The Coordinated Control of Circulating Fluidized Bed Boiler. with Intelligence Feedforward Control

Transcription

1 The Coordinated Control of Circulating Fluidized Bed Boiler with Intelligence Feedforward Control Xiao-Feng Li, Shi-He Chen, Qing Zhong * Guangdong Power Test and Research Institute. No. 8, Shuijun Gang, Dongfengdong Road, Guangzhou.516, P. R. China. leexfeng3@yahoo.com.cn Abstract: An coordinated control strategy has been proposed and successfully applied to 3MW circulating fluidized bed (CFB) units in China. The paper describes the new coordinated control (NCC) system to overcome the long settling time, strong coupling, nonlinearity and inconstancy encountered in CFB combustion. NCC is based on both fuzzy Feedforward (fuzzy-ff) control and the fuzzy-pid feedback control. The fuzzy-ff control path contains a set of multi-input single-output fuzzy inference systems obtained from steady-state input output plant data and is used to improve the dynamics of CFB unit. The control output is mainly determined by the fuzzy-ff path, diminishing the control effort on the PID controllers. The fuzzy-pid controller supplies the complementary control signal component for regulation and disturbance rejection in small neighborhoods of the commanded trajectories. A self-adapting algorithm and non-uniform grid scheduling is proposed in the fuzzy-pid controller. The strategy is implemented through the function code on many typical DCS (EDPF-NT, XDPS, etc). The industrial application results show that this strategy achieves better performance in the specific range of load variations. Keywords: Fuzzy Feedforward Control, Fuzzy-PID Controller, Non-uniform Grid Scheduling, Circulating Fluidized Bed Boiler, Boiler-Turbine Coordinated Control 1. INTRODUCTION Circulating Fluidized Bed (CFB) boiler has special advantages in its adaptability to fuel,low in-furnace desulphurization cost, low pollutant emissions, high combustion efficiency, and comprehensive ash utilization. At present, China has been the largest country in the total capacity of installed CFB boiler units in the world. Commercial power plants in China are typically designed with Primary Frequency Control and Automatic Generation Control (AGC), which sets tight demands for the unit control. The grid frequency control requires quick response of unit power output in presence of frequency fluctuation, whereas the CFB boiler steam pressure change is typically a few minutes or 2 minutes later than the fuel rate change. The lagged CFB steam generation results in the problem of load control or AGC control in CFB unit. Hence, a new coordinated-control (CC) scheme should be introduced to replace the slower response of CFB by the faster turbo-generator response for load tracking under automatic generation control conditions. This paper presents the New Coordinated Control (NCC) scheme for CFB boiler application. The scheme consists of the fuzzy-pid controller and a unit Intelligence Feedforward controller that effectively reduces the interaction among different control loops throughout AGC control. The boiler fuzzy-pid controller with self-adapting algorithm and non-uniform grid scheduling is designed to improve the dynamics of boiler, and the turbine fuzzy-pid controller with self-adapting algorithm is designed to improve the dynamics of the turbine. The Fuzzy-FF control path in NCC scheme is founded on steady-state models using a set of multi-input single-output fuzzy inference systems. These non-linear models are established by predetermined relationships, which are primarily obtained from the large amount of process data and operating practice. The control scheme has been implemented in many typical DCS (EDPF-NT, XDPS, etc) (Li X. F. 27, 29) and applied to many 135 MW or 3 MW CFB Units in China for about two years. 2. NEW COORDINATED CONTROL SYSTEM It has been long recognized that the long residence time of a buring coal particle in CFB boiler creates a significant time lag to the firing rate in presence of load change. To eliminate or reduce the adverse Copyright by the International Federation of Automatic Control (IFAC) 724

