Life extending control of boiler–turbine systems via model predictive methods

Abstract

Boiler–turbine units constitute a critical component of a co-generation system. During operation, especially the start-up operation, these units are subject to high-temperature variations that aggravate the stress of the material used in their construction and thus a negative effect in their life spans. This paper is about designing a life extending controller (LEC) to obtain a good tradeoff between the life of a boiler–turbine system and control system performance. Because the boiler system is multivariate and there exist some constraints on the inputs to plant, model predictive control theory is used. For practical consideration, the original controller of the boiler system designed for dynamic performance is taken as a pre-compensator. For ease in LEC design, the pre-compensated closed-loop system is reduced in order using the system identification method. Finally, the resulting system is extensively simulated and tested on a computer, using a sophisticated nonlinear model of the boiler system.

Introduction

Boiler–turbine systems are used to generate electric power in Syncrude Canada Ltd. (SCL) integrated energy facility which is located in Mildred Lake, Alberta. In SCL, the utility plant consists of a boiler system, a header system and an electricity generating system. The setup of the overall utility plant is shown in Fig. 1.

The boiler system consists of three utility boilers, two carbon monoxide (CO) boilers and two once-through steam generators (OTSG). The 900# (6.2 MPa) header collects steam from the boiler system and then distributes it for three different usages: (i) to extract bitumen from oil sands; (ii) to numerous turbines to generate electricity; (iii) to three other headers to generate steam at different pressure levels (6.20, 4.14, 1.03 and 0.34 MPa). The overall plant, like many similar ones available worldwide, is thus a rather complex, nonlinear, interconnected system.

Undoubtedly, boiler–turbine units are a critical component in this system. These units are susceptible to damage due to extensive use in a high-temperature environment, during normal operation these units are subject to high-temperature about $500{}^{\circ }\mathrm{C}$. It should be noted that because the highest temperature and pressure variations happen during the start-up operation, especially the cold start-up (thick-walled components like turbine rotors and steam headers of boilers are critical), the highest service lifetime consumption occurs. How to reduce their damage and extend their life spans is of great interest to engineers. In this paper, we design an additional controller in order to get a significant improvement in service life of a turbine at the expense of a small reduction in the system performance. This secondary controller will be refereed to as a life extending controller (LEC).

LEC is a new area of research that integrates system science and mechanics of materials. Because of its practical importance, it has attracted a great deal of attention from industry and academia (Noll et al., 1991; Kallappa, Holmes, & Ray, 1997; Holmes & Ray, 1998). Because fatigue damage is cycle-dependent, fatigue damage models are usually based on stress–strain hysteresis loops. In contrast, most control theories, e.g., the model predictive control (MPC), are formulated in the time domain. To fit the control framework, the damage dynamics should be expressed as a vector differential equation with respect to time. Ray, Wu, Carpino, and Lorenzo (1994) proposed a continuum fatigue/damage modeling method for use in LEC. The advantages of this approach are that it allows the damage model to be incorporated within the optimization problem and that the accumulated damage between any two instants of time can be derived even if the stress–strain hysteresis loop is not closed. However, because the so-called rainflow cycle counting is used in this method, it is not quite suitable for real-time control (Bannantine, Comer, & Handrock, 1990). Li, Chen, Marquez, and Gooden (2003) introduced an improved rainflow counting method to overcome this drawback. Also, Lorenzo (1994) proposed an alternative to this approach. Compared with Ray's method, his theory is simpler and thus more suitable for control system design. For this reason, in this paper we will build one damage model of the boiler–turbine system by making use of Lorenzo's method.

The operating principle of the utility boiler can be found in Tan, Marquez, Chen, and Gooden (2001). In brief, the boiler system is responsible for the steam production, whose quantity (measured by its flow rate) and quality (measured by its pressure and temperature) can be controlled by four inputs: the feedwater flow, fuel flow, attemperator spray flow and air flow. Like a traditional controller design, difficulties to design an LEC for a boiler system are mainly due to the following two facts: first, a boiler system includes multi-inputs and multi-outputs, and interactions exist among different inputs and outputs; second, there exist input or output constraints, such as physical limits of actuators, safety margins, limited manufacture tolerances.

In order to cope with multivariable and constrained control problems, in recent years MPC has been investigated and successfully tested. The main idea of MPC is to use a model of the plant to estimate system's evolution, and accordingly, select the command input. Prediction, thus is handled according to the so-called receding horizon scheme (Bemporad, 1998). Culter and Remaker (1980) proposed a form of predictive control algorithm, called dynamic matrix control (DMC). DMC went on to become one of the most well-known commercial MPC products. Since then, a few varieties of MPC were suggested by different researchers (Richalet, Rault, Testud, & Papon, 1978; Sripada & Fisher, 1985; Clarke, Mohtadi, & Tuffs, 1987; Shah, Mohtadi, & Clarke, 1993). DMC approach emphasizes optimal plant operation under constraints, and computed the control signal by repeatedly solving a quadratic programming problem, that fits our problem very well. Thus, DMC is used in this paper.

In many industrial facilities such as the boiler–turbine system at hand, control systems are already in operation for good dynamic performance. An important and practical question is: How to enhance the structural durability with existing control systems still in place? We use a hierarchical structure to handle this problem. In this structure, the LEC is at a higher level with the existing controller in place. In fact, we will show in Section 3.1 that the plant is ill-conditioned, scaling or pre-compensating is necessary before controller design; fortunately, the original controller can be used as a pre-compensator. After compensation, the system becomes very complex, being 19th order. For ease in LEC design, we reduce the system using an identification approach before LEC design.

This paper is organized as follows. In Section 2, we apply Lorenzo's damage modeling theory to construct a damage model for our problem. In Section 3, we concentrate on designing an LEC with the MPC theory. Finally, in Section 4, simulation results with a sophisticated nonlinear model are presented.