number of electricity-hungry devices have highlighted and elevated the need of addressing the demand response management problem in residential smart grid systems. In this thesis, a novel contract-theoretic demand response management (DRM) framework in residential smart grid systems is introduced based on the principles of labor economics. The residential households produce and consume electricity, acting as dynamic prosumers. Initially, the prosumers' personal electricity generation and consumption characteristics are captured by introducing the concept of prosumers' types. Then, the prosumers' and the electricity market's profit is depicted in representative utility functions. Based on the labor economics principles, Contract Theory is adopted to design the interactions among the electricity market, which offers personalized rewards to the prosumers in order to buy electricity at an announced price, and the prosumers, who offer their "effort" by paying for the purchased electricity. The contract-theoretic DRM problem is formulated as a maximization problem of the electricity market's utility, while jointly guaranteeing the optimal satisfaction of the prosumers, under the scenarios of complete and incomplete information from the electricity market's perspective regarding knowing or not the prosumers' types, respectively. The corresponding optimization problems are solved following a convex optimization approach and the optimal contracts, i.e., rewards and efforts, are determined. Detailed numerical results obtained via modelling and simulation, highlight the key operation features and superiority of the proposed framework.

This work seeks to quantify the technical and economic feasibility of integrating nuclear power plants with renewable energy generation and heat storage. A variety of technoeconomic models were developed for this purpose, including cash-flow based project financing models; capital cost scaling models for heat storage systems; thermodynamic models of heat storage charging, discharging, and standing losses; an electric grid unit commitment and economic dispatch model for modeling prices; and statistical models of load, wind, and solar PV at various locations for Monte Carlo market analyses. Based on the assumptions of recent market conditions in three major electricity markets in the United States, all systems evaluated would not be economically viable. Average electricity prices varied from around $25/MWh (ERCOT) to $42/MWh (Mid-Columbia) and $54/MWh (PJM). For integrating a new nuclear power plant with wind and hydrogen production, the very high capital costs of the nuclear plant led to multi-decade payback times, and the assumed cost of hydrogen at $2.50/kg was too low to justify the capital investment in the associated chemical plant. For the heat storage systems evaluated, power-related capital costs varied $1,200–2,300/kW, and energy-related capital costs ranged $36–140/kWh. Increasing the size of the power block had significant positive effects on net revenue, while increasing the size of the energy storage reservoir was generally not helpful beyond five hours of storage time. However, when considering the entire cash flow of a system beyond its operating revenue, no heat storage system had enough free cash flow to break even in the markets and scenarios analyzed, with the smallest system having the best overall economic performance. In contrast to other studies where only a single representative year of load and weather data is used, this work showed that the historical variability of those time series can lead to significant spreads in the distributions of results. Selecting a “lucky” year could yield revenue estimates three times greater than the median. However, the results were more sensitive to the capital cost assumptions than to the weather year. Integrating renewable energy generation and/or heat storage with nuclear power plants is unlikely to be economically feasible in the near term without significant changes in energy markets, significant decreases in capital costs of power turbines, or significant increases in development or operating subsidies for these system concepts.

A smart grid can help utilities conserve energy, cut costs, increase reliability and accountability, and increase efficiency in processes. Nevertheless, the growing use of IT-based electrical power systems increases the vulnerabilities of cyber security causes which increases the importance of cyber security in interconnected power grids. Utilities should understand protection of the smart grid, including vulnerable areas, strategic issues, layered security. However, there is a lack of awareness over the country since the inhabitants less informed about its importance in our life to transform into a smart grid and being part of this plan. Knowing that the government started already contributing to be part of the smart grid, the whole country should be awarded of its significance and the strategies consequences of such vital change. This project aims to build a program that is compatible to all smart phones where it can provide clear information of the smart grid and ways to be part of it. The program will have an interactive interface showing the user the consumption of electricity instantaneously. Besides, for the providers, the program will show how are they contributing to the grid and ways of reducing the electricity bills. This project discusses important information about the smart grid and features that the program should have to provide a better experience to the customers but mainly the large energy users that play a relatively larger rule in consuming electricity.

Electric power grids are becoming increasingly stressed due to political and environmental difficulties in upgrading transmission capacity. This challenge receives even more interest with the paradigm change of increasing renewable energy sources and demand response (DR) programs. Among DR technologies, existing DR programs are primarily designed for industrial and commercial customers. However, household energy consumption accounts for 38% of total electricity consumption in the U.S., suggesting a significant missed opportunity. This dissertation presents an in-depth study to investigate managing power system congestion and residential DR program under uncertainty.First, an interval optimization model is presented for available transfer capability (ATC) evaluation under uncertainties. The conventional approaches of ATC assessment include deterministic and probabilistic methods. However, the proposed interval optimization model can effectively reduce the accuracy requirements on the renewable forecasting, and lead to acceptable interval results by mitigating the impacts of wind forecasting and modeling errors. Second, a distributed and scalable residential DR program is proposed for reducing the peak load at the utility level. The proposed control approach has the following features: 1) it has a distributed control scheme with limited data exchange among agents to ensure scalability and data privacy, and 2) it reduces the utility peak load and customers’ electricity bills while considering household temperature dynamics and network flow.Third, the impacts of weather and customers’ behavior uncertainties on residential DR are also studied in this dissertation. A new stochastic programming-alternating direction method of multipliers (SP-ADMM) algorithm is proposed to solve problems related to weather and uncertain customer behavior. The case study suggests that the performance of residential DR programs can be further improved by considering these stochastic parameters.Finally, a deep deterministic policy gradient-based (DDPG-based) HVAC control strategy is presented for residential DR programs. Simulation results demonstrate that the DDPG-based approach can considerably reduce system peak load, and it requires much less input information than the model-based methods. Also, it only takes each agent less than 3 seconds to make HVAC control actions. Therefore, the proposed approach is applicable to online controls or the cases where accurate building models or weather forecast information are not available.

