Performance forecasting and optimization for metro station chilled water systems under extreme heat based on field data analysis

Rapid urbanization and expanding rail networks drive surging energy demands, necessitating efficiency research. When extreme heatwaves coincide with peak passenger periods, metro station air conditioning systems suffer from severe load fluctuations and performance degradation. Current research on chiller efficiency focuses primarily on standard operating conditions, often overlooking dynamic performance under extreme heat. Furthermore, a lack of high-temperature datasets restricts the adaptability and accuracy of existing predictive models. In response, this study proposes an integrated framework encompassing performance prediction, high-temperature performance degradation warning and operational optimization. To precisely predict COP, energy consumption (E), and degradation rate (η), a multi-output prediction model called N-BEATS-XGBoost is created by integrating gradient boosted trees with deep temporal feature extraction. By combining temporal feature extraction from N-BEATS with gradient boosting from XGBoost, the model achieves high predictive accuracy, with R2 values of 98.5 %, 99.1 %, and 99.1 %, and SMAPE values of 1.56 %, 1.93 %, and 1.47 % for COP, E, and η, respectively. The increasing frequency and intensity of high-temperature events necessitate optimized system operation under such conditions. Criteria distinguishing functional failure from performance degradation were established, and an IQR-based method identified a degradation-rate warning threshold of 39.8 %, effectively demarcating normal operation from potential degradation. An NSGA-II-based dual-objective optimization framework was applied to historical high-temperature data, targeting energy consumption (E) and degradation rate (η). The results indicate reductions of 3.64 % in E and 10.06 % in η, demonstrating substantial potential for performance improvement under extreme heat. These findings address a critical challenge at the nexus of climate resilience and energy efficiency.