Probabilistic Forecasting-Based Stochastic Nonlinear Model Predictive Control for Power Systems With Intermittent Renewables and Energy Storage
Kiet Tuan Hoang, Christian Ankerstjerne Thilker, Brage Rugstad Knudsen, Lars Imsland
Abstract
Managing hybrid power systems with significant intermittent power production is challenging. To address this, a probabilistic forecasting-based stochastic nonlinear model predictive control (SNMPC) scheme is proposed where data-driven Lamperti-transformed stochastic differential equations (SDEs) are employed as nonlinear grey-box models for the intermittent renewable source. This allows the control scheme to consider forecasting renewable power production that follows non-Gaussian distributions. In more detail, integrating Lamperti-transformed SDEs in the SNMPC framework enables the method to 1) propagate and forecast the non-Gaussian uncontrollable renewable power output mean and uncertainty based on past data, future uncertain numerical weather predictions and current observations and 2) formulate tight probabilistic constraints based on said mean and uncertainty for satisfying some exogenous power demand. The method is demonstrated in simulation on a nonlinear offshore hybrid power system (OHPS) case study, consisting of controllable gas turbines, uncontrollable intermittent offshore wind production, and electric batteries with wind speed and power data from the real operation of a wind farm in Denmark.