mRMR和PSO算法对神经网络预测模型优化效果
Optimization Effect of mRMR and PSO Algorithms on Neural Network Prediction Models
杜润琪 1于丹 1刘益民 2岑悦2
作者信息
- 1. 北京建筑大学,北京 102627
- 2. 中国建筑科学研究院有限公司,北京 100013
- 折叠
摘要
提出利用最大相关和最小冗余(mRMR)算法、粒子群优化(PSO)算法,对BP神经网络预测模型进行优化.对某住宅楼进行供热负荷预测,评价3种神经网络预测模型(BP神经网络预测模型、mRMR-BP神经网络预测模型、PSO-mRMR-BP神经网络预测模型)的预测效果.在3种神经网络预测模型中,BP神经网络预测模型的预测效果最差,PSO-mRMR-BP神经网络预测模型的预测效果最佳.与BP神经网络预测模型相比,经过mRMR算法对输入变量进行筛选以及PSO算法对初始参数进行优化,PSO-mRMR-BP神经网络预测模型的预测效果显著提高.
Abstract
It is proposed to use the maximum relevance minimum redundancy(mRMR)algorithm and the particle swarm optimization(PSO)algorithm to optimize the BP neural network prediction model.The heating load of a residential building is predicted,and the prediction effects of three neural network prediction models(BP neural network prediction model,mRMR-BP neural network prediction model,and PSO-mRMR-BP neural network prediction model)are evaluated.Among the three neural network prediction models,the BP neural network prediction model has the worst prediction effect,and the PSO-mRMR-BP neural network prediction model has the best prediction effect.Compared with the BP neural network prediction model,through the mRMR algorithm to screen input variables and the PSO algorithm to opti-mize the initial parameters,the prediction effect of the PSO-mRMR-BP neural network prediction model is sig-nificantly improved.
关键词
供热负荷/预测/BP神经网络/mRMR算法/PSO算法Key words
heating load/prediction/BP neural network/mRMR algorithm/PSO algorithm引用本文复制引用
出版年
2024
国家科技期刊平台
NSTL
维普
万方数据