灰色BP神经网络模型的优化及负荷预测  

Optimization and power load forecasting of gray BP neural network model

作　　者：周德强[1] 武本令 

机构地区：[1]长江大学信息与数学学院,湖北荆州434023 [2]许昌市高级技工学校,河南许昌461000

出　　处：《电力系统保护与控制》2011年第21期65-69,共5页Power System Protection and Control

基　　金：湖北省教育厅科学技术研究项目(D20111305;Q20101309)

摘　　要：为了提高样本数据较少情况下中长期负荷预测的预测精度,分析了传统GM(1,1)预测模型的缺点,提出了一种适用于中长期负荷预测的GM(1,1)优化建模方法。用一个与GM(1,1)模型的时间响应式具有相同形式的连续函数,拟合灰色系统的原始离散数据,将连续函数映射到神经网络,构建了GM(1,1)模型的灰参数与BP网络权值的对应关系。用已知负荷作为训练样本,利用BP算法对网络进行优化,当网络收敛时,提取优化的灰参数,实现了应用GM(1,1)模型对中长期负荷预测的优化建模。算例分析结果表明该方法是可行且有效的。In order to enhance the forecasting accuracy of medium and long-term power load forecasting under the conditions of small sample data, the shortcomings of traditional GM （ 1, 1 ） forecasting model are analyzed. A GM （ 1, 1 ） optimization modeling method for medium and long-term load forecasting is proposed. By use of a continuous function whose form is the same as the GM （ 1, 1 ） time response formula, the raw discrete data is fitted. The continuous function is mapped to a BP neural network, and the corresponding relation between GM （ 1, 1 ） model gray parameters and BP network weights is established. With known load data as training sample, the network is optimized by means of BP algorithm, when the BP network convergence, optimized gray parameters can be extracted, therefore, the optimization modeling of GM （ 1, 1 ） for medium and long-term load forecasting is realized. The example results show that the method is feasible and effective.

关 键 词：GM(1  1)模型 灰色BP神经网络模型 BP算法 优化建模 中长期负荷预测 

分 类 号：TM715[电气工程—电力系统及自动化]