改进GA-PSO算法的船舶电力系统故障诊断

为解决船舶电力系统故障识别的准确性以及快速性问题,在BP神经网络预测的基础上,提出一种改进的粒子群(PSO)和遗传算法(GA)混合优化BP神经网络的方法。改进包括两方面：一是对粒子群的惯性权重和学习因子进行改进；二是对遗传算法的变异概率和交叉概率进行改进。对发生故障时的三相电压信号进行小波包分解,提取各频率段的能量熵作为故障特征。经测试,优化后的算法诊断准确率明显提高,神经网络训练次数和误差减小,验证了改进GA-PSO-BP算法的可靠性,以及用于船舶电力系统故障诊断的实用性。

Fault diagnosis of ship power system based on improved GA-PSO algorithm

In order to solve the problem of accuracy and rapidity of ship power system fault identification, based on BP neural network prediction, an improved particle swarm optimization (PSO) and genetic algorithm (GA) hybrid BP neural network method is proposed. The improvement includes two aspects: one is to improve the inertia weight and learning factor of the particle swarm; the other is to improve the mutation probability and crossover probability of the genetic algorithm. Wavelet packet decomposition is performed on the three-phase voltage signal at the time of failure, and the energy entropy of each frequency segment is extracted as a fault feature. After testing, the accuracy of the optimized algorithm is obviously improved, the number of neural network training and the error are reduced, and the reliability of the improved GA-PSO-BP algorithm and the practicability for fault diagnosis of the ship power system are verified.