Recent Articles and Multimedia
Paper Title: Gaussian Process Supported Stochastic MPC for Distribution Grids
Authors: Moritz Wenzel, Edoardo De Din, Marcel Zimmer, and Andrea Benigni
The control of distribution grids faces the challenge of increasing uncertainties related to the rising share of distributed renewable generation. At the same time, including probabilistic information into the solution of the system-determining power flow equations further complicates an already computationally complex task. This paper proposes an algorithm to solve these two interwoven problems by approximating non-linear parts of the power flow equations with a Gaussian Process Regression (GPR), which enables the efficient application of a stochastic tube algorithm to deal with forecast and measurement uncertainties in the Model Predictive Control (MPC).
READ MORE ABOUT THIS PAPER HERE
The sequence of the proposed algorithm is mostly executed offline, starting with the distribution grid model and uncertainty model, and training the GPR first, followed by the computation of the stochastic tubes. Hence, only the relatively computationally lightweight online MPC is executed in real-time.