基于神经网络空间环境设计预测舒适度模型建构  

The Construction of the Comfort Modelof Interior Space Environment Design Prediction Basedon Neural Network

作　　者：韦鸾鸾[1] 高静静[2] WEI Luanluan;GAO Jingjing(Anhui Business and Technology College,Hefei230041,China;HuaibeiVocational and Technical College,Huaibei Anhui235000,China)

机构地区：[1]安徽工商职业学院,安徽合肥230041 [2]淮北职业技术学院,安徽淮北235000

出　　处：《佳木斯大学学报（自然科学版）》2024年第11期30-33,87,共5页Journal of Jiamusi University：Natural Science Edition

基　　金：2022年度高等学校省级质量工程项目(2022zygzsj008);2022年度安徽省高等学校省级质量工程教学研究项目(2022jyxm162)。

摘　　要：通过环境舒适度指标计算,获取影响空间环境设计舒适度的5个PMV变量指标;将其作为采用粒子群优化算法改进权值阈值后的RBF神经网络模型的输入,经归一化变形法完成PMV变换后,结合舒适度评价标准获取最终空间环境设计的舒适度等级。经测试分析得出:所建构模型的空间环境设计的舒适度预测输出值和期望输出值的拟合度极高,平均预测误差仅为0.21%;并且能够准确地预测出不同时刻的空间环境设计的舒适度。Through the calculation of environmental comfort index,five PMV variable indexes that affect the comfort degree of indoor space environment design are obtained.They are taken as the input of RBF neural network model after the weight threshold is improved by particle swarm optimization algorithm.After PMV transformation is completed by normalized deformation method,combined with the comfort evaluation standard.The final comfort level of indoor space environment design is obtained.Through the test and analysis,it is concluded that the fitting degree of the comfort prediction output value and the expected output value of the model is very high,and the average prediction error is only 0.21%,and it can accurately predict the comfort degree of the interior space environment design at different times.

关 键 词：神经网络 环境舒适度 预测模型 PMV指标 变量指标 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程]