基于EUNITE竞赛数据的中期电力负荷预测  

Mid-term load forecasting based on EUNITE competition data

作　　者：李炎[1] 翟永杰[2] 周倩[2] 韩璞[2] 

机构地区：[1]华北电力大学后勤与资产管理处,河北保定071003 [2]华北电力大学控制科学与工程学院,河北保定071003

出　　处：《华北电力大学学报（自然科学版）》2007年第4期22-26,共5页Journal of North China Electric Power University：Natural Science Edition

基　　金：华北电力大学博士学位教师科研基金资助项目(200512014)

摘　　要：应用自组织模糊神经网络(SOFNN)算法,基于欧洲智能技术网络(EUNITE)竞赛数据进行了中期电力负荷预测的应用研究。算法能够自动决定神经模型的结构并得出模型的参数,具有很好的实用价值。研究了训练数据选取和输入特征向量编码等实际应用问题,结果表明负荷预测精度高,优于竞赛的优胜者,之后提出了结合周平均负荷预测修正日负荷预测的方法,精度得到进一步地提高。The self-organizing fuzzy neural network （SOFNN） algorithm was applied for load forecasting with the data of EUNITE competition. The algorithm can automatically determine the model structure and identify the model parameters. The training data selection and input variables encoding problems were also discussed in details. The results show that SOFNN can outperform winners＇ models by providing very promising prediction accuracy. Further more, the average load value in each week was forecasted to revise the daily load. Through this procedure, better forecasting accuracy can be achieved.

关 键 词：模糊神经网络 自组织 负荷预测 

分 类 号：TM715[电气工程—电力系统及自动化]