Chaotic characteristics of electricity price and its forecasting model

Abstract

In electricity market the price of electricity depends on the joint effect of many factors with its evolving process characterized by very complicated random motion. To reveal the inherent law behind the seemingly random evolving process, first the chaotic feature of electricity price is verified with the chaos theory. The Lyapunov exponents and the fractal dimensions of the attractors are extracted from the phase space reconstructed by the single variable time series of electricity price. Here it can be seen that the electricity price possesses chaotic characteristics, and such a chaotic behavior is further verified by the surrogate data method, providing the basis for performing the short-term forecast of electricity price with the help of the chaos theory. Then a more accurate phase space is reconstructed by the multi-variable time series constituted by the electricity price and its correlated factors, i.e., the system load and the available generating capacity time series. By tracing the evolving trend of the adjacent phase points in the phase space, the global and local electricity price forecasting models based on the recurrent neural network are established, with which the electricity prices in the New England electricity market are successfully predicted.