Short-term building cooling load prediction model based on DwdAdam-ILSTM algorithm: A case study of a commercial building

Abstract

Traditional prediction methods of building cooling load fail to consider load changes caused by the time factor, which results in the failure of the best predicting effect and portability. To explore a more accurate cooling load predicting approach, in this paper, a new hybrid model based on decoupled weight decay Adam optimized improved LSTM neural network, denoted as DwdAdam-ILSTM, is proposed. First, the cross-correlation function is used to extract features with a high degree as the input data. Then, the decoupling weight decay adaptive moment estimation (DwdAdam) optimization algorithm is utilized to optimize the parameters of the improved long short term memory (ILSTM) prediction model. Next, the ILSTM is employed to establish a prediction model. Finally, to verify the validity and prediction accuracy of the proposed model, this paper selects the hourly cooling load data of a large commercial building in Xi'an as the research sample. Results of the experiment show that the Mean Absolute Error (MAE), Mean Absolute Percentage Error (MAPE), Mean Square Error (MSE), Variation Coefficient of Root Mean Square Error (CV-RMSE) and Range of Relative Errors (E-R) of DwdAdam-ILSTM predicting the cooling load are 2.2%, 2.74, 5.2%, 0.061, 0.216. Compared with three version variant LSTM neural networks, the MSE and E-R values of this model are decreased by 74%-53%, 42%-49%, and 69%-64%, which has better performance in accuracy and stability. It testifies that the proposed strategy is reliable and effective. Additionally, this model can provide important guidance for daily energy management and multi-energy scheduling of buildings, which is helpful to the application of cooling load prediction.