Model Free Adaptive Predictive Control of Desulfurization Slurry pH Based on CPS Framework

Jian Liu， Xiaoli Li,2,3,N， Kang Wang， Fuqiang Wang and Guimei Cui

（1. Faculty of Information Technology, Beijing University of Technology, Beijing 100124, China；2. Beijing Key Laboratory of Computational Intelligence and Intelligent System, Ministry of Education, Beijing 100124, China；3. Beijing Advanced Innovation Center for Future Internet Technology, Beijing University of Technology, Beijing 100124, China；4. Technology Research Center, Shenhua Guohua (Beijing) Electric Power Research Institute Corporation, Beijing 100025, China；5. School of Information Engineering, Inner Mongolia University of Science and Technology, Baotou 014010, China）

Abstract: In order to improve the slurry pH control accuracy of the absorption tower in the wet flue gas desulfurization process, a model free adaptive predictive control algorithm for the desulfur?ization slurry pH which is based on a cyber physical systems framework is proposed. First, aiming to address system characteristics of non?linearity and pure hysteresis in slurry pH change process, a model free adaptive predictive control algorithm based on compact form dynamic linearization is proposed by combining model free adaptive control algorithm with model predictive control al?gorithm. Then, by integrating information resources with the physical resources in the absorption tower slurry pH control process, an absorption tower slurry pH optimization control system based on cyber physical systems is constructed. It is turned out that the model free adaptive predictive control algorithm under the framework of the cyber physical systems can effectively realize the high?precision tracking control of the slurry pH of the absorption tower, and it has strong robust?ness.

Key words: wet flue gas desulfurization；slurry pH；cyber physical systems；model free adaptive predictive control；tracking control

In recent years, the frequent appearance of haze in China has had a serious impact on people's health and living environment. As the major emission source of air pollutants in China,thermal power plants have been highly valued by the Chinese government for environmental pro?tection. Sulfur dioxide is one of the main factors of acid rain and haze. The national emission standard of sulfur dioxide is getting higher and higher, so it is very important to control sulfur dioxide emission from coal?fired power plants.Limestone gypsum wet flue gas desulfurization(WFGD)[1?2]technology is widely used in thermal power plants because of its mature technology,high reliability and strong adaptability. The technology involves mixing sulfur dioxide in flue gas with limestone slurry in the desulfurization tower, which produces a good environmental pro?tection effect. From the processing perspective,the principle of limestone gypsum wet desulfuriz?ation technology is an acid?base neutralization re?action, where the pH control of the whole reac?tion process is very important. From the process composition perspective, the desulfurization tower is bulky, and the reaction process is com?plex, resulting in the slurry pH having character?istics of nonlinear, time?variant and large inertia.Therefore, developing solutions to accurately con?trol the slurry pH has become the focus of the technology, and it is also a problem that must be solved to improve the desulfurization efficiency.

At present, there are two main methods for coal?fired power plants to control the pH value of slurry: one is manual control, the other is PID control. Manual control requires the power plant staff to have significant working knowledge of the machinery and has high requirements for the control of the slurry valve opening. During the control process, it is easy to cause sudden in?creases and decreases of limestone slurry, which leads to large fluctuation of slurry pH in the ab?sorption tower. PID control improves the auto?mation level of slurry pH control to a certain ex?tent. However, the PID control parameters are difficult to adjust for the controlled object with nonlinear, time?varying and large inertia, which affects the control accuracy and describes its poor adaptive ability. In recent years, many ex?perts and scholars have tried to introduce ad?vanced algorithms into the pH control of ab?sorber slurry, and some results have been achieved. In Ref. [3] and Ref. [4], neural network and fuzzy control are combined with traditional PID control respectively to ensure the control ac?curacy of slurry pH in the absorption tower. In Ref. [5], the dynamic matrix control algorithm was applied to the slurry pH control system in the absorption tower, and good control effect is achieved. In Ref. [6], a control strategy based on dynamic superposition of slow coarse adjustment?fast fine adjustment?feedforward control is pro?posed for the pH control process of the absorp?tion tower slurry, so that the slurry pH of ab?sorption tower is kept in a stable range. In Ref. [7],a variable parameter adaptive Smith control sys?tem is proposed to adjust the pH value of the desulfurization slurry, which addresses the prob?lems of large inertia and lag in the pH control of desulfurization slurry. In Ref. [8], the desulfuriza?tion pH control system is designed based on a configuration predictive control algorithm, and the control effect is obviously better than the current mainstream cascade PID control method.

