To address the issue of non-energy-efficient behavior resulting from human operations during system runtime, this study utilizes cumulative operational data from the 2022 cooling season to establish a back propagation neural network energy consumption prediction model. The model is subsequently validated, with the average error meeting precision requirements. Leveraging the energy consumption prediction model, a genetic algorithm is employed for optimization, and the results are compared with those obtained through manual experiential adjustments. The findings indicate that, energy savings are achieved through parameter adjustments using the genetic algorithm outperform those based on human experience. Notably, within the load interval accounting for the longest operational duration (from 30% to 50%), energy savings reach 7.84%.