Symbolic Regression for Cooling Load Forecasting: Addressing Data Efficiency and Cold Start Challenges

Abstract: Traditional machine learning and deep learning models require significant training data, limiting their applicability in new or data-scarce environments. To overcome this, a symbolic regression-based approach is presented to develop a data-efficient and interpretable forecasting model for cooling loads. Equipment and weather data collected from a large hospital is used for training and validation, demonstrating the model’s effectiveness in minimizing data dependency and addressing cold start challenges. The proposed framework outperforms traditional black-box models in terms of interpretability, data efficiency, and adaptability, making it suitable for real-time deployment and broader energy management applications.