日负荷曲线预测是制定日调度计划等工作的基础, 典型日负荷曲线的预测对电网规划设计具有重要的参考价值, 其预测的准确性直接影响到电网建设和运行的经济效益。通过对历史数据的预处理、初始聚类中心的设置以及最优聚类数目的确定, 建立典型日负荷曲线的聚类预测模型。基于历史数据的聚类结果以及待预测日的温度、降水量、湿度等相关参数, 得出典型日负荷曲线预测结果。将该模型应用于上海电网典型日负荷曲线预测, 显示聚类算法能够全面地考虑相关的因素, 得到较为准确的预测结果。

The forecasting of daily load curve is the fundament for daily scheduling of power system, forecasting of typical daily load curve could also bring important reference value for distribution network planning and the prediction accuracy could exert a direct impact on the economic efficiency of the power system construction and operation. A forecasting model for typical daily load curve based on clustering theory is proposed in this paper, which would combine with the pretreatment of historical data, the set of initial cluster centers, and the determination of the optimal number of clusters. The daily load curve could be provided based on the clustering result of historical data, temperature, precipitation, humidity, and other relevant parameters of the predicted day. This model presented here is applied to a regional grid of a south city of china, it is demonstrated that the clustering algorithm could fully consider the relevant factors and then get the more accurate forecasting results.