基于LSSVM和马尔可夫链的母线负荷短期预测

Bus load short-term forecast based on LSSVM and Markov chain

提出了一种母线负荷短期预测混合算法。采用最小二乘支持向量回归方法(LSSVM)进行短期母线负荷预测,并提出一种广义网格搜索算法对模型参数进行优化;由历史预测误差组成误差序列,将历史预测误差序列看作是一个符合马尔可夫过程的时间序列,采用马尔可夫链方法对未来的预测误差进行估计,采用误差估计结果对上一步LSSVM的预测结果进行修正,得出最终预测结果。经算例分析证明,所提方法能显著提高预测精度。

A hybrid method is proposed for bus load short-term forecast.It firstly adopts the least squares support vector machines（LSSVM）method to forecast bus load and puts forward a generalized grid-search algorithm to optimize the selection of model parameter.Then regarding the series of history forecast error as a process of Markov chain,it adopts the Markov chain method to forecast the potential forecast error produced by LSSVM model.At last,it adopts the results from Markov Chain Method to modify the results from LSSVM and get the final forecast load.The case study in the end of this paper proves that the proposed hybrid method is available to satisfactory forecast accuracy.