Electricity markets are particularly complex because they must accommodate the underlying physics that govern the electric power system. These physics present non-convexities in the social welfare maximization problem, also called the economic dispatch problem, solved by the Independent System Operator (ISO), which is the social planner in this context. The non-convexity of this problem presents difficulties in computing the social welfare maximizing dispatch as well as difficulties in deriving a pricing structure that satisfies certain economic requirements such as revenue adequacy of the ISO and non-negative operating profits for market participants. This dissertation analyzes two sources of non-convexity that pertain to two separate market changes that have been recently proposed in Texas. Both proposals pertain to the real-time electricity market, which clears every 5-minutes and is myopic in the sense that only the demand at the end of the upcoming 5-minute interval is considered and no future time intervals are considered in the social welfare maximization problem. The Electric Reliability Council of Texas (ERCOT) is the ISO in Texas and currently neglects resistive losses along transmission lines when formulating the economic dispatch problem. The first part of this dissertation regards a proposed market change to incorporate transmission losses into the economic dispatch problem. Two general approaches are considered to accommodate associated non-convexity. Similar to current practice, the first approach is based on a marginal pricing structure and uses convex approximations that facilitate efficient computation. By utilizing various approximations, the aforementioned economic requirements are proven to be satisfied approximately. The second approach is based on an alternative pricing structure in which prices are chosen to explicitly minimize the worst-case violation of these economic requirements. For example the prices may be chosen to minimize the potential revenue shortfall of the ISO. These alternative prices are termed convex hull prices and can be approximated by use of convex relaxations. The economic dispatch problem currently used by ERCOT does not endogenously represent operating reserves to handle contingencies that may occur. Instead, operating reserves are currently optimized separately from the electric power generation dispatch. The second part of this dissertation regards a proposed market change to co-optimize reserve and generation dispatch in a social welfare maximization problem called a co-optimization problem. Implementation of the real-time co-optimization problem is being pursued simultaneously with a new definition of the primary frequency responsive reserve types considered in the market. One of these reserve types intends to accommodate standard droop control. Another of these reserve types is newly introduced and intends to facilitate participation of fast-acting batteries in primary frequency response. This dissertation derives reserve requirements from first principles that capture the coupling of these two reserve types as well as their ramping abilities. The newly proposed non-convex requirements represent limits on the ramp-constrained primary frequency responsive reserve procurement. Placing these non-convex requirements into a co-optimization problem is proven to result in the satisfaction of the aforementioned economic requirements.

In this dissertation, we study some problems concerning the integration of renewable energy. First, we design and analyze a two-stage mechanism for selling a stochastic resource where penalties are paid at the second stage to compensate for shortfalls in the amounts contracted in the first stage.We consider a setting in which a renewable generator, whose generation is a random variable with some probability distribution, is selling its generation to a number of flexible buyers through a two-stage electricity market. The system operator runs an auction in which the seller bids its probability distribution and the buyers submit their valuation functions. The system operator inputs these bids to a stochastic program to decide on an efficient allocation that maximizes social welfare, and then uses Myerson's payment to price electricity.The closed-form solution we obtain for the allocation and payment rules of our mechanism allows us to analyze the mechanism and prove its incentive compatibility in dominant strategies, individual rationality, and budget balancedness.We also investigate the sensitivity of the mechanism to the seller’s manipulation of its distribution. We show through running some simulations that increasing the number of buyers in the system reduces the gains the seller obtains from manipulation.In this work, we also analyze the effects of limited ramping capacity on the price of electricity. We study the problem of dynamic economic dispatch with ramping constraints and stochastic demand, which can ramp up or ramp down significantly due to the presence renewable generation. In conclusion, we show the value of investing in ramping capacity and highlight the possibility of exercising market powers by generators monopolizing it. Our study of the evolution of prices in this setting leads to the understanding that limited ramping of some generators enhances the market power of the flexible ones.

Benefitting from the modernization of power system, electricity end-users, especially residential customers, have played more active roles in the demand side. With authority to choose among a diversity of electricity plans and knowledge of household power consuming patterns, residential customers are inspired to employ advanced cost management strategies, e.g. demand response (DR) and the switch in electricity plans, to reduce electricity expenditures. Considering the unavoidable limitations of DR, this thesis focuses on the strategy of switching plan, with the development of electricity plan recommender system (EPRS). EPRS recommends cost-effective electricity plans identified among all the active candidates according to preferences of neighbours, households similar to the user of EPRS in power consuming pattern. This principle is firstly achieved through clustering-based EPRS (C-EPRS), the user of which is ascribed to one of the predefined clusters and attributed with neighbours as all the cluster members. Electricity plans generally preferred by these neighbours are then selected for recommendation. Followed by C-ERPS, neighbourhood collaborative filtering-based EPRS (NCF-EPRS) is developed. With modified neighbourhood-based collaborative filtering (CF) algorithm, NCF-EPRS compares its user to other households through a novel similarity metric and employs the obtained similarities in neighbour detection and plan assessment. Considering the sparsity of practical electricity datasets, hybrid CF-based EPRS (HCF-EPRS) is constructed next by expanding NCF-ERPS with model-based CF algorithm for missing feature estimation. Multiple similarity metrics requiring dense inputs are allowed in HCF-EPRS. Finally, Bayesian HCF-EPRS (BHCF-EPRS) is proposed by applying Bayesian inference in feature estimation model and in relevance vector machine-based user classifier modelling to improve the time efficiency and discretising variables to receive higher feasibility. The efficiencies of all the EPRS models, especially the superiority of BHCF-EPRS, are evaluated through case studies in two real electricity markets. The resulting EPRSs are applicable for ordinary residential customers to identify cheap electricity plans and for retail companies to assess cost competitiveness in practical markets. With the profits brought by EPRS, both customers and retailers are motivated in retail competitions, which further contributes to the deregulation of the electricity market.

