Conservative Sparse Neural Network Embedded Frequency-Constrained Unit Commitment With Distributed Energy Resources

The increasing penetration of distributed energy resources (DERs) will decrease the rotational inertia power system and further degrade frequency stability. To address above issues, this paper leverages advanced neural network (NN) to learn dynamics incorporates NN facilitate reliable operation. This proposes conservative sparse (CSNN) embedded frequency-constrained unit commitment (FCUC) with converter-based DERs, including learning optimization stages. In stage, it samples parameters, calculates corresponding frequency, characterizes stability region sampled parameters using convex hulls ensure avoid extrapolation. For conservativeness, positive prediction error penalty is added loss function prevent possible requirement violation. sparsity, topology pruning employed eliminate unnecessary connections for solving acceleration. trained CSNN transformed into mixed-integer linear constraints big-M method then incorporated establish data-enhanced model. case study verifies 1) effectiveness proposed model in terms high accuracy, fewer significant acceleration; 2) stable operation against violation under contingency.