With the implementation of energy-saving and emission reduction policies, intelligent lighting systems in buildings are facing higher energy-saving requirements. However, general particle optimization algorithms have some limitations when dealing with multi-objective optimization problems. Therefore, this study improves the particle swarm optimization algorithm using penalty functions and optimizes the vector machine model to form a particle swarm-support vector machine model for energy consumption optimization control of building intelligent lighting systems. In comparison with the baseline model, the improved model reduced the average lighting energy consumption by about 5.7%, and the actual lighting power fluctuated between 4150W and 4400W. After 24 training iterations, the prediction accuracy of the model reached 90%. In addition, the importance of factors such as building orientation and lighting intensity decreased from 0.07 and 0.06 to 0.06 and 0.05, respectively, demonstrating a more balanced impact on the lighting system. Therefore, the model not only maintains comfortable lighting conditions while effectively optimizing the lighting energy use of buildings, but also improves energy efficiency.
