电力短期负荷的多变量时间序列预测方法研究

在混沌理论中,负荷预测模型的建立通常由一单变量时间序列的相空间重构来实现,然而在实际过程中往往难以确定单变量时间序列是否包含了重构动力系统的全部信息,特别是在有限时间序列存在噪声时。因此,本文将单变量时间序列方法拓展到多变量时间序列中,进行多变量时间序列的相空间重构,计算了各时间序列的延迟时间和嵌入维数,建立了局域线性预测模型。预测结果检验显示,该方法具有较好的预测效果。

Nonlinear prediction of short-term electrical load on multivariate time series

Load forecasting model may be founded by phase space reconstruction to univariate time series in chaotic theory.But in practical process,one may not absolutely make sure the scale time series is enough to reconstruct phase space,specially noises existing.So multivariate time series is extended from univariate time series in order to reconstruct multivariate time series.Simultaneously delay time and embedding dimension are counted.Then local linear forecasting model is established.The result shows the method is given better effect.