Management of Demand Response Programs in the Electricity Industr

Daily electricity load profile is characterized by peak hours which are periods in which electrical power is expected to be provided for a sustained period at a significantly higher than average supply level. As a result, satisfying the electricity demand throughout the day will entail utility companies to build additional plants that are only used during the highest peak hours of the year or to buy high-priced wholesale energy. Further, such costs will increase given the expected growth of electricity demand in the next decades. To avoid these additional costs and address the resulting supply-demand mismatch, utility companies have designed Demand Response Programs (DRP) which are programs that incentivize customers to shift their electricity demand from peak hours to off-peak hours. In this work, I study the problem of an electricity utility company that offers DRP to its commercial and industrial customers with the objective of reducing its electricity costs.;In Chapter 1, I give an overview of the electricity industry in the United States and describe the important role that DRP play in improving the electric grid reliability and reducing the costs of electricity generation for the utility companies. I then describe and formulate the problem of an electricity retailer that offers interruptible demand response programs, which are a type of DRP, to their commercial and industrial customers. These programs consist of the Base Interruptible Program (BIP) and the Agricultural and Pumping Interruptible Program (API). Using these contracts, enrolled customers agree to curtail their consumption by a pre-specified load when instructed and obtain in return financial payments from the utility company. The operational challenges of these programs are in their implementation and management due to the large number of interruption possibilities, the uncertainty in electricity demand and the limited number of interruptions the electricity retailer. To address these challenges, I propose and describe the solution adopted to solve the dynamic program. The approach I use consists of a certainty equivalence algorithm that had two components: an electric load forecasting model and the deterministic model of the dynamic program which I discuss in chapters 2 and 3 respectively.;In Chapter 2, I present an electric load forecasting model in the context of demand response for both the short and long term horizons. The short term model consists of predicting by nonparametric regression the hourly electricity demand at the start of a given day using the previous day load and same day temperature as the driving variables. The long term forecasting model consists of first predicting the peak load through multivariate and semiparametric regression taking into account the temperature variable and calendar effects. Then, I approximate the hourly load profile by nonparametric regression using the predicted peak load. Further, I construct the peak load distribution by temperature simulation and kernel density approximation. The proposed methodolgy had been used to forecast the short and lonh term electricty demand as well as the probability distribution of the peak load for the area seved by the Southern California Edison (SCE) electric utility company. The performance of the methodology is evaluated by comparing the forecasts resuls to the ones of the California Independent System Operator (CAISO) for the area served by SCE.;In Chapter 3, I study the problem of implementing these contracts by determining their execution policy using a certainty equivalence approach. A central component of the certainty equivalence algorithm is the deterministic problem in which the electricity demand in known. Given that this problem is NP-hard, we propose a heuristic that efficiently solves the deterministic problem and test its efficiency by determining the optimality gap with a lower bound. Using the developed electricity forecasting and demand simulation models we then solve the certainty equivalence model in order to devise near optimal strategies for executing such contracts and verify its effectiveness.