遗传神经网络下光伏功率高比例异常数据检测  

Detection of abnormal data with high proportion of photovoltaic power based on genetic neural network

作　　者：汪鹤[1] 董晓峰 沈健[1] WANG He;DONG Xiaofeng;SHEN Jian(Guodian Nanrui Technology Co.,Ltd.,Nanjing 211106,China;Suzhou Power Supply Branch,State Grid Jiangsu Electric Power Co.,Ltd.,Suzhou 215000,China)

机构地区：[1]国电南瑞科技股份有限公司,江苏南京211106 [2]国网江苏省电力有限公司苏州供电分公司,江苏苏州215000

出　　处：《电子设计工程》2024年第22期87-90,95,共5页Electronic Design Engineering

基　　金：江苏省自然科学基金资助项目(BK20221165)。

摘　　要：在天气、设备以及多个因素的影响下,大部分光伏功率易出现异常。因此,该文提出基于遗传神经网络的光伏功率高比例异常数据检测方法。分析光伏功率高比例异常数据聚集特征,并利用遗传神经网络架构确定光伏发电条件概率分布。结合最小化估算区间原理,在确定高比例异常光伏概率分布情况下,估计光伏概率所在区间。利用遗传算法调整神经网络权值,确定神经细胞异常分数以及离群点异常分数平均值,从而判断当前数据是否为高比例异常数据,并得到光伏功率高比例异常数据检测结果。实验结果表明,该文方法能够有效检测出异常数据,误差小,实际应用效果好。Under the influence of weather,equipment,and multiple factors,most of the photovoltaic power will experience abnormalities.Therefore,this article proposes a genetic neural network⁃based method for detecting high proportion abnormal data of photovoltaic power.The aggregation characteristics of abnormal data with high proportion of photovoltaic power are analyzed,and the Conditional probability distribution of photovoltaic power generation is determined using genetic neural network architecture.Based on the principle of minimizing the estimation interval,estimate the interval where the photovoltaic probability is located when determining the probability distribution of high proportion abnormal photovoltaics.Genetic algorithm is used to adjust the weights of the neural network to determine the average abnormal scores of neural cells and outliers,in order to determine whether the current data is high proportion abnormal data and obtain the detection results of photovoltaic power high proportion abnormal data.The experimental results show that the proposed method can effectively detect abnormal data with small errors and good practical application results.

关 键 词：遗传神经网络 雨雪光伏功率 高比例异常数据 分布置信度 

分 类 号：TM93[电气工程—电力电子与电力传动] TN-9[电子电信]