改进的人工免疫分类算法在故障类型识别中的应用

Application of improved artificial immune network classifier for failure identification

提出将小波包变换和改进的免疫算法相结合,对输电线路故障类型进行识别。运用小波包将电压故障信号分解,提取三相的小波奇异熵作为免疫网络的抗原,利用免疫网络抗原-抗体识别原理进行故障类型识别。仿真结果表明:在相同实验条件下,与传统的ANN网络和SVM相比,该算法具有自适应连续学习的功能,对故障诊断系统可以连续不断的补充新样本。并且此故障类型识别方法不受系统运行方式、过渡电阻和故障位置等影响,具有较强的通用性,较高的精度,识别速度快和算法简单易实现。

Wavelet packet transform and the improved immune algorithm are combined for transmission line fault type identification. Wavelet packet is used to decompose fault voltages signals,and then fault three-phase wavelet singular entropy is extracted and regarded as the antigen of immune network.Fault type is recognized based on antigen-antibody recognition principle of immune network.Simulation shows that compared with traditional ANN and SVMs,the algorithm has adaptive and continuous learning ability so that the new samples can be continuously added to the fault diagnosis system; and the algorithm for fault phase selection is free from the influence of the factors such as the operation mode,the transition resistance and the fault location.It is highly universal,of high accuracy,and is fast in calculation and easy to realize.