Prediction Of Fresh Air Load Of Large Public Buildings Based On Neural Network

For large public buildings,fresh air load accounts for a high proportion of air conditioning energy consumption.Predicting fresh air load is an effective way to reduce building energy consumption and develop green buildings.This project is aimed at three different types of large public buildings: shopping malls,hotels,and offices.Fresh air load for building load influence degree is analyzed.Analytic hierarchy process is used to analyze the main factors affecting the change of fresh air load,the fresh air load is predicted based on neural network,the main research contents are as follows:DeST software is used to simulate and analyze the building load of commercial building,hotel building and office building respectively.The result of simulation analysis of Jia Zhaoye Mall in Shenyang City,the total load in summer is 36% of the fresh air load,the total winter load is 83.5% of the fresh air load.The results of the Shenyang friendship hotel simulation is shown,the total load in summer is 35% of the fresh air load,the total winter load is 62.3% of the fresh air load.The simulation results of Fushun office building are analyzed,the total load in summer is 38% of the fresh air load,the total winter load is 68.7% of the fresh air load.In summary,fresh air load has a great influence on the load of large public buildings.so it is a great significance to reduce the energy consumption of the building load of the fresh air load.The analytic hierarchy process is used to analyze the weight of the factors affecting the fresh air load,Through the characteristics of fresh air load,related information and expert questionnaire method,the influencing factors of fresh air load are determined,including outdoor temperature,outdoor humidity,unit model,number of indoor personnel,heat disturbance,etc.According to the calculation results,The most influential factor for fresh air load is the number of people,reaching 27.65%.Secondly,outdoor temperature and hu midity accounted for 12.74%.Therefore,these three factors were selected as the established forecast model input parameters will largely improve the model prediction accuracy.The applicability of artificial neural network is analyzed,the advantages and disadvantages of various neural network algorithms is compared.According to the dynamic changes of fresh air load of large public buildings.Input layer parameters,hidden layer parameters,values of various weight thresholds are determined.The prediction model of Elman neural network is established,the actual data is fed into the model for prediction,The prediction results show,the most of the absolute errors are between-0.5 and 0.5,and most of the relative errors are lower than 7.00%.The time error of the prediction result in the large change of personnel flow is the largest.The predicted results are consistent with those of AHP and are feasible.However,a single prediction model is prone to slow convergence and fall into limited solutions,the accuracy of the results is affected.Elman neural network is easy to be targeted by trapped solutions,Genetic algorithm is used to optimize the original model.Through continuous iteration of model weights and thresholds,the value with the highest fitness is taken into the prediction model.The prediction results show,the absolute error can be controlled within 0.5,the relative error is controlled within 5%.By comparing with the original prediction model.The optimized prediction model of Elman neural network is more accurate,and also meet the characteristics of fresh air load generation,the model is feasible.