Building energy management: Integrating K-means with sparrow search algorithm optimized BiLSTM for anomaly detection and prediction

Currently, building energy conservation and emission reduction are crucial for environmental protection and sustainable development. In response to the difficulty in accurately quantifying building energy consumption and the low efficiency of existing management, this study develops a building energy management system that integrates hybrid anomaly detection and intelligent prediction algorithms. The outcomes indicate that the highest determination coefficient of the prediction model reaches 0.96, and the mean absolute error of energy consumption prediction in spring is 3.33%, with a mean absolute percentage error of 1.01%. Compared with benchmark models such as sparrow search algorithm and particle swarm optimization algorithm, this method reduces the number of convergence iterations by 40.8%, further reduces the average absolute error by 14.2%, saves training time by 44.7%, and reduces memory usage by 18.5%. This method has significant advantages in building energy consumption prediction and anomaly detection, which can effectively improve the efficiency of building energy management and provide data support for energy efficiency enhancement and pollutant reduction. This provides new solutions for energy efficiency optimization of university buildings, promoting the realization of green buildings and energy efficiency enhancement and pollutant reduction.