The primary aim of this dissertation is to deliver a technique to augment power system test cases with realistic open-source data to represent a deregulated power system. These test cases are intended to be used by power system researchers who require a test case that is capable of performing economic and environmental analysis on a bulk-power level. These test cases are capable of estimating the cost of bulk-energy for economic analysis and harmful greenhouse gas (GHG) and air polluting (AP) emissions for environmental sustainability analysis. These cases are developed for simulations that are intended to be at the transmission level where the independent system operator (ISO) has control. In the second part of this dissertation, an aggregator based demand response (DR) model is studied as-a-service to the bulk-power market, and its economic benefit is estimated using the augmented test cases. The augmentation technique presented in this dissertation has three-layer data over the existing generator information in a test case. The first layer of augmented data replaces the cost functions of the test case generators with functions developed based on the generator offers from a real electricity market. An unsupervised learning technique had to be implemented to classify the market offer data because the identity of the generators is masked to honor a fair market policy. The offer data was converted to cost functions and is sampled statistically such that the test cases represent a similar generator supply curve as the real power system. In addition to the cost functions layer, the test case generator data has an augmented generator fuel-turbine data. This data in a test case will represent the energy sources and generator technology of the system that the test case is intended to emulate. The hourly energy mix of the electricity market is utilized to augment the generator fuel-turbine type of test case generators. Because the number and capacities of test case generators may not represent the real system, assigning one fuel-turbine type to one test case generator will not result in a right energy mix. The augmentation technique creates an additional layer of information for each test case generator which can represent multiple fuel-types. The third layer of augmented data on test cases contains the heat curve and emission information. With all these layers of data, the test case is capable of representing the dynamic cost nature of a deregulated power system and is able to dispatch generators similar to the real power system. PJM interconnection data was chosen to implement the proposed augmentation technique. The marginal cost result from optimal power flow (OPF) is compared with the marginal cost of energy of the PJM interconnection along with the GHG and AP emissions. Smart-grids have opened opportunities for end customers to participate in the power system operation. DR is one of the activities that the end customers can perform to participate in the electricity market. Revenue earned from energy markets has been relatively low compared to DR used for capacity markets and ancillary services. An aggregated DR model participating in the bulk-power market as a service through a pool-based entity called demand response exchange (DRX) is proposed to improve the benefits of DR to the market. The economic benefits to the market entities have been studied using the proposed augmented test cases. The key contributions of this dissertation are: power systems test case generator data for researchers who do not have access to the real power system data, a technique that utilizes only open-source data to develop augmented data for any test case to represent the dispatch of a real power system in terms of cost, and emissions, a DR model capable of improving the revenue for DR participants in the bulk-energy market.

On top of Smart Grid technologies and new market mechanism design, the further deregulation of retail electricity market at distribution level will play a important role in promoting energy system transformation in a socioeconomic way. In today’s retail electricity market, customers have very limited ”energy choice,” or freedom to choose different types of energy services. Although the installation of distributed energy resources (DERs) has become prevalent in many regions, most customers and prosumers who have local energy generation and possible surplus can still only choose to trade with utility companies.They either purchase energy from or sell energy surplus back to the utilities directly while suffering from some price gap. The key to providing more energy trading freedom and open innovation in the retail electricity market is to develop new consumer-centric business models and possibly a localized energy trading platform. This dissertation is exactly pursuing these ideas and proposing a holistic localized electricity retail market to push the next-generation retail electricity market infrastructure to be a level playing field, where all customers have an equal opportunity to actively participate directly. This dissertation also studied and discussed opportunities of many emerging technologies, such as reinforcement learning and deep reinforcement learning, for intelligent energy system operation. Some improvement suggestion of the modeling framework and methodology are included as well.

