基于相似日搜索的改进LMD与ESN相结合的短期电力负荷预测模型

The Short-term Power Load Forecasting Model of Combing ILMD and ESN Based on Similar Days Searching

短期电力负荷容易受到自然因素及社会因素的影响,这使得负荷预测比较困难.为了提高短期负荷的预测精度,提出了基于相似日搜索的改进局部均值分解(ILMD)和回声状态网络(ESN)相结合的短期电力负荷预测模型.首先用模糊聚类分析将与预测日最相似的多个日期筛选出来.然后把这些相似日的整点负荷数据按照时间先后排成一组数据序列,用改进的LMD进行分解,对分解出的各个分量分别建立一个ESN网络,对每一个网络分别训练并进行预测.最后把每个网络的预测结果累加起来就是最终的预测值.实验证明此方法能有效提高预测精度.

Short-term power load was easily influenced by natural factors and social factors, which made load forecast more difficult.In order to improve the accuracy of short-term power load prediction, the forecasting mode of combing improved local mean decomposition （ILMD） and echo state network （ESN） based on similar days searching was proposed.Firstly, the days most similar to the forecasted date were selected by fuzzy cluster analysis.A data sequence was formed by uniting the similar days＇ hourly loads together according to their time orders.Then, the ILMD was used to decompose the data sequence into several independent components, and an ESN was established for each component, separately.Each network was trained with similar daily load data.Using each trained network to predict the value of the corresponding component, the final result of prediction was the accumulation of all components predict values.Experiments showed that this method could effectively improve the prediction accuracy.