基于非高斯分布GARCH模型的负荷预测

提出一种基于非高斯分布的广义自回归条件异方差(GARCH)模型的短期负荷预测方法.在论证自回归条件异方差(ARCH)效应存在性的基础上,将标准GARCH模型的正态条件分布假设推广为非高斯条件分布的形式(t分布、广义误差分布).用极大似然估计获得ARCH族各模型的参数估计,建立了非高斯分布假设GARCH模型(GARCH-t,GARCH-GED).比较了ARMA、标准GARCH、非高斯分布GARCH模型的预测能力,分析平均预测误差、最大预测误差能力等指标显示GARCH-GED模型表现最出色.算例表明,基于非高斯分布GARCH负荷预测模型是有效而可行的.

Load forecast based on GARCH model with non-Gaussian distribution

A short-term load forecast method based on GARCH(Generalized AutoRegressive Conditional Heteroscedasticity) model with non -Gaussian distributions is proposed.Based on the discussion of ARCH effect existence,the Gaussian conditional distribution assumption of standard GARCH model is generalized to non -Gaussian alternative distribution(t distribution,generalized error distribution).The model parameter estimations are obtained by Maximum Likelihood Estimation and the improved GARCH models with non -Gaussian distribution,such as GARCH -t and GARCH -GED,are established.The forecast performances are compared among ARMA,standard GARCH,GARCH-t and GARCH-GED,and their mean forecast errors and maximum forecast errors are analyzed,which indicate GARCH-GED is superior to other models.Calculation example shows that the load forecast method based on GARCH model with non -Gaussian distribution is efficient and feasible.