The prediction of solar radiation heat in green building windows suffers from low real-time performance and poor accuracy.In order to solve these problems and improve the practicability of the research,this paper proposes a prediction algorithm for indoor solar radiation heat based on data processing and data prediction,namely the PKM_SVR algorithm,which is based on the correlation analysis and clustering analysis.Firstly,the algorithm preprocesses the external environmental data,and then constructs the main building in Dynamic Daylighting.Secondly,the algorithm matches the outdoor environmental data,and calculates the indoor radiant heat block by block in this state by selecting the window transmittance.Then,the data of indoor radiant heat and outdoor environment are processed by correlation analysis and clustering,and the processed data are divided into training and test sets.Finally,the initial prediction model of the indoor radiant heat of optimized PKM_SVR is constructed based on the cross-validation meth-od.The simulation results show that the ten-fold crossover algorithm is the most efficient for model optimization,when G=0.692,C=5.824,and R2=0.925.Simulation results show that,compared with other baseline algorithm models,the PKM_SVR model has the lowest RMSE index parameters,which are reduced by 23.1% and 31.3% compared with ANN and SVM models,respectively,and has higher R2 index parameters,which is increased by 10.7% on aver-age.To sum up,the PKM_SVR algorithm effectively improves the accuracy of indoor heat prediction,and has important simulation value.
维普
万方数据