2 impact, it is possible to use the technology that can automatically adjusts the proportional gain and the integration time of a PID controller,but the method is not always satisfactory. The use of feedforward control in combination with feedback control is known as an advanced control method. This is necessary to ensure fast load tracking performance of combustion system under varying load conditions, despite the long time delays before the effects of the steam pressure change can be observed. However, Fuzzy-FF control is not easy in practice because various adjustments must be made, such as manual operation,bias control,or the like. As is well known in the art, a signal for Fuzzy-FF control cannot assume all arbitrary values. (Fig. 1) Fig.1. The Fuzzy-FF Control Scheme of New Coordinated Control system 3. INTELLIGENCE FEEDFORWARD CONTROL ALGORITHM 3.1. Firing rate fuzzy anticipation controller Inputs to the firing rate fuzzy anticipation Feedforward controller should be determined first. The load error is a good indication of the error between power supply and power demand under varying load conditions. Hence, the load error and the derivative of load error are taken as the inputs to CFB firing rate fuzzy anticipation controller. It determines the action to be taken by the anticipation controller by considering the following factors: the load error MW E and the derivative of the load error MW E-D. The firing rate fuzzy anticipation controller is implemented as a two-input one-output Sugeno-type fuzzy system with constant consequent inference rules. The universes of discourse are determined after several simulation experiments, as well as membership functions for input Fuzzification (Fig. 2) OUT Firing Rate Anticipation Controller MW_D RATE MW_D Fig.2. Firing Rate Anticipation Controller Output Surface 725

3 The inference rule is expressed by: If MW E is and MW E-D is, then FFB-ANTI = 3.2. Firing rate load demand fuzzy Feedforward controller Several design methods may be used to design the fuzzy inference systems in the Feedforward control processor from input output data. The method known as the table look-up table is used here. Steady-state data along the whole operating range are used to design the demand Feedforward Controller. The inference rule is expressed by: If MW D is and MW D-R is, then FFB-ULD-B = Fuzzy Feedforward from load demand can be adjusted by the fuel flow rate to improve pressure control during load changes. The load change rate demand takes into account the differential coefficient to promote the pressure dynamic response during full load changes Valve servo load demand fuzzy Feedforward controller Input selection for the valve servo fuzzy FF controller is made the same way as that for the firing rate fuzzy FF controller. In this fuzzy FF case inference rules are established from the relationship between load demand and DEH control servo output, and the final control servo output can be fixed with fuzzified load demand and derivative of load demand. The inference rule is expressed by: If MW D is and MW D-R is, then FFB-ULD-T = 3.4. Abnormal Situation Fuzzy controllor (ASFC) The fuel quality and load changes in the bed temperature control should be taken into consideration in AGC controls of a CFB unit to stabilize the combustion and reduce emissions by balancing the amounts of the air and fuel. Bed temperature must be maintained at a high enough level to ensure an adequate combustion rate and low enough to avoid the sintering or the melting of the bed. If the amount of fuel is incorrect, the fuel input rate is trimmed based on the bed temperature. So, if there is a step change in fuel quality, or if the bed temperature or the throttle pressure goes outside the normal operating range during rapid load change conditions, the Abnormal Situation Fuzzy controller (AFSC) may be useful for correcting the process back to normal operations as soon as possible. It provides an action to deal with the abnormality by integrated consideration of the bed temperature T-b and the throttle pressure P-t. The rule is: IF T-b over T-b_Limit is OR P-t over P-t_Limit is THEN OUT_ ASFC is P-T Abnormal Situation Fuzzy Controllor % T-B OUT % Fig.3. Abnormal Situation Fuzzy Controller Output Surface Under abnormal operating conditions ASFC takes into action. When the operating point moves from the inside to the boundary of an abnormal operation region, the output of ASFC decreases gradually close to zero, and the ASFC forces the NCC system to slow down the rate of the changing load dement. This means that the load changes from the fully allowable to the completely forbidden state. When the operating point is close to a constraint, the output of ASFC is close to zero, and the changing load dement schedule halts. 4. FUZZY PID WITH NON-UNIFORM GRID SCHEDULING Fuzzy self-adjusting PID control schemes with uniform grid scheduling have enjoyed considerable success in industry (Li X. F 27, 29). However, the method is less effective for boiler control in CFB units since the combustion dynamics exhibits a severe non-linearity. In this section, a non-uniformed approach based on the fuzzification and inference mechanism of the T-S-type fuzzy system is presented to solve the non-linearity problem encountered in boiler control. Thus, a controller scheduling procedure based on current reference to the process is more desirable. The core of the approach is a MISO T-S-type fuzzy system with inference rules as shown below IF η is A n, THEN u n = K pn + K in e+ K de dn 726

