基于集成学习算法的供暖室内温度预测研究

Research on Prediction of Heating Indoor Temperature Based on Ensemble Learning Algorithm

提出利用基于多元线性回归模型和MLP神经网络模型的集成学习算法对供暖室内温度预测进行研究。以北京市某小区作为研究对象,选取30 d供暖数据和室外温度数据,将预测时刻之前6个时刻的室外温度、一级管网供水温度、一级管网回水温度、二级管网供水温度、二级管网回水温度,共30个特征值作为模型的输入,将下一时刻的室内温度作为模型的输出。研究结果表明,采用集成学习模型的平均相对误差和均方误差均小于单个模型的多元线性回归模型和MLP神经网络模型,预测效果较好,平均相对误差为0.8022%,均方误差为0.057665℃2。

An ensemble learning algorithm based on multiple linear regression model and MLP neural network model is used to study the prediction of heating indoor temperature.Taking a residential area in Beijing as the research object,30-day heating data and outdoor temperature data are selected,a total of 30 eigenvalues,including the outdoor temperature,water supply temperature of primary heating network,return water temperature of primary heating network,water supply temperature of secondary heating network and return water temperature of secondary heating network at the 6 times before the prediction,are taken as input parameters of the model,and the indoor temperature at the next time is taken as the output parameter of the model.The study results show that the mean relative error and mean square error of the ensemble learning model are both smaller than that of multiple linear regression model and MLP neural network model of single model,and the prediction effect is better,with a mean relative error of 0.8022%and a mean square error of 0.057665℃2.