The slurry pH of the absorption tower is a complex controlled object with characteristics of non?linearity, large inertia, hysteresis, time?vary?ing, etc., so it is difficult to establish an accurate mathematical model. For unknown nonlinear sys?tems, in which establishing accurate mathematic?al models are difficult tasks, model free adaptive control (MFAC)[9?11]is an effective data?driven control method. By introducing the pseudo par?tial derivative (PPD) and using the input and output data of the controlled system, the equival?ent dynamic linear data model of the nonlinear system is established. It is not necessary to estab?lish the process model of the system in order to realize the adaptive control of the system. Due to the large volume and complex reaction process of the desulfurization tower, there is hysteresis evid?ent in the detection performed by the various sensors. The change in slurry pH value can?not be obtained in real time. Therefore, there is a large inertia and extensive lag time in the slurry pH change process of absorption tower. Given the characteristics of large inertia and long delay time in the slurry pH, model predictive control(MPC)[12?14], which is mainly composed of three links: model prediction, rolling optimization and feedback correction, is a good control method in this context. It can effectively deal with multi?variable, constrained systems with long lag prob?lems and has good performance in dynamic con?trol. In view of the nonlinearity and long lag time of the slurry pH change process, a model free ad?aptive predictive control algorithm based on compact form dynamic linearization (CFDL?MFAPC)[10]is proposed by combining the ad?vantages of model free adaptive control al?gorithm with the advantages of model predictive control algorithm in this paper. The combined control algorithm is used for slurry pH tracking control of absorption tower.

Cyber physical systems (CPS)[15?17]are com?plex systems composed of physical space and in?formation space. Physical space and information space map each other, interact timely and co?operate efficiently. By integrating advanced sens?ing, computing, communication, control, and oth?er information technologies and automatic con?trol technologies, cyber physical systems can be realized with on?demand response, rapid itera?tion and dynamic optimization of resource alloca?tion and operation in the system. CPS is a com?plex system with multi?disciplinary integrations.Different countries and institutions have defined CPS according to their current situation and re?search direction. He Jifeng, academician of the Chinese Academy of Sciences, defines CPS as a controllable, reliable and scalable network?based physical equipment system, which deeply integ?rates computing, communication and control cap?abilities on the basis of environmental perception.It achieves deep integration and real?time inter?action through the feedback loop of interaction between the computing process and physical pro?cess to add or expand new functions. It can mon?itor or control a physical entity in a safe, reliable,efficient and real?time manner. At present, cyber physical systems are widely used in road traffic,energy, aerospace, medical, robot, industrial automation, smart grid and other fields. The ba?sic structure of the information physical system is shown in Fig. 1. Due to the rapid development of network technology and the application of a large number of basic automation equipment, a large amount of data in the flue gas desulfurization process can be obtained in real time. This data contain a large amount of potential information reflecting the relationship between production op?eration rules and process parameters. In order to make good use of the production data and im?prove the automation level of flue gas desulfuriz?ation and the control accuracy of the slurry pH of the absorption tower, CPS technology is intro?duced into the slurry pH control process of the absorption tower in WFGD. Through CPS’s per?ception, computing, communication, precise con?trol, remote coordination and autonomy and oth?er functions, the traditional WFGD system in?dustrial network and communication network are interactively integrated to realize the two?way flow and effective utilization of information in the converged network. By constructing a closed?loop enabling system with the function of state perception, real?time analysis, scientific decision and precise execution, the intelligent optimal control of slurry pH of absorption tower based on CPS framework is realized.

Fig. 1 Cyber physical systems

1 Analysis of pH Control in WFGD Process

1.1 WFGD process

The limestone?gypsum WFGD system mainly includes the flue gas system, the absorp?tion tower and oxidation air system, desulfuriza?tion slurry preparation system, gypsum dehydra?tion system and wastewater treatment system,etc. The process flow of WFGD is shown in Fig. 2.The flue gas discharged from the boiler first passes through the deduster to remove dust and then enters the bottom of the absorption tower through the induced draft fan, booster fan and heat exchanger. The flue gas moves from bottom to top after entering the bottom of absorption tower. The limestone slurry enters the absorp?tion tower and is sent from the bottom to the spray layer via the circulation pump to be sprayed from the top to the bottom in the form of droplets. The limestone slurry and the flue gas are placed in full contact in order to produce a chemical reaction, and calcium sulfite is gener?ated after the reaction. The calcium sulfite is ox?idized from the air blown by the oxidation fan to generate calcium sulfate. The calcium sulfate is then crystallized to produce wet gypsum. After dehydration, the wet gypsum becomes solid gypsum for building materials. The desulfurized flue gas first passes through the demister at the top of the absorption tower to remove moisture,and then heats up through the heat exchanger and is discharged into the atmosphere through the flue and chimney. After desulfurization, the clean flue gas passes through the mist eliminator at the top of the absorption tower to remove moisture. Then, it is heated to 85 ℃ by the heat exchanger and discharged into the atmosphere through the flue and chimney.

