多信息量分布式小波神经网络在电力系统故障测距中的应用

Application of distributed wavelet neural network based on multi-information in electric power system fault locating

提出了一种利用单端多信息量的测距方案。该方案采用分布式小波神经网络,通过模块化设计,充分考虑了高压线路中影响测距精度的各个因素,并针对常规测距中对于线路始端和终端测距精度不高的缺点,给出了运用泛化层解决的方案,实现了高压输电线路的高精度测距要求。该方案可以避免常规方案中出现伪根,迭代不收敛,以及消除对端系统运行方式和助增电流影响导致测距误差大等不足。该方法可以根据现场运行数据进行训练,具有再学习的能力。ATP和Matlab仿真表明,该方案测距精度高,适应性强,性能可靠。

A fault locating scheme is proposed,which uses multi-information of single terminal. Considering the factors influencing HV line fault-locating precision,this scheme adopts distributed wavelet neural network and modular architecture. Aiming at the low locating accuracy of the fault near both ends,generalization is constructed to realize the high-precision fault locating. It avoids false root and divergence in iteration,and eliminates the great influences of load current and operating mode of opposite system on locating error. It is trained with field operation data and learns repeatedly. The simulation with ATP and Matlab indicates that the scheme is precise, adaptable and reliable.