Short term forecast of electricity price based on mixed autoregressive moving average and hidden periodic model

The mixed autoregressive moving average (ARMA) and hidden periodicity model is chosen to predict a short term series of electricity price. Eliminating the trend influence, a complex hidden periodicity model of price sequences can be obtained with discrete Fourier transformation. The periodicity parameters of price sequences are calculated by a simple detection for periodogram. To take consider the impact of historical information on the present states into account, the ARMA model is used to fit the residual stochastic component. The Akaike's information criterion is employed to determine the number of order in autoregressive moving average model, whose parameters are estimated by moment approach, where the pre-assumption on periodicity scale gets unnecessary. The number and size of periodicities is only derived from a simple computation model. The model proposed is verified by actual data of electricity price in PJM power market in USA.