Fig. 2 WFGD process flow

1.2 Hammerstein model of slurry pH control system

In the limestone?gypsum wet flue gas desul?furization process, the measurement and control of the slurry pH of absorption tower are the key factors that affect the desulfurization rate and the quality of the final product gypsum. The higher the pH value, the better the absorption of sulfur dioxide and the higher the desulfurization rate. However, if the slurry pH remains high for a long time, the quality of gypsum will be de?creased. The lower the pH value, the better the dissolution of limestone, but the absorption of sulfur dioxide is suppressed, and the desulfuriza?tion rate will decrease. In actual production, the slurry pH is usually maintained between 5.0 and 6.0. In the desulfurization process, the most im?portant factor affecting the slurry pH of the ab?sorption tower is the limestone slurry flow which is mainly controlled by the opening of the slurry valve. Therefore, the slurry pH control of the ab?sorption tower can be regarded as the opening controll of the limestone slurry valve. In this pa?per, the limestone slurry valve opening is taken as the control input and the limestone slurry pH value as the system output.

Hammerstein model[18]describing the slurry pH control system of the absorption tower is

2 Tracking Control of Slurry pH Based on CPS Architecture

2.1 CPS hardware structure design of slurry pH control system

In WFGD process, the slurry pH has com?plex characteristics of time?variant, long delay time, large inertia and nonlinear, which makes the conventional PID control unable to meet the control requirements. Many power generation en?terprises rely on operators to manually adjust the limestone slurry supply or slurry circulating pump frequency to reduce sulfur dioxide emis?sion, which not only wastes human resources, but also increases the production costs and reduces economic benefits. Through the cooperation and integration of calculation, network and optimiza?tion control methods, a CPS framework for slurry pH control process of the absorption tower is pro?posed to realize timely monitoring, reliable trans?mission, optimal control and comprehensive treatment of system information. The CPS framework is shown in Fig. 3. Through the dis?tributed data acquisition system, the data of each sensor in the production process is transmit?ted to the integrated controller through the field bus. Through the optimization control method,the control signal of slurry pH value control pro?cess is obtained. The control signal is transmit?ted to each actuator through the field bus for regulation and control. The two?way data trans?mission is carried out by industrial Ethernet, and the industrial touch screen displays the operat?ing status of the system and performs on?site monitoring and control of the production opera?tion. At the same time, the wireless transmission module and industrial cloud technology are ap?plied in the CPS design. Remote computers and mobile devices are used to monitor the produc?tion process. Using this method, the equipment performance can be fully exploited, and the pro?duction efficiency, product quality and economic benefits can be improved.

Fig. 3 CPS hardware structure of slurry pH control system of absorption tower

2.2 Design of CFDL-MAFC controller

It is assumed that the discrete nonlinear in?put?output system of the slurry pH control pro?cess is

From the practical point of view, the above assumption of control object is reasonable and acceptable. Assumption 1 is a typical constraint condition for general nonlinear systems in con?trol system design. Assumption 2 is a restriction on the upper bound of the change rate of system input and output. From the energy perspective,change in the bounded slurry valve opening can only cause change in the bounded slurry pH in the system.

Based on Eq. (4), the following N?step for?ward prediction equation can be obtained

If the control increment of two adjacent mo?ments is too large, the input of the system will jump sharply, and the burden of the actuator will increase. Therefore, the following control input criterion functions are considered

Eqs. (10)(11)(13) (15)–(19) constitute the CFDL?MFAPC control scheme of a slurry pH control system. The stability analysis of the al?gorithm is basically the same as that in Ref. [10],so we will not repeat it here.

2.3 Comprehensive optimization control scheme of slurry pH based on CPS

By integrating the physical resources, con?trol algorithm, industrial network and communic?ation system of the absorption tower slurry pH control process, an optimized control scheme for the slurry pH of absorption tower based on the CPS architecture is proposed in this paper, as shown in Fig. 4. The information relayed from the sulfur dioxide concentration sensor, temperat?ure sensor, pH sensor, pressure sensor and slurry density sensor is uploaded to the Ethernet in real time. Data storage and real?time calculation are carried out via Ethernet, and the data informa?tion is filtered and processed. The decision con?trol unit designs the model free adaptive control?ler according to the data information obtained.The controller parameters and expected slurry pH are set manually or automatically to realize the optimal control decision over the slurry pH of the absorption tower. The optimized control sig?nal is then transmitted to each actuator via Eth?ernet to complete the control task. In this way, a closed loop of perception?analysis?decision?execu?tion is formed to realize the intelligent optimal control of the pH control process in the absorp?tion tower slurry under the CPS framework. At the same time, the production control data on the Industrial Ethernet is passed to the decision detection unit in the form of reports and charts to complete the visualization of the production process and control decisions and to realize re?mote monitoring and control of the entire pro?cess.

