The Model Predictive Controller (MPC) is gaining importance as a primary controller in a hierarchical control of the standalone micro-grid. The reasons being its ability to perform online optimization, better handling of the system constraints, and the inherent nature to consider the predicted behavior of the system while taking the control actions. The existing MPC designs at the primary control level of the standalone micro-grid control are decentralized in nature with small prediction horizons. They are intended for inverter micro-grids with linear models. However, the decentralized MPC with small prediction horizons leads to slow response and oscillatory behavior in nonlinear micro-grids with both inverter-DG’s and conventional synchronous-DG’s.

This research work focused on the MPC formulations with large prediction horizons for centralized primary control of standalone micro-grids with nonlinear models. For a micro-grid with the nonlinear model, a centralized MPC design with a large prediction horizon leads to a non-convex nonlinear optimal control problem at each sample, which is challenging to solve online. In this work, centralized MPC designs based on linear time-invariant (LTI) and linear time-variant (LTV) approximations of the nonlinear micro-grid model are used, which approximates the non-convex optimal control problem into a quadratic problem which is easy to solve online. To improve the computational viability of the MPC, the work used the orthonormal special functions called Laguerre functions and Kautz functions in the MPC formulation. These functions decrease the number of optimal variables in the control problem without compromising the performance. The work also covers the other auxiliary requirements for the implementation of online MPC. The performance of the proposed MPC formulations was analyzed in an eight-bus standalone micro-grid with one synchronous-DG and one photovoltaic-DG for the load disturbance and source intermittency scenarios.