基于混合算法与支持向量机的电力变压器故障诊断

Fault Diagnosis of Power Transformer Based on Hybrid Algorithm and Support Vector Machine

为了改进、优化支持向量机的核函数参数σ以及惩罚因子C,提出了基于粒子群与差分进化相结合的混合优化算法(PSODE),从而获得最优的故障诊断模型。通过引入四种基准测试函数:Sphere函数、Rosenbrock函数、Rastrigin函数、Griewank函数对PSO、DE、PSODE的性能进行测试,仿真结果表明PSODE是一种具有较强优化性能的算法。为了进一步验证该方法的有效性,通过对油中溶解的H2、CH4、C2H6、C2H4、C2H2的含量进行分析,可以较准确地识别低温过热、中温过热、高温过热、局部放电、火花放电、电弧放电以及正常状态。

In order to improve and optimize the kemel function of support vector machines（SVM） is a parameter sigma and penalty factor C, proposing a hybrid optimization algorithm（PSODE） which based on the combination with differential evolution and support vector machine, obtaining the optimal model of fault diagnosis. By introducing four benchmark functions：sphere function, rosenbrock function and rastrigin function, griewank function to test the performance of the PSO, DE, PSODE, the simulation results show that performance of PSODE is a kind of strong optimization algorithm. In order to further verify the effectiveness of the method, analysing the content of n2, CH4, C2H6, C2H4, C2H2, which dis- solved in oil can accurately identify low overheating, medium temperature overheating, high temperature overheating, par- tial discharge, spark discharge and arc discharge and normal state.