Price Process of Electricity

Abstract

In general, representing a price over several time periods as a series of in-dependent random variables is a poor model, as it ignores correlation in the uncertainties between time periods. A more conventional model is to use a price process [6]. Developing a price process for electricity, however, is quite difficult as there is very little empirical data available. Furthermore, the de-mand for electricity is not the same over all hours of the day. As shown in Figure 6.1, the price is generally lowest in the early morning hours and high-est in the early evening. This relation between hour of day and price closely follows the load profile (shown in Figure 6.2). In this book, price models are developed to account for the change in average price as a function of hour of day in addition to standard price models which treat all time steps equally. Additionally, the variance of the price estimate for the next time step may be significantly reduced by incorporating additional variables which are cor-related with price, such as day of week or temperature. To create a price model for electricity, hourly data from the Pennsylvania-Jersey-Maryland (PJM) market [27], which opened on April 1, 1997, will be used.