Data-Driven Linear-Time-Variant MPC Method for Voltage and Power Regulation in Active Distribution Networks

This paper proposes a data-driven voltage and power regulation method based on model predictive control (MPC) for active distribution networks (ADNs), which doesn’t rely on accurate system model parameters. To approximate the system model from data with a good balance between simplicity and expressiveness, we propose a new piecewise linear (PWL) regression method combing the support vector linear regression (SVR) model, which overcomes the issues of measurement outliers and data collinearity in the traditional PWL regression method. With the PWL model trained offline, a novel data-driven voltage and power regulation method based on the linear time-varying MPC scheme is proposed in this paper, which can effectively coordinate fast and slow regulating resources with a low computational burden. The convergence of the proposed data-driven MPC based scheme is also theoretically clarified. The effectiveness of the proposed method is demonstrated via numerical simulations using several distribution systems.