基于密度峰值聚类的超短期工业负荷预测

Ultra Short-Term Industrial Power Prediction Based on Density Peak Clustering

针对水泥工业提出一种使用密度峰值聚类和广义回归神经网络进行超短期负荷预测的方法,可以为大用户购电提供重要依据和参考。鉴于传统聚类很容易进入局部鞍点并且非常依赖于初始化数据,而密度峰值聚类是一种具有快速收敛,高鲁棒性,无需人为设置最佳聚类数等优势。所以采用密度峰值聚类算法分析负荷数据,然后对每一类簇构建广义回归神经网络预测模型得到预测结果。由仿真软件得出仿真结果表明,所提方法具有较高的预测精度,可以用于指导用户合理购电。

In this paper,a method of super short term load forecasting using density peak clustering and generalized regression neural network is proposed for cement industry,which can provide important basis and reference for large users to purchase electricity.In view of theSince traditional clustering is easy to enter the local saddle point and is very dependent on the initialization data,while density peak clustering has the advantages of fast convergence,high robustness,and no need to set the optimal clustering number artificially.Therefore,this paper adoptsed density peak clustering algorithm to analyze load data,and then builds built a generalized regression neural network prediction model for each type of cluster to get the prediction results.The simulation results of the simulation software show that the proposed method has high prediction accuracy and can be used to guide users to purchase electricity reasonably.