Association Network Inference Of Power Equipment Monitoring Parameters And Its Topological Analysis

By using the method of complex network research complex system has become an important research topic, and there are some successful applications, but have not yet applied to electric power equipment monitoring system. Electric power equipment's parameters are numerous,but the relationship between these parameters is unknown, therefore,electric power equipment monitoring parameters associated network inference and structure analysis is a challenging task. Around this topic,this paper makes the following main aspects:First of all, the relationship between the parameters identification of the commonly used method and the analysis of network structure,summarizes the related theory. Main content includes: (1) monitoring parameter binary relation analysis: correlation coefficient and partial correlation coefficient, the neural network, the transfer entropy, etc. (3)correlation identification method of multual information: the definition of entropy, mutual information and mutual information estimation and DPI algorithm; (4) the complex network topology characteristics of commonly used: node degree and degree distribution, average path length, clustering coefficient and node betweenness centrality.Second, deeper analysis for the power equipment monitoring parameters. Main content includes: data preprocessing, monitoring parameters of time series analysis and comparative analysis.Moreover, based on the correlation coefficient and mutual information respectively the associated network inference of parameters.Main content including implementation methods and steps. And analyzes the mutual information to reconstruct the rationality of the network,advantages.Finally, the inferences from associated network adopt some measures in the structure analysis. Main content including important node analysis, network robustness analysis, subgroup analysis and so on. And in view of the important node analysis, using multiple linear regression model and regression tree structure and prediction model, using the NMSE to evaluate performance of the model.This article is based on the monitoring data of power equipment, and the use of two methods for parameters associated network inference. Not only explains the relevant theory research, and has carried on the concrete implementation. Inferences from by comparing with the actual equipment network, the rationality of the method is validated. Through analyzing the characteristic of the structure of the network, to reveal the phenomenon of hidden behind the data. Therefore, this article for the power equipment monitoring parameters for the further study of the relation network has a certain reference significance.