Comparative Study Of Energy Consumption Prediction Methods For Office Buildings In Cold Regions

The increasingly prominent environmental and resource problems have made people realize the importance of achieving the goal of "carbon peaking and carbon neutrality",and the construction industry occupies a relatively high percentage of the carbon emissions in the country,occupying 50.9% of the overall domestic consumption,so energy-saving design and renovation of buildings are important for achieving the double carbon goal and reduce energy consumption.Therefore,energy-efficient construction and retrofitting are essential to meeting the dual-carbon target and enhancing energy efficiency.Energy consumption prediction in buildings is an important part of energy-efficient design and renovation,and it is necessary to choose a suitable energy consumption prediction method in the scheme design stage.This report takes office buildings in cold regions as an example,determines key design parameters affecting building energy consumption based on statistical principles,creates an office building power prediction model for cold regions by studying three typical energy demand estimation methods,introduces prediction model evaluation indexes,and compares and analyzes different algorithmic prediction models to determine the best prediction method for office buildings in cold regions in China.The main research contents are summarized as follows:(1)The initial model of office building energy consumption in cold regions was constructed by De ST-c software,and parameters such as outdoor meteorological parameters,indoor thermal disturbance number,and thermal performance of envelope structure were determined based on relevant standard codes and research on reference buildings.Through orthogonal tests,numerical adjustments were made in the range of values of energy consumption influencing factors to provide data support for subsequent analysis of the significance of building energy consumption influencing factors and input parameters of the prediction model.The significance of the factors influencing the two energy consumption indexes of building energy consumption and air conditioning energy consumption were analyzed separately by the ANOVA method to screen out the key energy consumption influencing factors and establish the input parameters of the energy utilization projection module.(2)To further improve the accuracy of the energy consumption prediction model,parametric simulation(Je Plus)was used to obtain multiple sets of building energy consumption data to provide data support for the prediction model establishment.Based on the significance analysis and parametric simulation results,BP(back propagation)artificial neural network,LSTM(long short-term memory)neural network and GA-BP(genetic algorithm error back propagation)neural network prediction models were established,and the prediction models established by the three algorithms were proved to be able to predict the energy consumption of office buildings in cold regions by linear fitting and relative error distribution between the predicted and simulated values.(3)The model evaluation indexes RMSE(root mean square error)and MAPE(mean absolute percentage error)are introduced to evaluate the three prediction models,and the results show that the prediction accuracy of the models constructed by the three algorithms based on BP,LSTM and GA-BP are all above 95%,among which the prediction accuracy of the model constructed by the GA-BP algorithm is the highest,with the maximum value of RMSE being 0.359 and the value of MAPE not The maximum value of RMSE is 0.359 and the value of MAPE is not more than 0.2885%,and the model evaluation indexes are all smaller than the other two algorithms.Therefore,for the prediction of energy consumption in office buildings in cold regions,the use of GA-BP neural nets should be given priority.