Research On Data-Driven Control Strategy Of Large-Capacity Circulating Fluidized Bed Units

As a kind of highly effective and clean coal combustion technology, circulating fluidized bed boiler (CFBB) has shown its extraordinarily promising prospect in the power generation field. The tendency of large-scale and commercialization makes the design and further research of combustion control system much more significant. The combustion of CFBB, with characteristics such as nonlinearity, time-variant parameter, great inertia, large delay and coupling, is quite a complex process whose automatic control is the focus of current researches. Data-driven control theory, which is the current research hotspot, can design the controller merely by I/O data of the system, instead of the model precise information. Aiming at solving the control problem of CFBB combustion, this paper utilizes two-degrees-of-freedom data-driven PID control strategy.Based on the existing researches of the CFBB structure and technological process, the thesis makes conclusions of CFBB mathematical model of each part and CFB controlled variable, which providing fundamental basis for application of the two-degrees-of-freedom data-driven PID in CFBB combustion process.According to the I/O data of the controlled object, the data-driven PID can automatically generate and updatte the PID parameters in real time by adopting the data-driven algorithm. Configuration, theory, design and parameter tuning of the data-driven PID controller are discussed in detail. Data-driven PID and the two-degrees-of-freedom PID, by which both the tracking performance and the disturbance rejection can be considered, are combined to achieve the automatic online tuning of all parameters. Simulations on the nonlinear system show its effectiveness.Taking the coupling problem into consideration, the inversed decoupling method is utilized, after which the general system can be equivalently seen as two single loops. Then, the two-degrees-of-freedom data-driven PID controllers are applied to each loop. Tracking performance, disturbance rejection, operation adaptability and robust performance are taken into consideration during the simulations. Results have shown that the proposed method has a simple structure and is easy to engineering implementation, also has better adaptability and control quality when compared with traditional PID.