基于双层贝叶斯分类的空间负荷预测  

Spatial Electric Load Forecasting Based on Double-level Bayesian Classification

作　　者：陶文斌[1] 张粒子[1] 潘弘[2] 李振元 郑华[1] 

机构地区：[1]华北电力大学电力工程系,北京市昌平区102206 [2]浙江省电力试验研究所,浙江省杭州市310014 [3]吉林电力调度通信中心,吉林省长春市130021

出　　处：《中国电机工程学报》2007年第7期13-17,共5页Proceedings of the CSEE

摘　　要：采用分类模型进行空间负荷预测。由于决定负荷密度的各属性之间并非完全独立,传统分类器分类效果并不理想。为此,该文采用双层贝叶斯分类模型,该模型是朴素贝叶斯分类模型的扩展,通过选取关键属性将单一属性集划分为强属性集和弱属性集,弱化了各属性间相互独立的假设,提高了对样本分类的正确性。给出了构建和训练该模型的方法,在正确分类的基础上,文中采用类内相似度方法进行负荷密度的修正。实践和算例证明,该文模型和算法适合进行空间负荷预测。Double-level Bayesian classification model was built to forecast the spatial electric load in this paper. This model is an improvement of naive Bayes classifier. It divided the attributes set into strong attributes set and weak attributes set by selecting the key attributes. It weakened the attribute independence assumption. This model highly improved the accuracy of the classification. On the base of classifying accurately, the method of similar degree in the class was used to correct the load density. The practical results showed the model and the method in this paper were effective and high accuracy in spatial electric load forecasting.

关 键 词：电力系统 空间负荷预测 区间相似度 双层贝叶斯分类 类内相似度 

分 类 号：TM73[电气工程—电力系统及自动化]