With the continuous development of Smart Grid, especially the emergence of Energy Internet, there is an increasing amount of measurement data available collected from power system end-users. Through data mining techniques, these measurement data can enable a better understanding of the load composition and end-user consumption behaviours, and therefore would provide great potentials for developing more flexible and targeted or even customized pricing schemes for electricity retail. This research starts with a comprehensive literature survey on decision-making for electricity retailers. Publications on electricity retailing in the last two decades are surveyed and discussed in detail. Then, key business framework of electricity retailers is studied. It elaborates the typical business process of electricity retailers and its procedure of creating a new sales agreement. Considering the drawbacks of existing load data mining methods, a new non-intrusive load monitoring method is proposed which is able to cope with the big load data in the Smart Grid environment. After obtained the status of all identified appliances, a statistical residential load model is developed. With this load model, the appliance identification results can be conveniently used in demand-side management and developing electricity retailing strategies. Next, this research proposes the idea of using the results of residential appliance identification and end-user behaviour analysis to help retail pricing. The problem of designing customized pricing strategies for different residential users is investigated based on the identification results of residential electric appliances and classifications of end-users according to their consumption behaviours. A novel framework of customizing electricity retail prices is proposed. When to customize retail prices through appliance identification, load data at least sampled at every minute is needed. Differently, this research explores another data mining technique to customize electricity retail prices using the half-hourly sampled electricity consumption data. A model of customizing electricity retail prices based on load profile clustering analysis is developed. Electricity usage data collected by the Smart Grid, Smart City (SGSC) project in Australia is used to demonstrate the feasibility and efficiency of the developed models and algorithms.

With the increasing integration of variable renewable energy, the electricity market supply and demand dynamics is expected to shift accordingly. Consequently, the renewable share in the global total energy mix is predicted to grow substantially over the next decade. This will boost the required capacity for the frequency regulation ancillary services and lead to an increase in variable renewable energy curtailment level if no action is taken. On the other hand, with the advance in the Energy Storage System (ESS) technologies, ESS is expected to play an important role in the electricity market in a high renewable energy penetration future. However, the economic viability of the ESS depends on both the revenue streams of the market applications and its operation cost profile (i.e. calendar life cost and cycle life cost). In this thesis, we firstly propose a dynamic ESS revenue estimation method and investigate and quantity the ESS’s revenue opportunities in the electricity spot market (price arbitrage) and frequency regulation ancillary service markets while the ESS is assumed to have a quick charge and discharge capability. While electric vehicle (EV) is gaining more popularity, a collection of retired EV battery packs provides an economic option for meeting the additional frequency regulation needs. We then propose a ESS market operation model that takes into account of both the revenue streams of the spot market and the frequency regulation ancillary service markets, and the ESS operation cost profile (a battery operation cost estimator is built to evaluate the potential impacts of market operations on battery lifespan). Specifically, the model is designed for retired EV lithium batteries under the Australian national electricity market framework. Finally, a hybrid ESS sizing framework is proposed for the combined wind-storage system. The operation is fully driven by the revenue opportunities including price arbitrage and storing curtailed wind energy and the ESS cost profile. The hybrid ESS sizing framework scales two distinctive types of ESSs (collectively a hybrid ESS) simultaneously and is analysed under different market frameworks and generation mixes.

This thesis models consumer behavior in electricity markets from both a theoretical and practical perspective. The main underlying theme of this thesis is residential demand response, which is enabled by the development of smart grid technologies and the subsequent collection of large amounts of data. The work in this thesis touches on themes at the intersection of machine learning, economics, and game theory.In the first part of this thesis, we seek to analyze the interaction between different players in the modern smart grid. Topics discussed include the incentivization of residential users to elicit their private information, how peer effects affect residential energy consumption, as well as hedging strategies against quantity and price risks in the electric market. To support these expositions with even more fundamental analyses, a framework for budget-constrained and combinatorial multi-armed bandits is introduced. The motivation behind including this rather generic topic lies in the sequential and repetitive nature of interactions between different players in the smart grid, which - under certain assumptions - could be captured with this methodology.The second part of this thesis is concerned with the estimation of a residential Demand Response program carried out by OhmConnect, Inc. during a 14-month period. To evaluate the effects of monetary (and non-monetary) incentives on the reduction in electricity consumption, we first develop a short-term load forecasting method and compare various estimators. The estimation accuracy is further improved by incorporating mixtures of Gaussians and Hidden Markov Models into the estimators under consideration. Next, we develop a two-stage estimation framework to estimate individual treatment effects of Demand Response and compare the aggregated effect, namely the Average Treatment Effect, to the outcome of a randomized controlled trial. The ability to estimate individual treatment effects allows us to design an adaptive targeting framework, which seeks to maximize cost efficiency of this program. Lastly, the effect of moral suasion (non-monetary incentives) that only appeal to the environmental consciousness of users is explored.

Energy storage systems (ESS) have the potential to reconfigure how the electricity system is used, operated, and expanded. Most research on grid-connected ESS is focused on applications related to renewables integration and system reliability; much less is written on the current economic uses (e.g., peak shaving and energy arbitrage). While these latter applications may be profitable, current literature suggests they tend to increase grid emissions. This need not be the case. In this paper, I explore varying system resource mixes and ESS operational modes that enable carbon-neutral, or carbon-reducing, usage. Specifically, I model the carbon emissions induced by energy storage operated in three ways – energy arbitrage (EA), demand charge management (DCM), and carbon minimization (MinCO₂) – in 16 simulated electricity systems where wind and solar assets generate 17% to 81% of annual energy. Dispatch of a 1MW/4MWh battery is simulated for each operational mode and in each resource scenario (for a total of 64 combinations). I find that energy storage is carbon-neutral, or carbon-reducing, in systems generating 17% to 40% of annual energy from renewables, depending on operational mode. That said, carbon emissions vary significantly between operational modes and resource scenarios. In general, (1) DCM with a time-of-use energy rate increases emissions; (2) EA generally reduces emissions; and, (3) MinCO₂ and DCM with real-time energy pricing always reduce emissions. Moreover, economic dispatch of ESS attains only a portion of the maximum achievable environmental benefits, with MinCO₂ reducing system emissions by an average of 494lbs/MWh-stored more than the next-lowest operational mode. In addition, I find that greater exposure to wholesale energy prices generally reduces induced emissions, and that retail rate designs encouraging price exposure can reduce the carbon footprint of ESS without eroding the benefits offered by storage. These results indicate that the emissions induced by ESS should alleviate themselves in the coming years, as regulators encourage more efficient energy consumption and as more renewables are added to the grid.