Fig. 4 CPS structure of pH control process of absorption tower slurry

3 Simulation Research and Analysis

The Hammestein model parameters of the pH control system of the absorption tower slurry are

CFDL?MFAPC algorithm, CFDL?MFAC al?gorithm and PID algorithm are used for matlab simulation respectively. The simulation results of these three algorithms are analyzed and com?pared. The initial value of the system is set to

When the CFDL?MFAPC algorithm is used for simulation, the controller parameters are

The simulation results are shown in Fig. 5 and Fig. 6. Fig. 5 shows the tracking curve of the actual slurry pH value compared to the expected slurry pH value. It can be seen from the figure that the CFDL?MFAPC algorithm can suffi?ciently realize the tracking control of the system output slurry pH value to the expected pH value.In the whole tracking process, there is no over?shoot and high steady?state accuracy is achieved.Fig. 6 shows the error curve during the tracking process of the output slurry pH value compared to the expected pH value. It can be seen from the curve that the pH tracking error almost ap?proaches 0 except when the pH setting value changes suddenly, which indicates the effective?ness of CFDL?MFAPC algorithm for slurry pH control of the absorption tower.

Fig. 5 pH tracking curve with CFDL?MFAPC algorithm

Fig. 6 pH tracking error curve with CFDL?MFAPC algorithm

Fig. 7 pH tracking curve with CFDL?MFAC algorithm

Fig. 8 pH tracking error curve with CFDL?MFAC algorithm

Fig. 9 pH tracking curve with PID control algorithm

Fig. 10 pH tracking error curve with PID control algorithm

The simulation results of Fig. 5 to Fig. 10 are obtained without external disturbance.However, there is usually a certain degree of ex?ternal interference in the actual slurry pH con?trol process. The common factors that may cause external interference of the slurry pH control sys?tem of absorption tower include the degree of coal combustion in the boiler, the change in the sulfur content in the coal, and the change in the generator unit load, etc. Next, in the presence of non?repetitive disturbance w (k), the control ef?fect of different control algorithms in the process of slurry pH control is studied and analyzed.When there is disturbance w (k)=0.001 randn (1, 1)in the slurry pH control process, CFDL?MFAPC algorithm, CFDL?MFAC algorithm and PID con?trol algorithm are used for simulation respect?ively. The simulation results are shown in Fig. 11,Fig. 12 and Fig. 13. It can be seen from the three figures that when CFDL?MFAPC algorithm and CFDL?MFAC algorithm are used, the tracking control of the system output pH with respect to the expected pH is almost unaffected, while the tracking error is still controlled in a small range.However, when the PID control algorithm is used, the system output pH fluctuates greatly.The simulation results of the three algorithms show that the CFDL?MFAPC algorithm not only has a certain ability to suppress the external dis?turbance in the slurry pH control process, but also converges faster and overshoots less throughout the whole control process. Therefore,better control effects can be obtained by using CFDL?MFAPC algorithm. Therefore, compared to the other two algorithms, CFDL?MFAPC al?gorithm can achieve the best control effect in the slurry pH control process of the absorption tower.

Fig. 11 pH tracking curve with CFDL?MFAPC algorithm

Fig. 12 pH tracking curve with CFDL?MFAC algorithm

Fig. 13 pH tracking curve with PID control algorithm

4 Conclusion

In order to address characteristics of non?lin?earity and long lag time in the slurry pH change process of the absorption tower in limestone?gypsum WFGD, a model free adaptive predict?ive control algorithm is proposed. This control al?gorithm combines the advantages of model free adaptive control algorithm and model predictive control algorithm. In order to improve the auto?mation level of the WFGD process and the con?trol accuracy of the slurry pH of the absorption tower, CPS technology is introduced into the process of slurry pH control in WFGD, and a control method of the slurry pH of absorption tower based on CPS framework is proposed. The simulation results show that the model free ad?aptive predictive control algorithm can effect?ively realize the high?precision tracked control of the slurry pH. Compared to the CFDL?MFAC al?gorithm and the PID control algorithm, when the CFDL?MFAPC algorithm is used, the output overshoot of the system is small, the conver?gence speed is fast, and the external disturbance can be aptly suppressed. Therefore, CFDL?MFAPC algorithm is suitable for the slurry pH tracking control of the absorption tower.