基于K-means聚类的小波支持向量机配电网短期负荷预测及应用

Short-term Load Forecasting and Application for Distribution Network of Wavelet Support Vector Machine Based on K-means Clustering

配电主站因量测能力不足制约着潮流计算、负荷转供等应用功能的工程实用化水平。在智能配用电大数据环境下,针对配电网负荷波动性和随机性大的特点,提出K-means聚类和小波—支持向量机相结合的配电网短期负荷预测方法。通过横向聚类分析提取日负荷典型特征曲线,补全历史缺失数据;通过纵向聚类分析对历史相似日归类,挖掘外部环境因素对负荷的影响。采用小波变换将历史数据分解到不同的尺度上,结合聚类结果形成各分支训练样本进行支持向量机预测,各分支预测结果叠加生成最终预测结果。基于配电变压器负荷预测的结果对负荷转供方案进行安全校验,为线路检修计划提供指导。

Due to the insufficient measurement of power distribution main station,the practical level of power flow calculation,load transfer and other applications is restricted. In this paper,the short-term load forecasting method of distribution network under intelligent big data background is studied. In view of the characteristics of great volatility and randomness of distribution network load,the short-term load forecasting method combining K-means clustering and wavelet support vector machine is proposed. The typical characteristic curve of daily load is extracted by horizontal clustering analysis to complete the missing data. By vertical clustering analysis,the effects of external environmental factors on load were explored. The wavelet transform is used to decompose the historical data into different scales. Combining the clustering results,the branch training samples are formed to support the prediction of the support vector machine. The prediction result of each branch is superimposed to generate the final prediction result. Based on the forecast results of each load,the safety check of load transfer scheme is carried out to provide guidance and suggestions for line maintenance plan.