A hybrid PSO-LSTM-based electricity prediction and optimization technique for home appliances

With population growth and technological advancements, electricity demand in residential buildings has increased sharply. Accurate energy consumption forecasting allows building owners and operators to understand and predict the energy usage patterns of their buildings. However, the prevailing forecasting techniques have certain limitations that must be addressed for improved energy optimization. This paper proposes a three-stage energy optimization technique for individual home appliances. The first stage performs season-wise cluster analysis using a hierarchical clustering algorithm. Secondly, an electricity prediction approach has been designed and implemented using an adaptive long short-term memory (LSTM) model. The third stage deploys particle swarm optimization (PSO) to determine optimal hyperparameters. The proposed hybrid technique has been rigorously evaluated using a benchmark energy consumption dataset of individual home appliances. The comparative analysis with pevious state-of-the-art prediction models reveals the superiority of the proposed work. This technique can significantly contribute to achieving sustainable and optimal energy management in residential buildings.