摘要
现有配电网风险评价中普遍存在专家评价主观性强和无效数据处理不严谨2个现实难点,据此提出一种基于BP神经网络—动态贝叶斯网络的配电网运行风险预警模型。首先运用BP神经网络集进行评价属性的训练和测试,避免了模糊综合评价中指标权重设置和指标专家评价值主观性强的问题;其次将多组BP神经网络的输出进行融合处理,剔除差异性较大的观测数据,最后将部分数据缺失的时间片观测值矩阵作为变结构动态贝叶斯网络的输入,对指标隶属度在时间序列上的动态演化进行分析,达到对配电网运行风险动态评估预警的目标。在实例分析中,通过该模型与模糊综合评价方法的评价结果进行对比,证明了该模型有效可行。
There exist two problems in power distribution network operational risks assessment: high expert evaluation subjectivity and invalid data processing. Therefore,this paper presents a distribution network operational risk prediction model based on the dynamic Bayesian network of BP neural network. First,BP neural network group is used to train and test the evaluation properties in order to avoid the high subjectivity in the index weight setting of fuzzy comprehensive evaluation and in the expert evaluation index values. Then the output of multiple BP neural network is fusion processed and the observed data with large differences are excluded. Finally,the time-sliced observed data matrix with missing data is inputted to dynamic Bayesian network with variable structures. The dynamic evolution on time series of the index membership degree is analyzed for the early warning of dynamic evaluation of distribution network operation risk. Finally,the comparison with the fuzzy comprehensive evaluation in the example analysis proves the validity and feasibility of the proposed model.
出处
《华东电力》
北大核心
2014年第6期1109-1114,共6页
East China Electric Power
基金
国家自然科学基金项目(71071054
71271084)
中央高校基本科研业务费专项资金资助(2014MS40)~~
关键词
配电网运行
变结构贝叶斯网络
风险预警
数据缺失
power distribution operation
structure-variable dynamic bayesian networks
risk warning
missing data