基于LS-SVM的变压器最优维护周期研究

Research on Optimal Transformer Maintenance Scheme Based on LS-SVM

优化变压器维护周期可提高变电站可靠性,降低变电站运行维护成本。基于最小二乘支持向量机(least squaressupport vector machines,LS-SVM)算法,结合成本–效益分析法对变压器维护周期优化进行研究。首先利用统计数据构建变压器缺陷树,并将缺陷数据、停电时间及缺陷权重专家数据等综合起来作为基础,在采用两层动态自适应优化法确定LS-SVM参数后,利用LS-SVM算法对变更维护计划后的缺陷数进行预测,将维护变更的成本(效益比较结果)作为量化约束条件,确定变压器的最优维护周期。采用该算法对某供电公司变压器进行评估,对变压器年度维护计划进行修正并实用,取得了较好的效果。

Optimization of transformer maintenance scheme can enhance substation reliability and lower maintenance cost. Based on least squares support vector machines （LS-SVM） method, combing with cost-effect effectiveness method, research on optimized transformer maintenance scheme was carried out. The paper built a transformer deficiencies tree from statistical data. Aggregating deficiency data, outage duration, maintenance operation deficiency checking out rate and etc as basis, a two layer dynamical adjustment optimization method was used to choose LS-SVM parameters, then LS-SVM algorithm was used to predict deficiencies before and after scheme optimization, and cost effect measure was used to determine the best scheme. This algorithm has been used to evaluate transformer status in a power supply company and modify transformer maintenance scheme, and ideal result is achieved.