AUTHOR=Lin Zhanbin TITLE=Short-Term Prediction of Building Sub-Item Energy Consumption Based on the CEEMDAN-BiLSTM Method JOURNAL=Frontiers in Energy Research VOLUME=Volume 10 - 2022 YEAR=2022 URL=https://www.frontiersin.org/journals/energy-research/articles/10.3389/fenrg.2022.908544 DOI=10.3389/fenrg.2022.908544 ISSN=2296-598X ABSTRACT=In order to improve the accuracy of short-term prediction of building itemized energy consumption, a short-term prediction model of building itemized energy consumption based on CEEMDAN-BiLSTM method is proposed. The methodology includes selecting the lighting, air conditioning and power consumption data of a building in 2019 as sample then classifying the weather according to meteorological indicators, decomposing the energy consumption data into multiple components by CEEMDAN method, selecting strong correlation components and then sending them to BiLSTM network followed by superimposing the prediction results of each sub-component to get the final energy consumption prediction results. And finally building BP, SVM, LSTM, EMD-LSTM and CEEMDAN-LSTM models synchronously to compare the errors with the proposed models. The results show that weather factors have a great influence on the accuracy of building energy consumption prediction. When the weather fluctuates greatly, the prediction error of energy consumption of single model is larger. The energy consumption data is decomposed by CEEMDAN, and the detailed features can be fully extracted. Compared with CEEMDAN-LSTM model, CEEMDAN-BiLSTM model reduces eRMSE, eMAPE and eTIC by 4.1%, 9.441% and 1.3% respectively. The proposed model can effectively improve the accuracy of short-term prediction of building itemized energy consumption.