A Study on Improved Grey Wolf Optimizer-BP Neural Network for Building Energy Consumption Prediction

To address the issue of low accuracy in traditional building energy consumption prediction models, this study proposes an improved building energy consumption prediction model that combines an Enhanced Grey Wolf Optimization algorithm (IGWO) with a Back Propagation (BP) neural network. The WOA algorithm was improved by incorporating opposition-based learning, differential evolution algorithm, and introducing a control parameter λ, which effectively resolves the problems of premature convergence and local optima trapping while enhancing algorithm accuracy. The proposed method was validated using operational data from a commercial building complex in Xi'an, China. Empirical results demonstrate that compared to the traditional BP neural network, the IGWO-BP model reduced the Root Mean Square Percentage Error (RMSPE) by 14.1%. Evidently, this model exhibits strong generalization capability, superior prediction accuracy, and reliability, fully satisfying the requirements of practical engineering applications.