The power grid is undergoing a transformation from a monopolized control system to a more decentralized one. Distributed renewable energy generation, responsive loads, and distribution automation are posing a new challenge to the traditional centralized control method. To address these challenges, we propose two innovative centralized and decentralized solutions for the information infrastructure of the future electricity market. For the centralized approach, we investigate the applications of an open-source control system platform VOLTTRON in the areas of building control and electric vehicle charging. For the case study, we implement the VOLTTRON platform to solve the economic dispatch (ED) problem. The VOLTTRON platform is used as a central message bus and 16 single-board computers are used to simulate distributed generators and dispatchable loads. For the decentralized approach, we propose an innovative Bitcoin-style distributed transactional model “Bit-Energy” using radically different Internet-of-Things technologies (Blockchain and Ethereum’s smart contract). “Bit-Energy” enables transparent, auditable, and peer-to-peer energy transactions between active market participants. We implement a highly efficient buyer/seller matching algorithm. Case studies demonstrate the accuracy, robustness, effectiveness, and scalability of the proposed Bit-Energy platform under various operating conditions.

Recently, world-wide power systems have been undergone a paradigm change with increasing penetration of renewable energy. The renewable energy is clean with low operation cost while subject to significant variability and uncertainty. Therefore, integration of renewables presents various challenges in power systems. Meanwhile, to offset the uncertainty from renewables, demand response (DR) has gained considerable research interests because of DR’s flexibility to mitigate the uncertainty from renewables. In this dissertation, various power system problems using bi-level optimization are investigated considering the uncertainties from wind power and demand response. In power system planning, reactive power planning (RPP) under high-penetration wind power is studied in this dissertation. To properly model wind power uncertainty, a multi-scenario framework based on alternating current optimal power flow (ACOPF) considering the voltage stability constraint under the worst wind scenario and transmission N-1 contingency is developed. The objective of RPP in this work is to minimize the VAR investment and the expected generation cost. Benders decomposition is used to solve this model with an upper level problem for VAR allocation optimization and generation cost minimization as a lower problem. Then, several problems related wind power and demand response uncertainties under power market operation are investigated. These include: an efficient and effective method to calculate the LMP intervals under wind uncertainty is proposed; the load serving entities’ strategic bidding through a coupon-based demand response (CBDR) with which a load serving entity (LSE) may participate in the electricity market as strategic bidders by offering CBDR programs to customers; the impact of financial transmission right (FTR) with CBDR programs is also studied from the perspective of LSEs; and the stragegic scheduling of energy storages owned by LSEs considering the impact of charging and discharging on the bus LMP. In these problems, a bi-level optimization framework is presented with various objective functions representing different problems as the upper level problems and the ISO’s economic dispatch (ED) as the lower level problem. The bi-level model is addressed with mathematic program with equilibrium constraints (MPEC) model and mixed-integer linear programming (MILP), which can be easily solved with the available optimization software tool.

The emergence of cloud computing has established a growing trend towards building massive, energy-hungry, and geographically distributed data centers. Due to their enormous energy consumption, data centers are expected to have major impact on the electric grid by significantly increasing the load at locations where they are built. However, data centers also provide opportunities to help the grid with respect to robustness and load balancing. For instance, as data centers are major and yet flexible electric loads, they can be proper candidates to offer ancillary services, such as voluntary load reduction, to the smart grid. Also, data centers may better stabilize the price of energy in the electricity markets, and at the same time reduce their electricity cost by exploiting the diversity in the price of electricity in the day-ahead and real-time electricity markets. In this thesis, such potentials are investigated within an analytical profit maximization framework by developing new mathematical models based on queuing theory. The proposed models capture the trade-off between quality-of-service and power consumption in data centers. They are not only accurate, but also they posses convexity characteristics that facilitate joint optimization of data centers' service rates, demand levels and demand bids to different electricity markets. The analysis is further expanded to also develop a unified comprehensive energy portfolio optimization for data centers in the future smart grid. Specifically, it is shown how utilizing one energy option may affect selecting other energy options that are available to a data center. For example, we will show that the use of on-site storage and the deploymentof geographical workload distribution can particularly help data centers in utilizing high-risk energy options such as renewable generation. The analytical approach in this thesis takes into account service-level-agreements, risk management constraints, and also the statistical characteristics of the Internet workload and the electricity prices. Using empirical data, the performance of our proposed profit maximization models for data centers are evaluated, and the capability of data centers to benefit from participation in a variety of Demand Response programs is assessed.

