Model Based Control of the Circulating Fluidized Bed Boiler Jiří Findejs*, Vladimír Havlena* , **, Jan Jech*, Daniel Pachner* * Honeywell Laboratory, V Parku 2326/18, 148 00 Praha 4, Czech republic (e-mail: {Vladimir.Havlena|Daniel.Pachner}@honeywell.com) ** Faculty of Electrical Engineering, Czech Technical University of Prague Abstract: The paper describes Advanced Combustion Control (ACC) system for the circulating fluidized bed (CFB) boilers. ACC is based on the model-based predictive control technology. The CFB boiler dynamics is represented by a non-linear low order model which captures the key CFB boiler behavior. The model exhibits strong cross-interactions between the process variables, therefore the control based on single input single output PID loops is difficult; however, it is shown the CFB boiler can be successfully controlled using the multiple-inputs multiple-outputs (MIMO) model. The bed temperature, boiler power and the oxygen concentration in the flue gas can be controlled simultaneously and independently via the primary and secondary air flows and the fuel supply rate. Keywords: Circulating fluidized bed, control oriented modeling, combustion control, predictive control. 1. INTRODUCTION Fluidized bed combustors burn solid fuels in a turbulent mixture of gases, fuel particles and inert particles suspended in a strong stream of primary air. They evolved from the effort to achieve combustion temperatures below the threshold of nitrogen oxides forming, which is approximately at 1400ºC (Howard 1983). In fluidized bed combustor, the flue gases can be in effective contact with sulphur absorbents such as calcium dioxide produced by limestone calcination in the bed. This desulphurization mechanism can capture more than 95% of the SO x formed. As a result, the fluidized bed combustion technology can be successfully used with low quality coals, turfs, and biomass (Basu and Frazer, 1991) producing low pollutants concentration. Quoting Bittanti et al., (2000), “Fluidized beds are remarkably difficult to model, since the process is characterized by a series of complex thermal and mechanical interactions”. In spite of these facts, this paper presents a successful fluidized bed boiler control application based on the predictive control technology using a nonlinear model. The doctoral thesis by Karpanen (2000) provides an insight to a multi-fuel CFB control. The work suggests the fuzzy logic and the neural networks approaches to overcome the difficulties. It seems the PID control system performance degrades if the fuel quality is changed. The model based control presented in this paper contains parameters which can directly be interpreted as the fuel properties (mainly its heating value H). The model can adapt even to substantial changes of the heating value. The performance of adaptation has been verified in practice applying the approach presented to a boiler burning the coal/coke mixture of a variable ratio. From this perspective, our approach is a model-based alternative to the fuzzy logic and neural networks systems for the multi-fuel bio-waste firing boilers. The circulating fluidized bed (CFB) combustor shown in Figure 1 represent the most sophisticated fluidized bed based combustion technology. The retention of ashes in the cyclone and its reinjection to the bed lead to lower loss of unburned fuel particles and more efficient limestone utilization. Except of improved efficiency, the principles as well as the main characteristics are similar to those of the other fluidized bed boiler types. Thus, the principles described should not be limited to the CFB boiler only. The CFB boiler dynamics contains strong cross interactions of variables. Hence the MIMO model is necessary for a successful control. This paper is focused on a “lumped” CFB boiler model and the predictive control strategy based on utilization of the accumulated char in the bed. The concept of a ‘mathematical model of the CFB boiler’ exhibits extreme variability, either in its form, or complexity. The researchers in fluidization engineering focused on CFB boiler design consider usually partial differential equations based finite element method models describing the key variables like temperature, concentrations, pressure and velocity as time-varying three-dimensional fields defined in the combustion chamber (Bittanti et al., 2000). On the other hand, there exists the usual transfer matrix based approach prevalent in control engineering. The transfer functions between the key manipulated and controlled variables can be readily obtained through plant step testing without deep understanding the physical laws governing them. The concept of a ‘mathematical model of the CFB boiler’ exhibits extreme variability, either in its form, or complexity. The researchers in fluidization engineering focused on CFB