4 Where K pn, K in and K dn are the linguistic values of PID parameters: proportional, integral and derivative gains, respectively, and the scheduling variable η, to be selected according to the control application, represents a regions, and the linguistic term A r is a fuzzy set that represents r region along the space of the scheduling variable η. For the Boiler Master Controller, the actual steam flow is chosen as the scheduling variable. It should be noted that the load reference is readily available; it is a slowly-varying variable whose time dependency is easily handled; and it is well correlated to the changes in the process dynamics, thus making it the perfect candidate to be the scheduling variable η. Therefore, we are able to build the fuzzy rule in the following form. Thus, the fuzzy rule can be expressed as follows: IF current S_flow is S_PV n THEN u n = K pn + K pn e T in + K pn T de dn (4) Where K pn, T in and T dn are the linguistic values of PID parameters; S_PV n is the linguistic value of actual steam flow The particular examination that was used in Table 3 is given in Table 4. The universes of discourse of S_PV n are {6%, 9%}; the universes of discourse of K pn are {1.23, 1.45}, the universes of discourse of T Dn are {4.3, 23.8}, and the universes of discourse of T In are {14.8, 557.5}. Table 1. Tuning Value of boiler master Controller under Difference load Master A m =2.5;Φ m =3 o A m =2.7;Φ m =45 o A m =3;Φ m =6 o Flow K Pm T Im T Dm K Pm T Im T Dm K Pm T Im T Dm 9% % INTELLIGENCE FEEDFORWARD CONTROL IN CFB UNIT CONTROL The performance of the developed NCC control system in comparison with conventional CCS is summarized here. To test the performance of the NCC control system, it is helpful to find a simple model that can capture the essential dynamics, especially the coupling effect between the generated electricity and the throttle pressure. In Tian et al. (24) a nonlinear unit model is proposed. The model is built for subcritical units (throttle pressure between 15.7 MPa and 19.6 MPa, main steam temperature between 535 C and 565 C) with coal-fired, naturally-circulated drum boilers. The nonlinear model takes the following form: ddq τs K f = DQ + e B (5) ddb CB = K3PTμ + K1D (6) Q dn KT = N + K3PTμ (7) P = P K ( K D ) (8) 1.5 T D 2 1 Q Table 2. Nomenclature Parameters Description B Boiler firing rate (t/h) μ Throttle valve position (%) N P T P D C B K 1 K 2 K 3 K t K f τ Output of Turbine Master output (MW) Throttle pressure (MPa) Drum pressure (MPa) Boiler storage constant Constant related to boiler firing rate and megawatt output Superheater friction drop coefficient Constant related to throttle valve Time constant of The turbine (s) Time constant of the mill (s) Pure time delay of the mill (s) Output of Boiler Master Fig.4. Main Parameters Under Conventional CCS The model contains 5 variables (B,, P T, N, and P D ) and 7 parameters (K 1, K 2, K 3, K f, C B, and K t ). All the symbols are described in Table 1. The conventional CCS is a master slave control strategy in most power plants. In comparison, the conventional CCS control system has been simulated as shown in Fig. 4. It shows transient process with key variables when unit load increases from 18 MW to 29MW at a rate of 9MW/min. When there is 727

