A hybrid data-driven and physics-based method for industrial load modeling and aggregated flexibility evaluation

• A hybrid data-driven and physics-based method for modeling the regulatory characteristics of industrial loads. • Estimation of key parameters in physics-based load modeling using production process-level power data. • Tailored methods for quantifying aggregate flexibility for energy markets and incentive-based demand response programs. • Wasserstein distributionally robust chance-constrained optimization for addressing uncontrollable load uncertainty. Industrial loads, characterized by large energy consumption and strong flexibility, represent high-quality adjustable resources. However, current research faces two major challenges: first, the complex internal coupling relationships of industrial loads and data privacy constraints make it difficult to construct accurate regulation models; second, the aggregated flexibility evaluation of industrial loads with uncertainties proves highly challenging. To address these issues, this paper proposes a hybrid data-driven and physics-based method for industrial load modeling and aggregated flexibility evaluation. Initially, a data-driven industrial load regulation model is constructed using historical data. Subsequently, a physics-based model of industrial loads is developed according to production processes, with parameters estimated using production process-level historical data. By integrating data-driven and physics-based models, a more refined industrial load regulation model is obtained. Finally, for two typical application scenarios of industrial load-aggregated virtual power plants, flexibility region evaluation method and regulation capability metric evaluation method are proposed respectively. Simulation results demonstrate that the proposed method can effectively evaluate the flexible adjustment capabilities of industrial load-aggregated virtual power plants while establishing more precise regulation models.