基于贝叶斯网络的电力变压器局部放电故障检测

Partial discharge fault detection of power transformer based on bayesian network

摘要:

针对传统电力变压器故障检测方法对电力系统中潜藏的故障问题检测水平不足，准确率较低，无法及时准确的发现异常隐患等问题，本文提出了一种基于贝叶斯网络的变压器局部放电故障检测方法，首先通过传感器获取电力变压器不同状态下运行过程中的参数数据，对局部放电故障发生的概率和范围进行合理性评估，提取评估概率数据综合为样本数据集，构建贝叶斯网络故障树；根据逻辑规则转化为贝叶斯网络，推演计算故障节点之间的算例关系，利用贝叶斯原理抽取故障特征指标与异常概率之间的关联关系，利用模糊描述方法构建故障特征关联函数，计算可得故障特征模糊函数动态变化关系，实现对变压器故障发生的概率与位置信息的判断与确定。通过实验结果可以证明，通过贝叶斯网络对电力变压器局部放电故障检测的准确率均达到了85%以上，最高可达96%，说明该方法具有较高的检测准确率，能够有效提高电力变压器放电故障检测的有效性。

Abstract:

In view of the problems of the traditional power transformer fault detection methods, such as insufficient detection level, low accuracy, and inability to detect abnormal hidden dangers in a timely and accurate manner, this paper proposes a transformer partial discharge fault detection method based on Bayesian network, which first obtains the parameter data of the power transformer in different operating conditions through sensors, Reasonably evaluate the probability and scope of partial discharge fault occurrence, extract and synthesize the evaluation probability data into sample data set, and construct Bayesian network fault tree; Convert the logic rules into Bayesian network, deduce and calculate the example relationship between the fault nodes, extract the correlation between the fault characteristic index and the abnormal probability using Bayesian principle, construct the fault characteristic correlation function using fuzzy description method, calculate the dynamic change relationship of the fault characteristic fuzzy function, and realize the judgment and determination of the probability and location information of the transformer fault. The experimental results can prove that the accuracy of partial discharge fault detection of power transformer through Bayesian network is more than 85%, and the maximum is 96%, which shows that this method has high detection accuracy and can effectively improve the effectiveness of power transformer discharge fault detection.