基于神经网络的电力计量故障诊断及仿真研究

Electric power metering fault diagnosis and simulation based on neural network

针对传统的BP算法在收敛速度和易陷入局部最小值的问题,结合高压电力计量的重要性,提出一种基于改进BP网络的电力计量故障诊断算法。文章在分析电力计量系统工作原理和检测原理的基础上,对BP算法问题进行深入分析,然后结合故障诊断的理论,引入小波分析对特征信号进行处理,减少特征信号中的噪声和尖峰成分,提高故障诊断的正确率。在信号特征处理的基础上,结合BP算法的问题,提出引入PSO优化算法在全局搜索方面的优势,从而防止传统BP算法容易陷入局部极小值的问题,最终时期更靠近群体最优。最后通过和BP网络的诊断测试对比,验证本文设计算法的可行性与正确性,从而为当前电力计量诊断提供了借鉴。

A fault diagnosis algorithm for power metering based on improved BP neural network is proposed,aiming at the problem of convergence speed and easily failing into local minimum value of traditional BP algorithm,and considerimg the importance of high voltage power metering. In this regard,based on the analysis of the working principle and detection principle of the power metering system,the problems of the BP algorithm is thoroughly analyzed,and then combining the theory of fault diagnosis,introducing the wavelet analysis to process the characteristic signals,and reducing the noise and spikes in the characteristic signals to improve the accuracy of fault diagnosis. Based on the processing of signal features and combining with the problems of BP algorithm,this paper proposes to introduce the advantages of PSO optimization algorithm in global search so as to prevent the traditional BP algorithm from falling into the problem of local minimum value. In the final period,it is closer to the group optimality. Finally,by comparing with the diagnostic test of BP network,this paper verifies the feasibility and correctness of the proposed algorithm,which provides a reference for the current power metering diagnosis.