加权马尔可夫链在负荷预测中的应用

利用小波将月售电量序列进行分解,对不受气象因素影响的趋势项和周期项,分别采用趋势外推和周期图法进行预测,对受气象因素影响的随机项,应用模糊聚类的方法确定分级标准,将观测值之间的相关系数作为权值,采用加权马尔可夫模型进行预测,然后进行综合。该法避免了精度较低的月气象资料对负荷预测精度的影响,不仅得到了未来月售电量的具体值,而且得到了其所属的区间,因此更加符合实际。最后给出了预测实例,验证了所提方法的有效性。

Application of Markov chain with weights to load forecasting

The series of the monthly sales electric energy is resolved by utilizing the wavelet.To the trend term and cycle term free of meteorological factor,this paper applies trend extrapolate and cyclic graph methods to predict them separately.To the last term at random that is influenced by meteorological factor,this paper employs fuzzy sequential cluster method to set up the classification standard,then it regards the correlation coefficients of record values as weights and predicts the future loads by using Markov chain model with weights.And all results are synthesized at last.This method avoids the influence to the load forecasting prediction of the moon meteorological data,which has a bad prediction.Not only the concrete value of the monthly sales electric energy but its range in the future is gained.So the prediction is more practical.A forecasting instance is provided and the validity of the method proposed is tested finally. This project is supported by Electrical Science Foundation of Hunan Province(No.20043005).