Abstract:

With the development of new power systems, the capacity configuration of energy storage systems and power demand forecasting face high uncertainty and complex coupling relationships. To address this challenge, this paper proposes a machine learning-based collaborative modeling strategy, theoretically constructing a three-stage capacity planning framework of "estimation-confidence-feedback." By integrating deep learning and ensemble regression methods, a multi-scale, multi-quantile load forecasting system is established. Through rolling optimization and adaptive updating mechanisms, dynamic linkage between capacity and demand, as well as system self-learning, are achieved. This method emphasizes data-driven and closed-loop strategies, improving forecasting accuracy, capacity adaptability, and model robustness, providing theoretical support for the intelligent configuration of energy storage systems in dynamic environments.

Key words: energy storage systems, capacity planning, demand forecasting, machine learning, data-driven modeling