A Discrete Increment Model for Electricity Price Forecasting Based on Fractional Brownian Motion

Abstract

Electricity price forecasting is essential for all participants of the power consumption market. However, electricity price series has complex properties such as high volatility and non-stationarity that make forecasting turn out to be very difficult. In this study, we aim to forecast electricity price series by the discrete increment model of fractional Brownian motion (fBm). A specific feature of the fBm is that it represents a typical non-stationary stochastic process with long-range dependent (LRD) characteristics. Analysis of electricity price series has LRD characteristics, and the Hurst exponent can be calculated. The Hurst exponent is the key parameter of fBm, and it is a measure of self-similarity. The stochastic differential equation driven by fBm is discretized into the discrete increment model for electricity price forecasting. Other parameters of the discrete increment model can be evaluated by the maximum likelihood estimation (MLE). The performance of the proposed method is demonstrated by using the data from the U.S. Energy Information Administration. The validity of the proposed model was compared with other methods.