Analysis of French electric market using heteroscedastic models of time series

A dynamic long memory regression model with autoregressive errors is considered for the analysis of hourly electricity spot prices. The method provides reliable and accurate forecasts of hourly prices in the electricity market of France, Powernext day-ahead. The presence of significant autocorrelation in squared residual recommends to fit a conditional heteroscedastic time series model. These models together have appealing economic and statistical implications.