基于改进PSO-SVM的电能计量装置异常状态识别研究

Research on Recognition of Abnormal State of Electric Energy Metering Device Based on Improved PSO-SVM

电能计量装置出现异常状态后最终表现上传的数据异常,利用计量系统自动化数据进行分析,尽早识别装置异常有利于供电企业进行装置检修,提升运维能力。针对异常数据,人工使用判断规则对其进行诊断需要大量时间,计量装置出现问题后不能及时发现的问题,对已标记的异常数据进行研究,使用PSO-SVM模型对计量数据进行分析,判别装置状态异常。首先,针对粒子群算法易陷入局部最优的问题,将粒子移动的权重更新方式结合当前粒子的迭代次数和适应度值进行动态改变;其次,引入偏二叉树结构的支持向量机分类模型,并且树中每层的分类器使用改进后的粒子群算法寻找分类最佳超参数,增加分类准确率;最后,使用电能计量装置对改进的PSO-SVM模型进行验证,结果表明该方法能够较好地识别出异常数据。

The abnormal state of the electric energy metering device will result in abnormal data finally uploaded.Using the automated data of the metering system to analyze and identify the abnormality of the device as early as possible is beneficial to the power supply enterprise to perform device maintenance and improve the operation and maintenance capabilities.Aiming at solving problems,such as abnormal data that requires a lot of time to be diagnosed through judgment rules manually,and problems that cannot be discovered in time after the measurement device has faults:the marked abnormal data is studied.And the PSO-SVM model is used to analyze the measurement data to determine the abnormal state of the device.First,in view of the problem that the particle swarm algorithm is easy to fall into the local optimum,the weight update method of the particle movement is changed by combining the iterations and the fitness value of the current particle;second,the support vector machine classification model of the partial binary tree structure is introduced,and the classifiers at each level in the tree use the improved particle swarm algorithm to find the best hyper parameters to increase the accuracy of classification.Finally,the improved PSO-SVM model is verified by the abnormal data of the electric energy metering device,and the results show that it can identify abnormal data better.