The lithium-ion (Li-ion) battery has become an established technology in portable electronics and electric vehicle applications. At the same time, there is rising interest in grid-based battery energy storage to improve the flexibility of the electric grid and integrate intermittent sources of renewable energy. To provide information for energy storage developers, battery system operators, state policymakers, and the general public, this research develops a framework to characterize, operate, and evaluate Li-ion battery energy storage that is connected to the electric grid and participates in a wholesale electricity market. Methods are developed to characterize and model the voltage, temperature, and capacity degradation behavior of a Li-ion battery system. Then, an optimization program is developed to schedule Li-ion storage in an electricity market while modeling and controlling its operating state and rate of capacity loss. The optimization framework is used to simulate operation of Li-ion storage in Texas’s Electric Reliability Council of Texas (ERCOT) electricity from 2002–2014, and the market revenue potential and operating lifetime of Li-ion storage are approximated. It is shown that controlling capacity degradation in operational management can extend the lifetime of Li-ion battery modules by approximately 30–60% without significantly reducing market revenue potential. To test the reliability impact of distributed Li-ion storage, residential electricity data are used to approximate how long a battery system could isolate downstream electricity customers during an outage. Thousands of outage events are simulated to show the expected islanding duration for outages occurring at different times of day. The potential reliability benefit from avoided residential electric outages is calculated and found to be much smaller than the revenue potential from the electricity market, indicating that market applications should be prioritized over residential reliability applications in siting and operating a battery system. It is found that the net-present value (NPV) of a Li-ion battery system providing wholesale energy arbitrage in the ERCOT market is negative across a range of cost and benefit parameters. However, controlling capacity degradation in operational management of the battery system is found to increase its value by approximately $100/kWh of rated energy capacity. The NPV of a battery system providing a combination of energy and Fast Responding Regulation Service (FRRS) is found to be positive across a wide range of cost and benefit parameters, indicating a Li-ion battery system could most likely provide a combination of energy and FRRS service to the ERCOT electricity market at a profit. Controlling capacity degradation in operational management of the battery system for energy and FRRS is found to have little impact on its NPV. However, controlling capacity loss makes the NPV less sensitive to variation in the lifetime of the battery modules, reducing the risks associated with premature battery cell failure.

Emerging integration of renewable energy has prompted a wide range of research on the use of energy storage to compensate for the added uncertainty that accompanies these resources. In the Electric Reliability Council of Texas (ERCOT), compressed air energy storage (CAES) has drawn particular attention because Texas has suitable geology and also lacks appropriate resources and locations for pumped hydroelectric storage (PHS). While there have been studies on incorporation of renewable energy, utilization of energy storage, and dispatch optimization, this is the first body of work to integrate all these subjects along with the proven ability to recreate historical dispatch and price conditions. To quantify the operational behavior, economic feasibility, and environmental impacts of CAES, this work utilized sophisticated unit commitment and dispatch (UC&D) models that determine the least-cost dispatch for meeting a set of grid and generator constraints. This work first addressed the ability of these models to recreate historical dispatch and price conditions through a calibration analysis that incorporated major model improvements such as capacity availability and sophisticated treatment of combined heat and power (CHP) plants. These additions appreciably improved the consistency of the model results when compared to historical ERCOT conditions. An initial UC&D model was used to investigate the impacts on the dispatch of a future high wind generation scenario with the potential to utilize numerous CAES facilities. For all future natural gas prices considered, the addition of CAES led to reduced use of high marginal cost generator types, increased use of base-load generator types, and average reductions in the total operating costs of 3.7 million dollars per week. Additional analyses demonstrated the importance of allowing CAES to participate in all available energy and ancillary services (AS) markets and that a reduction in future thermal capacity would increase the use of CAES. A second UC&D model, which incorporated advanced features like variable marginal heat rates, was used to analyze the influence of future wind generation variability on the dispatch and resulting environmental impacts. This analysis revealed that higher amounts of wind variability led to an increase in the daily net load ramping requirements which resulted in less use of coal and nuclear generators in favor of faster ramping units along with reductions in emissions and water use. The changes to the net load also resulted in increased volatility of the energy and AS prices between daily minimum and maximum levels. These impacts were also found to increase with compounding intensity as higher levels of wind variability were reached. Lastly, the advanced UC&D model was also used to evaluate the operational behavior and potential economic feasibility of a first entrant conventional or adiabatic CAES system. Both storage systems were found to operate in a single mode that enabled very high utilization of their capacity indicating both systems have highly desirable characteristics. The results suggest that there is a positive case for the investment in a first entrant CAES facility in the ERCOT market.

The recent rapid increase in the penetration of renewable sources of generation in electricity markets has introduced a new challenge for system operators due to the inherent variability of these sources. An effective solution to this challenge is to use storage capacity to offset the variabilty. An additional advantage of storage is that it can also shift load from peak to off-peak periods and lower system costs. Since electric batteries are relatively expensive, a promising form of storage is to use deferrable demand devices to decouple the purchase of electricity from the delivery of an energy service, such as thermal storage for space conditioning and hot water. The smart-charging of electric vehicles represents another type of deferrable demand. Two additional advantages of deferrable demand are that it is relatively inexpensive and the potential amount of capacity is enormous. The objective of this dissertation is to evaluate how high penetrations of wind generation affect the costs of operating an electricity grid and to determine the economic value of different types of storage from the perspective of a system operator and of individual customers. The empirical analysis is based on a stochastic form of multi-period Security-Constrained Optimal Power Flow (SCOPF) using a reduction of the network in New York State and New England. In this model, the potential wind generation and electric load are both stochastic inputs, and the optimal dispatch of conventional generating units for both energy and reserves over 24 hours is determined endogenously to meet load and maintain system reliability. The results for a hot summer day show that adding wind capacity displaces fossil fuels and increases ramping requirements, but the net effect is lower operating costs. Energy storage reduces operating costs further by 1) buying more energy when electricity prices are low by shifting demand from peak to off-peak hours, 2) providing ancillary services such as ramping services to mitigate the variability of wind generation, and 3) lowering the amount of conventional generating capacity needed to maintain system adequacy at the system peak. This is true for both utility-scale storage and deferrable demand. Although utility-scale storage reduces the operating costs more than the same capacity of deferrable demand, the capital costs of storage are higher and the total of all costs are lower for deferrable demand. Customers with thermal storage for space and/or water heating get most of their savings from lower demand payments (i.e. reducing their demand at the system peak), and for customers with electric vehicles, the main savings are from buying less gasoline. However, encouraging customers to adopt deferrable demand devices will require charging them an efficient rate structure that reflects the true cost of supplying electricity. Comparing the bills paid by customers with different types of deferrable demand shows that an efficient rate structure provides positive economic incentives for investing in deferrable demand but a flat rate structure for energy only provides perverse incentives.