5 load demand increase, turbine control valve opens immediately to produce the required additional power output. The immediate action of turbine control valve causes load increase and inlet steam pressure decrease at first, and then inlet steam pressure restores to the preset value by the compensation of coal delivery. The recovery of inlet steam pressure leads to excessive steam flow to the turbine, and turbine control valve has to close to track the error of load. This action causes oscillations in boiler variables and undesirable load overshoot. Output of fuzzy-ff Output of Boiler Master Output of Turbine Master Fig.5. Main Parameters Under NCC Strategy It has been long recognized that the long residence time of a buring coal particle in CFB boiler creates a significant time lag to the firing rate in presence of load change. In CFB combustion, it usually takes about 15 to 3 minutes to observe the steam pressure change after the fuel rate is changed at the given boiler demand, while the heat stored in the CFB bed can be released immediately after the air flow rate is changed at the given boiler demand, and the CFB energy release can continue with the succeeding increase of adequate firing rate. The special CFB behavior is different from that of the pulverized coal combustion in that the inverted response effect does not occur. For this reason, solutions to this problem have been proposed and utilized, such as kickers in the air control and the feeder speed. The main purpose of kickers is to provide an overshoot of air flow for CFB boiler to immediately release the boiler energy at the sudden load change, and the consequent overshoot of fuel rate contributes to sustain the increased boiler energy release and to compensate the fuel rate lag in combustion control. The load control dynamics is improved by the advantage of kickers in design. Boiler Master Demand Fuel Flow Turbine Master Demand Fig.6. Main Parameters of Conventional CCS under Fixed Pressure Operation Mode at a Loading Rate of 4.5MW/min In NCC system, the boiler control loop functions as a fast-acting firing rate controller, since changing the fuel demand and releasing the stored energy in CFB bed meets short-term increases in load demand. Fig.5 shows the transient process with key process variables when load increases from 16Mw to 29MW at a rate of 9MW/min. The changing boiler master demand leads to a quick increase in load change, as energy stored in the boiler is being released. The steam pressure is restored to its original level by increasing the boiler master output, after being decreased. The above tests indicate that the NCC scheme shows a better behavior during rapid load change conditions. The overshoot and settling time of load control decreases significantly by utilizing this control scheme. Neither the undesired pressure/load fluctuations nor the longer settling time is observed. The CFB boiler-turbine process includes fairly strong nonlinearities due to the different stored energy at each plant load. Figure 6 shows the results of operation under the conventional CCS plus non-uniform grid scheduling PID controller (F-PID) when the unit load is changed over a wide range at the rate of 4.5 MW/min from 28 MW to 17 MW in Baolihua 2 3 MW CFB Power Plant in 28. Field test is done under constant pressure. The data shows that it is difficult to avoid a significant time lag to the firing rate, and the main steam over-pressure forces the pressure control valve to open to release wasted energy. The NCC strategy described in this paper is realized by using the 728

6 Distributed Control System (DCS), which has been successfully applied in the No.5 and No.6 unit in the Yunfu (2 3 MW CFB) Power Plant. H-limit Fuel Flow Turbine Master Output L-limit Fig.7. Main Process Variables of NCCS under Sliding Pressure Operation Mode at a loading rate of 6MW/min Sliding pressure mode is an instructive development in power plant coordinated contro1. The unit shows little nonlinearity with the throttle pressure setpoint varied accordingly when the unit load is changed from 17MW to 3MW. Figure 7 shows the results of operation under sliding pressure operation mode, wherein the load is changed from 3 MW to 18 MW at the rate of 6MW/min. The throttle pressure is varied according to the selected sliding pressure curve. The megawatt output follows the load demand swiftly and accurately, and other parameters are steady. The figure also shows how the system responds to a rapid load change by a relatively large increase in the boiler fuzzy-ff controllers output. As is seen from this figure, fuzzy-ff controller output is a key contributor in load control at the initial stage of load change, and F-PID controller output begins to participate after the contribution of fuzzy-ff is decreased significantly. The combined action from fuzzy-ff and F-PID controllers helps to reduce the risk of oscillation and large overshoot, together with the long time lag in combustion controls. From the field tests, it can be seen that the designed NCC system has advantages in fast load tracking performance, and it applies to both sliding pressure mode and fixed pressure mode with larger turndown range (6% to 1% of the maximum load).. 6. CONCLUSIONS The proposed NCC has been running in many 135 MW and 3 MW commercial CFB units. It can be seen from the results of this industrial application that this control system is superior to conventional CCS because of its enhanced adaptability and robustness to the process. Excellent control results have been obtained, control accuracy related to turbine throttle pressure and active megawatt have been me and the demands of Automatic Generation Control are satisfied,when the unit is running at 12MW at the 1oading rate of 6MW/min. Site operation results confirm the advantages and feasibility of the proposed NCC system in CFB unit controls. 7. REFERENCES Astrom, K.J. and R.D. Bell (1993). A nonlinear model for steam generation processes. Proceedings IFAC 12th Triennial World Congress. Sydney, Australia Halow, J. S. (1995): Dynamical system analysis of fluidized beds. Proceedings of 3rd SIAM Conference on Applications of Dynamical Systems, Philadelphia, PA, USA, 1995 Li X. F. etc. (27). Application of The Fuzzy-PID to The Power Plant. NAFIPS '7. Annual Meeting of the North American. 27 June San Diego, CA, USA pp Tan, W. etc. (1999) Analysis and control of a nonlinear boiler-turbine unit. Journal of Process Control, 15: ,