基于改进粒子群优化LS-SVM的变压器故障气体预测

Predicting method for dissolved gas in transformer oil based on modified particle swarm optimization

最小二乘支持向量机(LS-SVM)能较好的解决小样本、非线性数据特征的多分类问题,适用于电力变压器油色谱故障气体预测,但参数c与σ2的选取对预测结果影响较大,有必要对其进行优化选择。提出一种基于改进粒子群(MPSO)的参数寻优方法,并将其应用到变压器油中故障气体预测。改进粒子群算法在每次迭代中,将粒子群进行分类,不同类粒子采取不同加速因子,相比较经典粒子群算法,可以有效扩大粒子搜索区间,增强其局部搜索能力。最后进行了多组现场数据的实例分析,结果表明:基于MPSO进行参数优化后的LS-SVM预测准确率明显优于传统LS-SVM预测结果。

LS-SVM（least square support vector machines） is applied to solve the practical problems of small samples and non-linear prediction better,and it is suitable for the forecasting dissolved gas in transformer oil.But in this model,the selecting values of the parameters,c and σ^2,impact on the result of the diagnosis greatly,so it is necessary to optimize those parameters.A new modified particle swarm optimization（MPSO） is proposed in the paper to optimize those parameters.The MPSO dynamically classifies the swarm population according to whether the historical best position found by the particle is changed or not,different groups make use of different parameters.This tactics enlarges the search scope and avoids falling into a local optimal solution.Finally,using the optimized model analysis multiple sets of oil chromatogram data,the results show that： the accuracy of the LS-SVM based on the MPSO is better than traditional LS-SVM model.