The Texas ERCOT market is one of the most open, deregulated electricity markets in the world. This open market brought electricity costs down for Texas residents and businesses, creating a much more competitive economic climate. However, these low prices currently generate insufficient revenue for generators to finance construction of new or replacement generation assets. In the instance of combined cycle advanced natural gas, the Independent Market Monitor 2012 annual report estimated that a plant needed to generate 2.5 times as much as revenue it did in 2012 to incent new generation. This author argues that while the gap is still significant, the continuous changes to the ERCOT market since its inception make an historical examination like that used by the IMM less accurate. New market rules such as price caps or changes in fuel markets through new technologies like hydraulic fracturing create a very different valuation gap than a model based on historical activity alone. This analysis attempts to get a more accurate approximation of the gap through the use of publicly traded futures contracts for natural gas and electricity. Electricity futures reflect market expectations of revenue based on current and future market rules. Gas futures reflect price expectations in light of market changes like fracturing, potential LNG exports, and other changes. Financial positions can be maintained in both markets to give a fixed rate of return. Using this method, one can create a very conservative valuation model that still more accurately reflects market sentiment. This thesis starts with a brief history of ERCOT deregulation from the early 2000s to present in order to clarify for the reader the changes that have taken place in the market. It then demonstrates the futures-valuation model using an advanced combined cycle power plant as an example.

This work probes several aspects of the renewable resources and controllable loads. The investigation includes the impact of wind power in bidding process in a deregulated power market, the effect of load damping elements on power system frequency stability and security, and impact of controllable load on system operation from the viewpoint of economic volatility and physical security. In the first part, new bidding models are developed under two schemes for wind generation to analyze the competition among generation companies (GENCOs) with transmission constraints considered. The proposed method employs the supply function equilibrium (SFE) to model a GENCO’s bidding strategy. The bidding process is solved as a bi-level optimization problem. An intelligent search based on Genetic Algorithm (GA) and Monte Carlo simulation (MCS) is applied to obtain the solution. This model also considers the probabilistic variability of wind output. In the second part, the effect of frequency-sensitive load on system frequency using typical system frequency response (SFR) model is investigated. Theoretic analysis based on transfer functions shows that the frequency deviation under a variable load-damping coefficient is relatively small and bounded when the power system is essentially stable; while the frequency deviation can be accelerated when the power system is unstable after disturbance. For the stable case, the largest frequency dip under a perturbation and the corresponding critical time can be derived by inverse Laplace transformation using a full model considering effect of load-damping coefficient. Further, the error in evaluating the load-damping coefficient gives the largest impact on frequency deviation right at the time when the largest frequency dip occurs. In the last part, a new demand response model is presented. It models system economic dispatch as a feedback control process and introduces a flexible and adjustable load cost as a controlled signal to adjust load response. Compared to the conventional “one time use” static load dispatch model, this dynamic feedback demand response model can adjust load to desired level in finite discrete time steps. In addition, MCS and interval mathematics are applied to describing uncertainty of an individual end-user’s response to an ISO’s expected dispatch.

As an energy-importing developing country, Turkey depends heavily on imported petroleum and natural gas. The increase in the global petroleum price has affected the Turkish economy adversely in the last decade. Renewable energy is an important alternative in reducing Turkey’s energy dependency. Turkey’s strategies are improving domestic production and diversifying energy sources for the security of supply. New investments, especially in renewables, have been chosen to achieve these objectives. As a model for Turkey, Texas is the leader in non-hydroelectric renewable energy production in the U.S. and has one of the world’s most competitive electricity markets. However, wind generation creates unique challenges for the Electric Reliability Council of Texas (ERCOT), the transmission system operator of Texas. The market environment has forced the Public Utility Commission of Texas (PUCT) to develop unique deregulated energy markets. In 2005, the Texas Legislature passed Senate Bill 20, in part to break the deadlock between transmission and wind generation development. This legislation instructed the PUCT to establish Competitive Renewable Energy Zones (CREZs) throughout the State, and to designate new transmission projects to serve these zones. In this context, first of all, the electricity market development in Turkey is introduced in terms of renewable energy, especially wind power. Next, considering wind power, the progress in the Texas electricity market is investigated. Subsequently, we examine the development of CREZs in Texas from a regulatory perspective and discuss Texas’ policy initiatives, including the designation of CREZs. Finally, we review the impact of wind power on the primary electricity market of Texas and evaluate market conditions and barriers to renewable energy use in Turkey in order to extract suggestions. This experience may be particularly instructive to Turkey, which has a similar market structure on the supply and transmission sides. This study suggests ways that Turkey might handle renewable applications in combination with existing transmission constraints.

