1、英文翻译部分英文部分：Advanced control algorithms for steam temperature regulation of thermal power plants A.Sanchez-Lopez,G.Arroyo-Figueroa*,A.Villavicencio-RamirezInstituto de Investigaciones Electricas, Division de Sistemas de Control, Reforma No. 113, Colonia Palmira,Cuernavaca, Morelos 62490, MexicoReceiv
2、ed 5 February 2003; revised 6 April 2004; accepted 8 July 2004AbstractA model-based controller (Dynamic Matrix Control) and an intelligent controller (Fuzzy Logic Control) have been designed and implemented for steam temperature regulation of a 300 MW thermal power plant. The temperature regulation
3、is considered the most demanded control loop in the steam generation process. Both proposed controllers Dynamic Matrix Controller (DMC) and Fuzzy Logic Controller (FLC) were applied to regulate superheated and reheated steam temperature. The results show that the FLC controller has a better performa
4、nce than advanced model-based controller, such as DMC or a conventional PID controller. The main benefits are the reduction of the overshoot and the tighter regulation of the steam temperatures. FLC controllers can achieve good result for complex nonlinear processes with dynamic variation or with lo
5、ng delay times.Keywords: Thermal power plants; Power plant control; Steam temperature regulation; Predictive control; Fuzzy logic control1. IntroductionCurrent economic and environment factors put a stringer requirement on thermal power plants to be operated at a high level of efficiency and safety
6、at minimum cost. In addition, there are an increment of the age of thermal plants that affected the reliability and performance of the plants. These factors have increased the complexity of power control systems operations 1,2.Currently, the computer and information technology have been extensively
7、used in thermal plant process operation and control. Distributed control systems (DCS) and management information systems (MIS) have been playing an important role to show the plant status. The main function of DCS is to handle normal disturbances and maintain key process parameters in pre-specified
8、 local optimal levels. Despite their great success, DCS have little function for abnormal and non-routine operation because the classical proportional-integral-derivative (PID) controlis widely used by the DCS. PID controllers exhibit poor performance when applied to process containing unknown non-l
9、inearity and time delays. The complexity of these problems and the difficulties in implementing conventional controllers to eliminate variations in PID tuning motivate the use of other kind of controllers, such as model-based controllers and intelligent controllers. This paper proposes a model-based
10、 controller such as Dynamic Matrix Controller (DMC) and an intelligent controller based on fuzzy logic as an alternative control strategy applied to regulate the steam temperature of the thermal power plant. The temperature regulation is considered the most demanded control loop in the steam generat
11、ion process. The steam temperature deviation must be kept within a tight variation rank in order to assure safe operation, improve efficiency and increase the life span of the equipment. Moreover, there are many mutual interactions between steam temperature control loops that have been considered. O
12、ther important factor is the time delay. It is well know that the time delay makes the temperature loops hard to tune. The complexity of these problems and difficulties to implement PID conventional controllers motivate to research the use of model predictive controllerssuch as the DMC or intelligen
13、t control techniques such as the Fuzzy Logic Controller (FLC) as a solution for controlling systems in which time delays, and non-linear behavior need to be addressed 3,4. The paper is organized as follows. A brief description of the DMC is presented in Section 2. The FLC design is described in Sect
14、ion 3. Section 4 presents the implementation of both controllers DMC and FLC to regulate the superheated and reheated steam temperature of a thermal power plant. The performance of the FLC controller was evaluated against two other controllers, the conventional PID controller and the predictive DMC
15、controller. Results are presented in Section 5. Finally, the main set of conclusions according to the analysis and results derived from the performance of controllers is presented in Section 6.2. Dynamic matrix controlThe DMC is a kind of model-based predictive control (Fig. 1). This controller was
16、developed to improve control of oil refinement processes 5. The DMC and other predictive control techniques such as the Generalized Predictive Control 6 or Smith predictor 6 algorithms are based on past and present information of controlled and manipulated variables to predict the future state of th
