Multi-model based MPC to improve the ramp rate for renewable energy consumption

Accommodating mass renewable energy into the power grid requires coal-fired power plants to trace reference trajectory, which varies rapidly and is prescribed by the power dispatching system. The Coordinating Control System (CCS) is impotent because decoupling nets work well only on one linearized point of this nonlinear system. This paper presents a multi-observer prediction framework based on model predictive control (MPC) to address the tricky problem. In this framework, multiple state observers predict the output states within the predictive horizon. The arbitration mechanism chooses the right observer on the actual outputs of the system. The picked observer will calculate control inputs subject to energy constraints. Then, the issue is transformed into a quadratic programming formulation. Numerical computation based on gradient descent is used to obtain the optimal control inputs. The experimental results demonstrate that the outputs of coal-fired plants perform more flexibly under the proposed framework than CCS. This approach can improve the capability of equipment.