基于灰色关联指标筛选的BP神经网络中长期电力负荷滚动预测马尔可夫残差修正模型研究  

Error correcting Markov chains for medium/long term load rolling forecasting of BP models based on grey relational grade

作　　者：吕佳良[1] 张振刚[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：《华东电力》2008年第9期10-13,共4页East China Electric Power

基　　金：国家自然科学基金资助项目(70671039);高等学校博士点专项基金项目(20040079008)

摘　　要：针对中长期电力负荷预测样本量小、多因素影响的特点,利用灰色关联度筛选影响因素,建立基于BP神经网络算法的负荷预测模型,通过多因素变量及历史负荷变量序列进行滚动预测,得到的预测值明显优于单一预测方法,并通过马尔可夫过程对预测残差进行修正,使预测精度得到较大提高,研究实证表明,这种预测方法具有进行推广应用的价值。To deal with the features of med-long term power load forecasting such as sample shortage and being subject to multiple factors,the grey relational grade was used to screen influencing factors and construct the load forecasting model based on BP neural network algorithms.The predicted value obtained by multifactor variables and historical load variables is more accurate than that obtained by single forecasting method.Then the Markov process was used to modify the forecasting error,which makes the prediction accuracy increase greatly.

关 键 词：灰色关联度 BP神经网络 马尔可夫链 电力负荷 

分 类 号：TM715[电气工程—电力系统及自动化]