Short term load forecasting based on SV model

The volatility of load time series is noteworthy in load forecating analysis. Considering the characteristic of time-varying variance, a feasible method of short term load forecasting based on Stochastic Volatility (SV) models is presented. The Quasi Maximum Likelihood Estimate (QMLE) is brought in to specify the standard SV model. The model is transformed into state space form, and the Kalman filter is employed to estimate the parameter. Following different conditional distribution, the extended non-Gaussian SV model is proposed. Furthermore, the curve of dynamic volatility is illustrated and the time-varying characteristics in the volatility of load time series is analyzed. Based on the actual daily load data set of Nanjing, the SV type models are specified and daily forecasting is demonstrated, the forecast performance of SV model is compared with GARCH model by three summary index. The empirical results verifies the validity and feasibility of the proposed method.