电力短期负荷的多变量时间序列线性回归预测方法研究

根据单变量时间序列的相空间重构思想，提出了一种电力短期负荷的多变量时间序列相空间重构方案，同时针对每一分量时间序列采用互信息法进行最佳延迟时间的选择，最优嵌入维数则采用最小预测误差法进行确定。根据相点间的欧氏距离和关联度，提出了最近邻域点的优化选择方法，建立了多变量时间序列的一阶局域线性预测模型。通过重庆某地区电力系统短期负荷预测的计算实例表明，基于多变量时间序列的负荷预测方法与单变量负荷预测方法相比，具有较强的自适应能力和较好的预测效果。

The Research of Local Linear Model of Short-Term Electrical Load on Multivariate Time Series

According to the idea of phase space reconstruction of scalar time series, phase space reconstruction of multivariate time series based on short-term electric load is presented. The good time delay is chosen for each scalar time series by mutual information. The method to get the minimum embedding dimension is based on minimum forecasting errors. The choice of optimal neighbor points is presented in this paper according to Euclidean distance and correlation degree between neighbor points and forecasted points. Then one-rank local linear model based on multivariate time series is proposed. The daily load forecasting for a certain district in Chongqing is respectively carried out by the method presented in this paper. Through the analysis of the obtained forecasting results it is shown that the adaptive ability and the forecasting effect of the method presented in this paper is better than that of scalar time series.