Multi-contingency transient stability-constrained optimal power flow using multilayer feedforward neural networks

Abstract

Transient stability-constrained optimal power flow (TSC-OPF) aims at optimising the scheduling of generation with stability constraints to ensure a secure system in the event of contingencies. This paper proposes a new approach based on a critical clearing time (CCT) constraint that replaces the dynamic and transient stability constraints of the TSC-OPF problem. The CCT is computed by a multilayer feedforward neural network (MFNN) trained using Gauss-Newton approximation for Bayesian regularization. In order to ensure a uniform distribution of generated points in the input space to train the neural networks, a Sobol quasi-random sequence is adopted for data generation. The proposed method has the merit of removing the computational burden of dynamic simulation during optimisation. Multi-contingency can simply be handled by adding a CCT constraint for each contingency. Simulation results for the New England 10-machine 39-bus system show that TSC-OPF using MFNN has very fast convergence to optimal operating points with the desired CCT.