AI-Driven Affinely Adjustable Robust Many-Objective Scheduling Framework for Flexibility-Oriented Power Systems with High Renewable Penetration

The increasing penetration of renewable energy sources (RESs) has fundamentally transformed modern power system operation, introducing significant challenges associated with net-load volatility, rapid ramping events, and flexibility supply–demand imbalance. Traditional deterministic scheduling approaches lack adaptability to renewable uncertainty, while conventional unit commitment (UC)-based flexibility assessment methods, although accurate, are computationally intensive and unsuitable for large-scale scenario-rich environments. To overcome these limitations, this paper proposes a novel AI-driven, flexibility-oriented Affinely Adjustable Robust Optimization (AARO) framework integrated with Generalized Linear Polyhedron (GLP)-based uncertainty modeling and a hybrid MOABC–NSGA-III many-objective optimization strategy. The proposed architecture uniquely combines probabilistic residual load forecasting, spatiotemporal renewable correlation modeling, explicit ramp-based flexibility quantification, and robust many-objective scheduling within a unified and computationally efficient framework. Extensive simulation studies conducted on modified IEEE 30-, 57-, and 118-bus systems demonstrate clear superiority over state-of-the-art deterministic and UC-based methods. At 60% renewable penetration, the flexibility gap is reduced from 10.3% to 2.8%, representing a 72.8% improvement in flexibility adequacy. Renewable curtailment decreases by 58%, while the Flexibility Risk Index (FRI) is reduced by approximately 78%, indicating significantly enhanced reliability under uncertainty. Additionally, total operational cost and carbon emissions are reduced by 10.7% and 13.8%, respectively. From a computational perspective, the proposed framework achieves more than 80% reduction in execution time compared to classical UC-based flexibility studies. These results confirm that the proposed method effectively reduces conservatism, enhances robustness, improves Pareto optimality, and provides a scalable, technically rigorous solution for next-generation renewable-dominated power systems.