Research On Comprehensive Evaluation And Association Rule Mining Method On Operating State Of Low-voltage Distribution Network

In recent years,the development focus of power grid in China has shifted from transmission network to distribution network,and the transforming and upgrading of lowvoltage distribution network is unprecedented.The comprehensive evaluation can accurately reflect the low-voltage network operating state and the satisfaction degree of each index with the quantitative results,which is helpful for selecting bad operating state networks and transforming them targetedly.However,there is a lack of scientific and reasonable evaluation index system and effective weighting method.Association rules describe concisely the potential relationship between operation state levels and indices,combining association rules and comprehensive evaluation can provide decision support for power supply departments to efficiently and quickly select bad operating state networks from massive ones and to design trageted optimization program.Therefore,this paper focuses on the two primary links in the comprehensive evaluation of low-voltage network operating state,that is,index system and index weight,as well as association rules mining between operating state and indices.Firstly,considering one sidedness,poor applicability and lack of overall situation in the traditional low-voltage network evaluation index system,low-voltage network is divided into three sub objectives,including distribution transformer,low-voltage line and power consumers by using analysis method to comprehensively reflect the operating state;In order to ensure the applicability of the index system,the easily available statistical indices are used to describe the sub objectives;The distributive index is centralized to represent the satisfaction degree of the whole network on the index,so as to better meet the overall situation.Secondly,in view of the index redundancy in the initial index system,the comprehensive and independent index system is obtained by screening indices based on the rough set attribute reduction principle.Four key indices,transformer load rate,voltage drop per unit length,unbalance degree of three-phase load and qualified rate of power consumers voltage are selected by using heuristic reduction algorithm of attribute frequency in discernibility matrix,That the reductive index system is scientific and reasonable is verified.The reductive index system simplifies the comprehensive evaluation process,makes the evaluation results more accurate,and provides a clearer direction for the optimizing and transforming networks.Then,considering the poor operability of traditional objective weighting method for imprecise data,using the principle of attribute importance of rough set,the objective weight is determined by describing the classification ability of each index based on the principle of rough set attribute importance.Through "addition" integration,the subjective weights and objective weights are aggregated into comprehensive weights,and the comprehensive evaluation function of network operating state is thus obtained.In the example,the ranking of the evaluation results of 10 networks is consistent with that of the previous evaluation,which verifies the effectiveness of the evaluation method.The comprehensive evaluation results quantify the operating state,and scoring value of each index reflects its own satisfaction degree,which can provide basis for ranking networks from good to bad and targeted transforming.Finally,in consideration that comprehensive evaluation needs to collect all the indices data before evaluation results determined,association rule is introduced,and only a few indices are needed to deduce the operating state levels of some networks,which realizes estimating operating state fast.Considering the low efficiency of classical Apriori algorithm,the algorithm is improved by integrating the partition technology and knowledge type constraints.After analyzing the association rules of the training networks data set,several relationships between the operating state levels and indices are found,which provide simple and practical decision supports to select quickly the networks that need not be optimized or transformed and ones of bad operating state from massive networks.