智能技术在距离保护中的应用

针对传统的距离保护方法存在许多不足,提出了一种将BP网络与径向基函数网络相结合的距离保护模型。分别采用BP网络构建振荡识别子网络,采用RBFN构成的故障检测定位子网络,并采用附加动量项和自适应学习率等措施增加BP算法学习过程的稳定性和收敛速度;为增加网络训练的确定性,文中进行了大量的仿真计算,结果表明,该方法能够对电力系统的各种故障情况进行正确的识别并且比目前常用的神经网络式距离保护有更高的工作效率。文章最后也分析了神经网络技术在距离保护中存在的不确定性,指出了消除不确定性的措施。

Intelligent technique application in distance relay protection

For the disadvantages of the classic methods of distance relay protection, a distance relay protection mode of combining BP network and radial basis function network (RBFN)is proposed. The BP network is adopted to oscillate identification sub-network, the radial basis function network is used to fault position sub-network, and measures such as additional momentum item and self-adaptive algorithm are employed to increase the certainty of network. The experiment results show that this method is reliably used in the distance protection of power system and has higher work efficiency compared with the method based on commonly artificial neural network. In the end, the uncertainty of the ANN applied in the distance relay protection is analyzed, and the method to avoid the uncertainty is proposed.