基于IWOA优化LSSVM的煤矿变压器故障诊断研究

Research on Fault Diagnosis of Coal Mine Transformer Based on IWOA Optimized LSSVM

为了快速地分辨出变压器故障类型提高故障诊断的准确率,提出了一种改进的鲸鱼算法(IWOA)优化最小二乘支持向量机(LSSVM)的变压器故障诊断模型。利用核主成分分析(KPCA)对冗杂繁多的数据进行降维处理,减少无效特征;采用启发式概率、融合正弦函数优化的动态权重、优化比例系数对鲸鱼算法进行改进,提高其优化能力,并与鲸鱼算法(WOA)和粒子群算法(PSO)进行性能测试对比,验证算法有效性;利用改进的鲸鱼算法对LSSVM的相关超参数进行寻优求解,避免算法出现早熟问题,提高变压器故障诊断的准确度。并对模型进行模拟仿真实验,仿真结果表明准确率达到93.33%,相对于WOA-LSSVM模型和PSO-LSSVM模型分别提高了6.66%、10%,具有良好的故障诊断效果。

In order to quickly identify transformer fault types and improve the accuracy of fault diagnosis,an improved Whale Algorithm(IWOA)optimized least squares support vector machine(LSSVM)transformer fault diagnosis model is proposed.Using kernel principal component analysis(KPCA)to reduce the dimensionality of complex and diverse data and reduce invalid features.The Whale Algorithm by using heuristic probability,dynamic weights optimized by fusing sine functions,and optimized proportional coefficients are improved to enhance its optimization ability.Performance tests are conducted and the algorithm with whale algorithm(WOA)and particle swarm optimization(PSO)are compared to verify its effectiveness.Using the improved whale algorithm to optimize and solve the relevant hyperparameters of LSSVM,premature problems in the algorithm and improving the accuracy of transformer fault diagnosis are avoided.And simulation experiments are conducted on the model,and the simulation results show that the accuracy reaches 93.33%,which increases by 6.66% and 10% compared to the WOA-LSSVM model and PSO-LSSVM model,respectively,showing good fault diagnosis performance.