Data-driven distributed voltage regulation in electrical distribution systems

Abstract

Voltage regulation in modern electrical distribution systems faces significant challenges due to the widespread integration of distributed energy resources (DERs) such as solar photovoltaics and wind turbines. These resources introduce significant variability and uncertainty into power generation, complicating the maintenance of voltage stability. This paper introduces a novel Gaussian Process (GP)-based Model Predictive Control (MPC) framework for distributed voltage regulation. Utilizing Gaussian Process Regression (GPR), the proposed method models the complex, nonlinear power flow relationships to deal with the uncertainties of the system. The distributed framework divides the grid into small regions, each managed by a local controller, which coordinate using the Alternating Direction Method of Multipliers (ADMM) to ensure voltage regulation. This approach combines the scalability of ADMM with the accuracy of GPR, requiring fewer data points than traditional machine learning methods. Extensive simulations on various distribution networks demonstrate the effectiveness of the proposed method, outperforming centralized and other data-driven approaches in scalability and computational efficiency.