基于贝叶斯网络的电网故障诊断

针对电网故障诊断中存在的信息具有不确定性的问题,依据元件故障、保护动作和断路器跳闸之间的内在逻辑关系,由Noisy-Or和Noisy-And节点组成贝叶斯网络和采用类似训练多层前馈神经网络的误差反传算法进行诊断模型的参数学习,分别建立了线路、变压器和母线的通用故障诊断模型;依据元件-保护-断路器间的关联关系,给出了元件诊断贝叶斯网络的自动生成方法,最后对各个元件的诊断网络进行推理,以获得元件的故障概率值。实例仿真表明了该诊断方法的可行性和有效性,无论简单故障或多重故障,并且存在保护和断路器拒动、误动的情况下,都能得到合理有效的诊断结果。

Power system fault diagnosis based on Bayesian network

According to the internal logic relationships among element fault,protective relay action and circuit breaker trip,the general fault diagnosis models of transmission lines,busbars and transformers are respectively established to solve the information uncertainty problem in power system fault diagnosis,which organizes a special Bayesian network composed of Noisy-Or and Noisy-And nodes,and uses back propagation algorithm in parameter learning.The Bayesian network for each element diagnosis is generated automatically according to the relationships among element,protective relays and circuit breakers.By element diagnosis network reasoning,the element fault probability is obtained.Instance simulations show the feasibility and effectiveness of the proposed fault diagnosis method for both simple and complex faults,even when there is malfunction of protective relays or circuit breakers.