基因多点交叉遗传算法在变压器故障诊断中的应用

APPLICATION OF MULTI-POINT CRISS-CROSS GENETIC ALGORITHM IN TRANSFORMER FAULT DIAGNOSIS

文中提出了一种基因多点交叉遗传算法，设计了基于该遗传算法、自动调整网络参数、连接权重和偏差的最优神经网络，建立了一套集溶解气体分析(DGA)技术、遗传算法和神经网络为一体的变压器故障诊断系统。由于基因多点交叉遗传算法的全局搜索能力和神经网络的高度非线性映射属性，文中的故障诊断系统能够较好地自动识辨变压器油中溶解气体与故障的对应关系，离线试验和现场运行结果表明，该诊断系统对变压器的过热、放电和受潮等故障诊断有一定的准确性。

A novel multi-point criss-cross genetic algorithm (GA) is put forward. On this basis an optimal neural network which can automatically adjust the network parameters, connection weights and deviation is designed, then a transformer fault diagnosis system integrating the dissolved gas analysis (DGA) technology, GA and neural network is established. Because of the global search ability of multi-point criss-cross genetic algorithm and the highly nonlinear mapping attribute of the neural networks, the presented fault diagnosis system can automatically identify the corresponding relation between the gas dissolved in the transformer oil and the fault. The results of the off-line test and on-site operation show that this diagnosis system can be used to diagnose the overheating, discharging and wetting of transformers with certain accuracy.