17、e process.The DMC is based on a time domain model. This model is utilized to predict the future behavior of the process in a defined time horizon (Fig. 2). Based on this precept the control algorithm provides a way to define the process behavior in the time, predicting the controlled variables traje
18、ctory in function of previous control actions and current values of the process 7. Controlled behavior can be obtained calculating the suitable future control actions. To obtain the process model, the system is perturbed with an unitary step signal as an input disturbance (Fig. 3).This method is the
19、 most common and easy mean to obtain the dynamic matrix coefficients of the process. The control technique includes the followings procedures:(a) Obtaining the Dynamic Matrix model of the process. In this stage, a step signal is applied to the input of the process. The measurements obtained with thi
20、s activity represent the process behavior as well as the coefficients of the process state in time. This step is performed just once before the operation of the control algorithm in the process.(b) Determination of deviations in controlled variables. In this step, the deviation between the controlle
21、d variables of the process and their respective set points is measured.(c) Projection of future states of the process. The future behavior of each controlled variable is defined ina vector. This vector is based on previous control actions and current values of the process.(d) Calculation of control
22、movements. Control movements are obtained using the future vector of error and the dynamic matrix of the process. The equation developed to obtain the control movements is shown below:where A represents the dynamic matrix, AT the transpose matrix of A X the vector of future states of the process, f
23、a weighting factor, I the image matrix and D he future control actions. Further details about this equation are found in Ref. 5.(e) Control movements implementation. In this step the first element of the control movements vector is applied to manipulated variables. A DMC controller allows designers
24、the use of time domain information to create a process model. The mathematical method for prediction matches the predicted behavior and the actual behavior of the process to predict the next state of the process. However, the process model is not continuously updated because this involves recalculat
25、ions that can lead to an overload of processors and performance degradation.Discrepancies in the real behavior of the process and the predicted state are considered only in the current calculation of control movements. Thus, the controller is adjusted continuously based on deviations of the predicte
26、d and real behavior while the model remains static.3. Fuzzy logic controlFuzzy control is used when the process follows some general operating characteristic and a detailed process understanding is unknown or process model become overly complex. The capability to qualitatively capture the attributes
27、 of a control system based on observable phenomena and the capability to model the nonlinearities for the process are the main features of fuzzy control. The ability of Fuzzy Logic to capture system dynamics qualitatively and execute this qualitative schema in a real time situation is an attractive
28、feature for temperature control systems 8. The essential part of the FLC is a set of linguistic control rules related to the dual concepts of fuzzy implication and the compositional rule of inference 9.Essentially, the fuzzy controller provides an algorithm that can convert the linguistic control st
29、rategy, based on expert knowledge, into an automatic control strategy. In general, the basic configuration of a fuzzy controller has five main modules as it is shown in Fig. 4.In the first module, a quantization module converts to discrete values and normalizes the universe of discourse ofvarious ma
30、nipulated variables (Input). Then, a numerical fuzzy converter maps crisp data to fuzzy numbers characterized by a fuzzy set and a linguistic label (Fuzzification). In the next module, the inference engine applies the compositional rule of inference to the rule base in order to derive fuzzy values o
31、f the control signal from the input facts of the controller. Finally, a symbolic-numerical interface known as defuzzification module provides a numerical value of the control signal or increment in the control action. This is integrated by a fuzzy-numerical converter and a dequantization module (out
32、put).Thus the necessary steps to build a fuzzy control system are Refs. 10,11: (a) input and output variables representation in linguistic terms within a discourse universe;(b) definition of membership functions that will convert the process input variables to fuzzy sets; (c) knowledge base configur
33、ation; (d) design of the inference unit that will relate input data to fuzzy rules of the knowledge base; and(e) design of the module that will convert the fuzzy control actions into physical control actions.4. ImplementationThe control of the steam temperature is performed by two methods. One of th
