计及相似日与气象因素的电动汽车充电负荷聚类预测

电动汽车充电负荷精准预测对于电网调度、电力市场交易、充电站规划建设等具有实际意义。由于电动汽车充电负荷特性异于传统的电力负荷,两者负荷的规律性及影响因素的敏感性各有不同,有必要针对电动汽车充电负荷影响因素及预测模型开展针对性研究。考虑到不同类型电动汽车充电负荷时间序列特性及影响因素存在差异,构建考虑日类型、最高与最低温度的电动汽车充电负荷预测模型;采用模糊C均值(fuzzy C-means, FCM)聚类算法对充电负荷进行聚类分析,挖掘数据特征属性,提取相似日负荷;针对聚类后的相似日负荷采用最小二乘支持向量机(least square support vector machine, LS-SVM)进行预测。将所得的预测结果和测试集进行对比,结果显示,基于该模型的预测精度高于使用非聚类的LS-SVM方法,验证了预测模型的有效性。

Clustering Prediction of Electric Vehicle Charging Load Considering Similar Days and Meteorological Factors

Accurate prediction of electric vehicle charging load has practical significance for power grid dispatching, power market trading, charging station planning and construction. Because the characteristics of electric vehicle charging load are different from the traditional electric load, it is necessary to carry out targeted research on the influencing factors and prediction model of electric vehicle charging load. Considering the differences of time series characteristics and influencing factors of different types of electric vehicle charging load, a prediction model of electric vehicle charging load considering daily type, maximum and minimum temperature is established. Fuzzy C-means(FCM) is used to cluster the charging load, data feature attributes are mined, and similar daily load is extracted. The least square support vector machine(LS-SVM) is used to predict the similar daily load after clustering. The prediction results are compared with the test set, and the results show that the prediction accuracy of the proposed model is higher than that of the LS-SVM method, which verifies the effectiveness of the prediction model.