基于自适应PSO-LSSVM多维开关柜数据的状态评估方法

为有效分析开关柜的各种监测数据,减轻状态评估工作量,提出一种基于自适应粒子群优化和最小二乘向量机分析(PSO-LSSVM)的多维开关柜数据状态评估方法。首先对开关柜的多维历史监测数据进行分类,提取有效数据构建数据集,并利用K最近邻法(KNN)填补缺失数据,针对已有的开关柜历史评估结果,采用最小二乘支持向量机算法进行数据分析和训练,引入自适应粒子群算法,实现算法参数的优化,最终获得多维数据的开关柜评估模型。经实际数据检验,此评估方法能有效分析历史监测所累积的大规模数据,评估准确率较高,可有效筛选出需检修的开关柜,减轻了开关柜评估工作量,具有一定的实际应用意义。

State Evaluation Method of Multidimensional Switchgear Data Based on Adaptive PSO-LSSVM

In order to analyze the various monitoring data of the switchgear effectively and reduce the state evaluation work, this paper proposes an evaluation method of multidimensional switchgear data based on self-adaptation particle swarm optimization and least squares support vector machine(PSO-LSSVM). Firstly, the multidimensional historical monitoring data of the switchgear was classified and the available data was extracted to establish the data set. Then the missing data was filled by using the method of K-Nearest Neighbor (KNN). According to the existing history evaluation results, data was analyzed and trained by using the algorithm of LSSVM. The parameters were optimized by introducing the algorithm of the adaptive PSO. Ultimately the switchgear assessment model of multidimensional data was obtained. Through the practical data test, the proposed method can effectively analyze historical accumulated monitoring big data, and evaluation accuracy is high. It can effectively filter out the switchgear which needs an overhaul and reduce switchgear assessment work, which has a certain practical application significance.