As an energy-importing developing country, Turkey depends heavily on imported petroleum and natural gas. The increase in the global petroleum price has affected the Turkish economy adversely in the last decade. Renewable energy is an important alternative in reducing Turkey’s energy dependency. Turkey’s strategies are improving domestic production and diversifying energy sources for the security of supply. New investments, especially in renewables, have been chosen to achieve these objectives. As a model for Turkey, Texas is the leader in non-hydroelectric renewable energy production in the U.S. and has one of the world’s most competitive electricity markets. However, wind generation creates unique challenges for the Electric Reliability Council of Texas (ERCOT), the transmission system operator of Texas. The market environment has forced the Public Utility Commission of Texas (PUCT) to develop unique deregulated energy markets. In 2005, the Texas Legislature passed Senate Bill 20, in part to break the deadlock between transmission and wind generation development. This legislation instructed the PUCT to establish Competitive Renewable Energy Zones (CREZs) throughout the State, and to designate new transmission projects to serve these zones. In this context, first of all, the electricity market development in Turkey is introduced in terms of renewable energy, especially wind power. Next, considering wind power, the progress in the Texas electricity market is investigated. Subsequently, we examine the development of CREZs in Texas from a regulatory perspective and discuss Texas’ policy initiatives, including the designation of CREZs. Finally, we review the impact of wind power on the primary electricity market of Texas and evaluate market conditions and barriers to renewable energy use in Turkey in order to extract suggestions. This experience may be particularly instructive to Turkey, which has a similar market structure on the supply and transmission sides. This study suggests ways that Turkey might handle renewable applications in combination with existing transmission constraints.

Turkey wants to encourage renewable electricity generation to reduce dependence on imported natural gas and meet its highly growing power demand. The government’s objective is to increase the share of renewable resources in electricity generation to at least 30 percent by 2023, and the specific target for the installed wind energy capacity is 20 GW by that date. Fortunately, Turkey has an enormous wind energy potential to exploit for electricity generation. When from “good” to “outstanding” wind clusters are taken into account, the overall technical wind power generation capacity in Turkey is calculated to be nearly 48 GW. In this context, this thesis investigates whether policy instruments in the Turkish regulatory frame contribute to economic viability for wind power projects or not. The financial results point out that an electricity price of 7.3 USD cent/kWh, which is the guaranteed price for wind power generation by current regulations, does not make a typical onshore wind power plant located in a “good” windy resource economically viable. However, when locally produced wing blades and turbine towers are used in the project, the purchase price increases to 8.7 USD cent/kWh, and the project becomes economically viable. As a result, the local content element introduced in recent regulations promotes wind energy investments and helps government to reach its renewable target for 2023.

This work concerns the formulation and solution of a multi-period security-constrained optimal power flow problem for real-time electricity market operations. The solution of the proposed problem is intended to be part of the core pricing procedure for electricity trading in open markets where real energy, reactive energy, voltages support, and other system resources and services can all be traded in discrete bids and offers. Traditionally, real-time dispatching operations only involve solving single-period security-constrained optimal power flow problems. This work demonstrates the need for solving multi-period security-constrained optimal power flows. The nonsmoothness of the offer/bid-driven optimal power flow problem is studied. Three techniques, namely, a trust-region based augmented Lagrangian method, a step-controlled primal-dual interior point method, and a modified constrained cost variables method, are developed for reliable and efficient computation of large-scale nonsmooth optimal power flows. Numerical studies show that these techniques are reliable and better than some existing ones. To reduce the computational complexity, two decomposition techniques are proposed and studied. In the first one, the auxiliary problem principle method is extended to handle inequality constraints created from generator ramping limits. In the second one, binding time-coupling and contingency-coupling constraints are estimated, ranked, and filtered before the computation is decomposed and parallelized using standard block matrix computation techniques. According to experimental results, the most promising way of solving large-scale multi-period security-constrained optimal power flow problems in real time is to combine the second decomposition method with the modified constrained cost variables method. The optimal power flow formulation and relevant computation techniques proposed in this work balance the needs for: (1) deterministic convergence, (2) accurate computation of nodal prices, (3) support of both smooth and nonsmooth costings of a variety of resources and services, such as real energy, reactive energy, voltage support, etc., (4) full active and reactive power flow modeling of large-scale systems, and (5) satisfactory worst-case performance that meets the real-time dispatching requirement.

The pricing of options is generally dominated by the Black-Scholes approach. The characteristics of electricity prices, however, appear incompatible with the assumed process underlying the Black-Scholes model (ie. Geometric Brownian Motion). This thesis examines the potential for option pricing biases when applying the Geometric Brownian Motion (GBM) model in the context of the New Zealand electricity market. Two alternative models, Geometric Mean Reversion (GMR) and Geometric Mean Reversion with Jumps (GMRJ), are tested against the GBM. Beginning with an examination of the NZ electricity market and price process, this study identities and explains some of the characteristics of electricity prices. Then it identities and fits the appropriate price process models to the spot price data. The results indicate that the GMR model is the best representation of the electricity price process in the New Zealand context, while the GBM model is the least accurate. Moreover, simulating option prices for these models under different scenarios reveals that the GBM model significantly overprices electricity options in most cases. Option prices generated by the GMRJ model are generally not largely different from those of the GMR model.