C. Vázquez. Random Batch Method (RBM) and Model Predictive Control (MPC) for Converter-Dominated Power Systems (2024)
Abstract. This paper investigates the application of Data-based Reduced Order Models, the Random Batch Method (RBM) and Model Predictive Control (MPC) in Converter-Dominated Power Systems. The study evaluates their effectiveness in handling stiff systems, marking a pioneering effort in this domain. Several challenges emerge, particularly concerning the accuracy of approximations due to the stiffness of the state matrix A: block-wise decomposition strategies face limitations in accurately approximating the original, as evidenced by simulation results. This limitation questions the effectiveness of block-wise strategies, highlighting the need for alternative decomposition methods. Spectral decomposition showed potential but prompts a search for more efficient alternatives. Also, we show that combining MPC with RBM is a promising approach for optimal control problems constrained by state trajectories, as the latter efficiently handles large time horizons T by breaking down the time into smaller sub-intervals, and although a marginal improvement of the RBM-MPC over RBM was found, we have concluded that does not justify the increased computation time, emphasizing the need to understand factors influencing error accumulation. Furthermore, the study identifies a relationship between the RBM, the number of discretization points, and the switching parameter K, where increasing these parameters enhances accuracy but may lead to deterioration in quality. These results are in line with the theoretical framework. Overall, the study highlights the necessity for continued investigation to optimize control strategies for stiff systems such as power systems.