34、em is to spray water in the steam flow,mainly before the super-heater (Fig. 5). The sprayed water must be strictly regulated in order to avoid the steam temperature to exceed the design temperature range of G1% G5 8C). This guaranties the correct operation of the process, improvement of the efficien
35、cy and extension of the lifetime of the equipment. The excess of sprayed water in the process can result in degradation of the turbine. The water in liquid phase impacts on the turbines blades. The other process to control the steam temperature is to change the burner slope in the furnace, mainly in
36、 the reheated. The main objective ofthis manipulation is to keep constant the steam temperature when a change in load is made. The DMC, fuzzy logic and PID controllers were implemented in a full model simulator to control the superheated and reheated steam temperature. The simulator simulates sequen
37、tially the main process and control systems of a 300 MW fossil power plant. The simulator has the full models of each main element of the generation unit. These models let the simulator display the effects of a disturbance in each process variable.4.1. Dynamic matrix control (DMC)The matrix model of
38、 the process is the main component of the DMC. In this case the matrix model was obtained by a step signal in both the sprayed water flow and the burnersposition.Fig. 6 shows a block diagram of the DMC implementation in the steam superheating and reheating sections. The temperature deviations were u
39、sed as the controllers input. The sprayed water flow and slope of burners were used as the manipulated variables or controllers output. The DMC performance was implemented using a prediction horizon of 10 s, a weighting factor in the last control actions of 1.2 and considering the last 30 movements
40、executed. These parameters belong to the best available for this application in the study of the DMC performance 7.4.2. Fuzzy logic control (FLC)Seven fuzzy sets were chosen to define the states of the controlled and manipulated variables. The triangular membership functions and their linguistic rep
41、resentation are shown in Fig. 7. The fuzzy sets abbreviators belong to: NBZnegative big, NMZnegative medium, NSZnegative short, ZEZzero, PSZpositive short, PMZpositive medium and PBZpositive big.The design of the rule base in a fuzzy system is a very important part and a complex activity for control
42、 systems. Li et al. 11 proposed a methodology to develop the set of rules for a fuzzy controller based on a general model of a process rather than a subjective practical experience of human experts. The methodology includes analyzing the general dynamic behavior of a process, which can be classified
43、 as stable or unstable. In Fig. 7 the range of fuzzy sets are normalized to regulate the temperature within the 20% above or below the set point, the change of error within the G10%, and the control action are considered to be moved from completely close or 08 inclination to completely open or 908 o
44、f inclination in water flow valve and slope of burners, respectively. In the case of regulation of temperature, if therequirements change to regulate the temperature within a greater range, the methodology proposed by Li et al. 11 considers to apply a scale factor in the fuzzy sets. A time step resp
45、onse of a process can be classified as stable or unstable, as shown in Fig. 8. Characteristics of the four responses are contained in the response shown by the second stable response. The approach also uses an error state space representation to show the inclusion of the four responses in the second
46、 stableone (Fig. 9).A set of general rules can be built by using the general step response of a process (second order stable system):1. If the magnitude of the error and the speed of change is zero, then it is not necessary to apply any control action (keep the value of the manipulated variable).2.
47、If the magnitude of the error is close to zero in a satisfactory speed, then it is not necessary to apply any control action (keep the value of the manipulated variable).3. If the magnitude of the error is not close to the system equilibrium point (origin of the phase plane diagram) then the value o
48、f the manipulated variable is modified in function of the sign and magnitude of the error and speed of change. The fuzzy control rules were obtained observing the transitions in the temperature deviations and their change rates considering a general step response of a process instead of the response
49、 of the actual process to be controlled. The magnitude of the control action depends on the characteristics of the actual process to be controlled and it is decided during the construction of the fuzzy rules. A coarse variable (few labels or fuzzy regions) produces a large output or control action, while a fine variable produces small one.Fig. 10 shows a representative step response
陕公网安备 61072602000132号 违法和不良信息举报：0916-4228922