利用粗糙集约简算法改进BP神经网络负荷预测模型  

Use rough sets reduction algorithm to improve BP neural network load forecasting model

作　　者：李艳梅[1] 孙薇[1] 

机构地区：[1]华北电力大学工商管理学院,河北保定071003

出　　处：《电力需求侧管理》2008年第5期21-23,共3页Power Demand Side Management

摘　　要：传统神经网络预测模型受网络结构复杂性和样本复杂性的影响,容易出现"过学习"或低泛化能力。利用粗糙集理论中的几种属性约简算法对与负荷相关的各种历史数据进行约简,剔除与决策信息不相关的属性。实例证明该方法简化了BP神经网络的输入变量,从而缩短了神经网络模型的训练时间,提高了预测性能。Traditional neural network load forecasting model is affected by the complexity of the network structure and complexity of the samples, easily leads to a ＂over-study＂ or low-generalization; The method uses several attribute reduction algorithms in rough sets theory to reduce the various historical data associated with load, eliminates the attributes that are not relevant to decision-making information. Examples prove that this method simplifies the BP neural network input varifies, so as to shorten the neural network model of training time and improve the forecast performance.

关 键 词：负荷预测 BP神经网络 粗糙集约简算法 

分 类 号：F407.61[经济管理—产业经济]