Machine learning modeling and predictive control of nonlinear boiler-turbine unit

The boiler-turbine unit (BTU) is extensively utilizeed in power generation owing to its high reliability and operation flexibility. However, the control design of BTU system remains challenging due to its intricate dynamic properties, such as wide range of nonlinearities, large inertia, and strong coupling in multivariable systems. More challengingly, it is generally difficult and time-consuming to derive an accurate nonlinear model to capture the BTU dynamics and facilitate controller design. In this regard, machine learning methods show great potential in classical engineering fields, such as modeling nonlinear dynamic systems. In this article, a machine learning based model predictive control (MPC) scheme is presented for the nonlinear BTU. The machine learning method is first adopted to model the BTU using the input and output data directly, which can also be used to model a large class of nonlinear dynamic systems. Then, this data-driven model is deployed as a predictive model in the MPC for optimal control of the nonlinear BTU. Finally, a series of simulations are performed to validate the proposed method, including RNN training and testing, open-loop prediction simulation and closed-loop control simulation.