基于灰色神经网络与灰色关联度的中长期日负荷曲线预测  

Forecasting of mid-long-term daily load curve based on grey neural network and grey relational degree

作　　者：王丽 朱文广 杨为群 程虹 肖园 彭怀德 柯学 胡钋[2] WANG Li;ZHU Wenguang;YANG Weiqun;CHENG Hong;XIAO Yuan;PENG Huaide;KE Xue;HU Po(Economic and Technical Research Institute,State Grid Jiangxi Electric Power Company,Nanchang 330032,China;School of Electrical Engineering,Wuhan University,Wuhan 430072,China)

机构地区：[1]国网江西省电力公司经济技术研究院,江西南昌330032 [2]武汉大学电气工程学院,湖北武汉430072

出　　处：《武汉大学学报（工学版）》2019年第1期58-64,70,共8页Engineering Journal of Wuhan University

基　　金：国家高技术研究发展计划(863计划)资助项目(编号:2015AA050101);国家自然科学基金青年项目(编号:51207115)

摘　　要：用电结构变化和经济发展会深刻影响中长期的日负荷特性.采用加权平均法确定归一化之后的预测年基准曲线,利用非常适于少数据、多因素预测问题并具有高度非线性拟合特性的灰色神经网络,对中长期日负荷曲线的日特征参数进行预测,其中考虑了经济发展、用电结构的影响;并利用粒子群算法对灰色神经网络的参数进行初始化,以提高网络的全局搜索性能.引入灰色绝对关联度描述曲线的相似特性,基于日负荷特征参数约束,通过所构建的非线性规划模型进行中长期日负荷曲线预测.选用江西电网2006-2015年各季度日负荷数据进行测试,结果表明本方法具有较高的预测精度.Changes in power structure and economic development will profoundly affect the mid-long-term daily load characteristics.In this paper,the weighted average method is used to determine the forecasting annual reference curve after normalization.And we use the grey neural network which is very suitable for small data and multifactor prediction problem with high nonlinear fitting characteristics,and forecasting the daily characteristic parameters of mid-long-term daily load curves,taking into account the economic development,the impact of power structure.And the parameters of grey neural network are initialized by particle swarm optimization to improve the global search performance of the network.Based on the daily load characteristic parameter constraints,the mid-long-term daily load curve forecasting is carried out by constructing the nonlinear programming model by introducing the similarity characteristics of the grey absolute relational degree description curve.The results show that this method has high forecasting accuracy.The results show that this method has high forecasting accuracy.

关 键 词：用电结构 灰色神经网络 粒子群优化算法 灰色关联度 中长期日负荷曲线预测 

分 类 号：TM715[电气工程—电力系统及自动化]