Convolutional Neural Network-based RoCoF-Constrained Unit Commitment (2309.02650v1)

Abstract: The fast growth of inverter-based resources such as wind plants and solar farms will largely replace and reduce conventional synchronous generators in the future renewable energy-dominated power grid. Such transition will make the system operation and control much more complicated; and one key challenge is the low inertia issue that has been widely recognized. However, locational post-contingency rate of change of frequency (RoCoF) requirements to accommodate significant inertia reduction has not been fully investigated in the literature. This paper presents a convolutional neural network (CNN) based RoCoF-constrained unit commitment (CNN-RCUC) model to guarantee RoCoF stability following the worst generator outage event while ensuring operational efficiency. A generic CNN based predictor is first trained to track the highest locational RoCoF based on a high-fidelity simulation dataset. The RoCoF predictor is then formulated as MILP constraints into the unit commitment model. Case studies are carried out on the IEEE 24-bus system, and simulation results obtained with PSS/E indicate that the proposed method can ensure locational post-contingency RoCoF stability without conservativeness.