基于长短期记忆循环神经网络的变电站监控系统智能故障推理方法

Intelligent fault reasoning method for the substation monitoring system based on LSTM

针对变电站故障推理和分析应用存在人工总结的规则不全面、总结难度大、干扰信号多、故障推理配置可重用性低,以及故障推理往往需要考虑输入信号的时序性等问题,而采用传统机器学习算法无法有效解决此问题,提出一种基于长短期记忆循环神经网络(LSTM-RNN)、自然语言处理技术的变电站智能故障推理方法。分析了故障推理的应用场景,介绍了智能故障推理方法的整体架构、关键技术,并通过实际数据的应用试验进行了测试,验证了不依赖人工规则的智能故障推理方法的可行性,在信号时序可以记忆的场景中LSTM-RNN比其他机器学习算法有更好的适用性。

The substation fault reasoning rules and analysis applications summarized by human are incomplete,difficult and vulnerable to interfering signals. Fault reasoning’s configurations are of low reusability and have to take time sequence of input signals into consideration, which cannot be effectively solved by traditional machine learning algorithms. Thus, a substation intelligent fault reasoning method based on long-short-term memory recurrent neural network(LSTM-RNN) and natural language processing(NLP) technology is proposed. Based on the analysis on the application scenarios of fault reasoning, the overall architecture and key technologies of the intelligent fault reasoning method are expounded. The feasibility of the intelligent fault reasoning method that does not rely on man-made rules is verified by the data of application tests,and the LSTM-RNN works better than other machine learning algorithms in the scenarios that time sequence of signals can be memorized.