基于K-Adaboost数据挖掘的配电网负荷预测

Load Forecasting of Distribution Network Based on K-Adaboost Data Mining

气象因素是造成配电网负荷波动的主要原因,在利用斯皮尔曼相关系数分析气象因素与配电网负荷相关性的基础上,着重把握部分气象因素与配电网负荷的联系,针对这类气象因素对配电网负荷的影响,提出一种基于数据挖掘聚类分析和Adaboost的负荷预测方法。首先对历史负荷数据进行预处理,然后应用K均值聚类算法对待测点气象因素进行分析,选择与待测点同类气象因素的历史负荷作为训练样本,最后采用Adaboost算法建立配电网负荷预测模型。通过实例证明K-Adaboost预测模型比BP神经网络预测模型更加稳定并且更贴近实际负荷。

Meteorological factors are the main cause of load fluctuation in distribution network.On the basis of analyzing the correlation between meteorological factors and distribution network load with Spearman correlation coefficient,the relationship between some meteorological factors and distribution network load is emphatically grasped,in view of the influence of such weather factors on meteorological sensitive load,and a load forecasting method based on data mining clustering analysis and Adaboost is proposed.First,the historical load data is preprocessed,then the K means clustering algorithm is used to analyze the meteorological factors of the test points,and the historical load of the similar meteorological factors is selected as the training sample.Finally,the Adaboost algorithm is used to establish the distribution network load forecasting model.The example shows that the K-Adaboost prediction model is more stable and closer to the actual load than the BP neural network prediction model.