A hybrid forecasting model for general hospital electricity consumption based on mixed signal decomposition

Abstract

Current research into electricity consumption forecasting for General Hospital still has considerable scope for further development, particularly in its failure to incorporate hospital-specific energy usage characteristics as input variables. This study explores the impact of the usage frequency of sizeable medical equipment on the electricity demand of general hospitals. It proposes a hybrid forecasting algorithm that integrates the Improved Complete Ensemble Empirical Mode Decomposition with Adaptive Noise (ICEEMDAN) and Variational Mode Decomposition (VMD) for signal decomposition with the Hyperband-LSTM deep learning algorithm to enhance prediction accuracy. ICEEMDAN is employed for preprocessing the power consumption series, while VMD is used for the secondary decomposition of high-frequency signals within the series. The Hyperband Pruner is utilized to efficiently adjust the hyperparameters of the LSTM, which is then used for electricity consumption forecasting. The predictive performance of the developed method is assessed by comparing it with 15 different forecasting models. The results indicate that the proposed method demonstrates superior forecasting performance. Applying the model to a real-case scenario, it has reduced the hospital’s electricity consumption by about 15%, providing a referable energy management solution for other medical institutions.