GNN assisted frequency constrained unit commitment of multi-region power systems with high penetration of renewable energy sources

Abstract

The increasing integration of renewable energy resources (RES) into the power grid poses significant challenges in system frequency dynamics. Traditional frequency-constrained unit commitment models simplify the average system frequency and neglect the spatial characteristics of frequency dynamics, potentially underestimating the risk of contingencies. In this paper, we consider a nodal frequency response model to capture the frequency dynamics in the unit commitment problem. Nodal frequency dynamics and rate of change of frequency (RoCoF) expressions are converted into security constraints against worst contingency, which are then incorporated into the proposed muti-region frequency-constrained unit commitment (MR-FCUC) formulations. To improve the efficiency and performance of the MR-FCUC model, a decomposition algorithm is implemented to solve the proposed MR-FCUC efficiently. The subproblem of the original model confirms the frequency dynamics, and sensitivity cuts are refined based on the validation errors. Additionally, a GNN-based voltage phase angle predictor is incorporated to boost the computational efficiency of the FCUC model. Case study involving modified IEEE 24-bus and IEEE 118-bus systems illustrates the effectiveness of the proposed GNN-MR-FCUC model. Simulation results of test systems affirm that the frequency stability is guaranteed: the maximal RoCoF is mitigated within 0.5 Hz/s, and the lowest frequency nadir is maintained above 59.71 Hz.