Modeling Hourly Day-Ahead Electricity Prices: A Functional Data Analysis Approach

• Activity Number: 475
• Type: Contributed
• Date/Time: Wednesday, August 3, 2011 : 8:30 AM to 10:20 AM
• Sponsor: Business and Economic Statistics Section
• Author(s): Asitha Edirisinghe*+ and V. A. Samaranayake
• Companies: Missouri University of Science and Technology and Missouri university of Science and Technology
• Keywords: High Frequency Time Series ; Price Forecasting ; Electricity Market ; Spline Functions ; Irregular Seasonality ; Volatility

Abstract

Four years of hourly day-ahead electricity price data from the region supplied by the American Electric Power (AEP) company through the PJM Regional Transmission Organization (RTO) are modeled. The price data show several features, such as irregular seasonal behavior, weekly and daily cycles, as well as sensitivity to oil price fluctuations, making it unsuitable for modeling using techniques that assume stationarity. We adopt a structural time series approach to remove some of the non-stationary behavior in the daily aggregate series and model the residual hourly prices obtained by subtracting the predicted daily average from the hourly series. The series is then scaled to remove long-term fluctuations in the variance due to economic conditions and demand. The resulting series is then modeled using spline functions to estimate the daily cycles that change across summer, winter and shoulder months. A GARCH type model is also used to investigate the within-day conditional volatility that is observed in the price data. This case-study can be looked upon as providing a new paradigm for modeling high-frequency data with non-stationary and cyclical features that change over time.