电力系统短期负荷预测的混合模型神经元网络方法  

A HYBRID MODEL NEURAL NETWORK BASED APPROACH TO SHORT\|TERM LOAD FORECASTING

作　　者：赖晓平[1] 周鸿兴[2] 田发中 

机构地区：[1]山东大学威海分校控制工程系,威海264209 [2]山东大学数学与系统科学院,济南250100 [3]积成电子实验所,济南250100

出　　处：《电网技术》2000年第1期47-51,共5页Power System Technology

基　　金：国家自然科学基金!(69774002)

摘　　要：提出一种将线性模型方法和神经元网络方法相结合的负荷预测方法——混合模型神经元网络方法。该方法将一部分线性变化的负荷分量用线性模型描述，其它负荷分量用神经元网络建立，因而同时具有线性模型的优点和神经元网络的优点。将这一方法用于江苏省连云港市超前２４ｈ 负荷预测，取得了比单纯的神经元网络模型高的预测精度。A hybrid model neural network based approach to short\|term load forecasting is presented in this paper.This approach is a combination of linear model based method and neural network based method, thus it possesses the advantages from both methods. In this approach, some load components are described by linear models and the others are modeled by neural networks. Simulations show that with this approach a more accurate forecasting can be made than with simple neural network based method.

关 键 词：短期负荷预测 线性模型 电力系统 神经元网络 

分 类 号：TM715[电气工程—电力系统及自动化]