摘要
电能质量问题日益严重,实现稳态电能指标的预测对于保障供电质量有重大意义。以某低电压台区监测点为研究对象,提出一种基于随机森林(Random Forest,RF)特征优选和粒子群优化(Particle Swarm Optimization,PSO)反向传播(Back Propagation,BP)神经网络的电能质量稳态指标预测方法。根据监测点的环境因素、动态电能数据以及电能质量指标的历史数据等多维度特征进行标准化处理,通过RF算法实现特征选择后,进一步将优选特征输入到神经网络,并结合PSO算法优化BP神经网络的权值和阈值,构建神经网络预测模型。将该方法结合实例与传统神经网络进行对比试验,结果表明,所提预测方法的MAPE均低于3%,预测效果较传统神经网络更佳。
The problem of power quality is becoming more and more serious.It is of great significance to realize the prediction of steady-state power index for ensuring power supply quality.Based on Random Forest(RF)feature selection and Particle Swarm Optimization(PSO)Back Propagation,a new method based on Random Forest(RF)feature selection and Particle Swarm Optimization(PSO)Back Propagation is proposed.The power quality steady-state index prediction method of BP neural network is standardized according to the multi-dimensional features of the monitoring point,such as environmental factors,dynamic power data and historical data of power quality index.After feature selection is realized by RF algorithm,the preferred features are further input into the neural network,and the weight and threshold of BP neural network are optimized by combining the PSO algorithm.Construct the neural network prediction model.The results show that the MAPE of the proposed method is lower than 3%,and the prediction effect is better than that of the traditional neural network.
作者
向星宇
刘敬之
曲全磊
夏得青
罗政
黎朝晖
XIANG Xingyu;LIU Jingzhi;QU Quanlei;XIA Deqing;LUO Zheng;LI Zhaohui(College of Electrical and Information Engineering,Hunan University of Technology,Zhuzhou 412007,China;State Grid Qinghai Electric Power Research Institute,Xining 810001,China;College of Railway Transportation,Hunan University of Technology,Zhuzhou 412007,China)
出处
《电子设计工程》
2023年第22期116-120,共5页
Electronic Design Engineering
关键词
电能质量
特征选择
随机森林
粒子群优化
BP神经网络
powerquality
featureselection
random forest
particleswarm optimization
BP neuralnetwork
