基于改进FOA LSSVM的变压器故障诊断

针对基于油中溶解气体（DGA）变压器故障诊断方法存在的不足，提出了改进果蝇算法优化LSSVM的变压器故障诊断方法。为克服果蝇算法易陷入局部极值、收敛精度低的缺陷，引入线性递减步长、变异操作和混沌搜索策略进行改进。建立了LSSVM变压器故障诊断模型，并用改进的果蝇算法优化参数。仿真试验及对比研究表明，改进模型可准确、有效地识别变压器故障类型，相较其他模型（BPNN、FOA PNN和FOA LSSVM），该模型的准确率较高，更适合于变压器故障诊断。

Transformer Fault Diagnosis Based on LSSVM with Improved Fruit Fly Optimization

Considering the problems of traditional transformer fault diagnosis techniques based on dissolved gas analysis (DGA), a strategy based on LSSVM with improved fruit fly optimization for transformer fault diagnosis is proposed. In order to overcome the shortcomings of trapping into local extremum and slow convergence speed for fruit fly optimization algorithm, linear descending step, mutation operation and chaotic search are introduced to improve fruit fly optimization algorithm. The traditional transformer fault diagnosis model is established by least squares support vector machine and the parameters are optimized by improved fruit fly optimization algorithm. The test results show that transformer faults type can identified accurately and effectively, the accuracy is higher compared with other models (BPNN, FOA PNN and FOA LSSVM). So, the new method has better application value.