Prediction And Analysis Of Short-Term Load Data For Electric Heating In University Buildings

Electric heating,as a new type of energy load,caters to the comprehensive energy-saving and emission reduction plan proposed by China during the 14 th Five Year Plan period.Adopting electric heating in universities can meet the energy consumption needs of various areas on campus,and can achieve independent regulation to ensure the comfort of students and faculty.However,the fluctuation of electricity load is significant,and smart grids have considerable flexibility in energy production and distribution.There are many factors that affect the electricity heating load,making it difficult to accurately predict the daily electricity load data of universities.In addition,it is difficult for people to make judgments based on experience comparing the trend of single factors on the electric heating load for the factors that affect it.In order to solve the above problems,this paper proposes a random forest short-term electric heating load data prediction model based on isolated forest multivariate stepwise feature screening and an interpretable evaluation prediction model algorithm based on nonlinear multi-dimensional to predict and analyze the short-term electric heating load of colleges and universities.The main research content of this article is as follows:(1)In view of the fact that the data collected by the electric heating sensor may be abnormal and the multi-dimensional feature data will increase the learning cost of the model and reduce the generalization ability of the model,this paper proposes an isolated forest multivariate stepwise feature filtering algorithm to detect the outlier of the data and reduce the feature dimension.Then,on the premise of ensuring the prediction accuracy of the model,the random forest regression algorithm is fused and the prediction results are output.Finally,the effectiveness of the model was demonstrated through comparative experiments.(2)In view of the lack of feature analysis in the research of electric heating load data prediction and the impact of categorical variable data on short-term electric heating load trend,this paper,based on the algorithm of the Temporary Fusion Transformer model,combines the One Lot encoder and the SM Taylor Softmax function to realize the optimization of VSN variable selection network in the model.Then,the Hyperband algorithm was applied to optimize the parameters of the model,and a nonlinear multi-dimensional interpretable evaluation prediction model was proposed.Further optimization was carried out in terms of prediction accuracy,and the interpretability of the model in predicting electric heating loads was enhanced.Provide theoretical support for the planning of electric heating energy in the power system and universities.(3)This paper conducts experimental verification on two real university electric heating load datasets.The experimental results show the practical value of the random forest short-term electric heating load data prediction model proposed in this paper,which is based on the isolated forest multivariate stepwise feature screening.The experimental results demonstrate the effectiveness of the nonlinear multi-dimensional interpretable evaluation prediction model.