Data-driven Approach to Chance-constrained AC Optimal Power Flow with Optimization of AGC Participation Factors

The AGC participation factors’ optimization, coupled with the renewable energy uncertainties, changes the distribution of status variables. Consequently, it poses both modeling and computational challenges for existing chance-constrained alternating current optimal power flow (CC-ACOPF) methods. This paper proposes a data-driven method to reformulate and solve the CC-ACOPF, optimizing participation factors for automatic generation control (AGC) without prior assumptions about uncertainties. Based on historical data, a surrogate model is constructed to describe the relationship between uncertainties and chance constraints. A linearized power flow model, including reactive power and voltage magnitude, is introduced to reduce the dimensionality of the surrogate model. By embedding the surrogate model into the chance-constrained optimization, a tractable data-driven CC-ACOPF model is developed. A two-stage, data-driven Monte Carlo-based method is established to correct errors in power flow linearization and surrogate learning. Numerical results are presented for the PJM 5-bus and IEEE 118-bus systems to validate the effectiveness and robustness of the proposed method.