Control action simulation for industrial boiler ... · Control action simulation for industrial boiler ... a practical depiction of the drum water level dynamics. ... simulation for

ISSN 1 746-7233, England, UKWorld Journal of Modelling and Simulation

Vol. 13 (2017) No. 1, pp. 27-36

Control action simulation for industrial boiler performance analysis∗

Susmita Das1, Sourajit Ghosh2†, Biswarup Neogi3, Amira S. Ashour4, Nilanjan Dey5

1 Department of Electronics & Instrumentation Engg., Narula Institute of Technology, Kolkata, India2 S & H Manufacturing & Trading Pvt. Ltd, Kolkata, India

3 Department of Electronics & Communication Engineering, JIS College of Engg., Kolkata, India4 Department of Electronics and Electrical Communications Engineering, Faculty of Engineering, Tanta University,

Egypt5 Department of Information Technology, Techno India College of Technology, Kolkata, West Bengal, India

(Received December 21 2015, Accepted August 15 2016)

Abstract. This paper is constructed with the effort of development of controlled system on the applicationof a traditional Zeigler-Nichols (Z-N) tuning method implemented for a boiler-turbine power plant to achievebetter efficiency of the system. Extensive study and analysis of the state space representation to mathemati-cal modeling approach is developed for the system introducing simulation method. A proportional-integral-derivative (PID) tuner and a state feedback controller for an existing non-linear MIMO (multi-input/multi-output) plant model are considered for the analysis of the system performance, which can be generalizedto design perfect industrial system. The contribution of the proposed study is to tune the boiler-turbine sys-tem model using Zeigler-Nichols(Z-N) method with PID tuning for step input. Different tuning parametersand robustness parameters are observed after the tuning adjustments implemented on the system. Finally, theMIMO system can be split toward SISO (single input/single output) system. In addition, the controllabilityand observability approaches are depicted along with system performance analysis with step input. The exper-imental results for the control systems design with the Z-N control algorithm are represented in the z-domainfrom continuous domain for discrete domain stability analysis in future.

Keywords: Z-N tuning, Proportional-integral-derivative (PID), multi-input/ multi-output (MIMO), statespace, Z-domain, controllability, observability

1 Introduction

Processes engaged in industrial boiler-turbine system to produce steam in power plants are non-linearas well as very much complex. The combustion process analysis is based on the burner configuration andthe shape/ size of the combustion chamber. In real-time applications, the control phenomena achieved a hugeattention for its robust control of the power plants[3]. Typically, industrial automatic control is considered asa compulsory condition for controlled operation. It minimizes the material fatigue and the number of staff aswell as enables efficient power plant system. The most critical task in the power plant is the controller design.Its design and tuning are challenging and very much essential for the power plants specifically boiler-turbineoperated plants due to their typical non-linearity and multi-variability along with the multiple control aims.However, the PID compensators conventional controls provide an adequate response without any flexibility toachieve superior performance over a wide zone of operation. Tuning methods must be applied for the boiler-turbine plants to investigate the proper adjustment of the controller designing with controlling parameters

∗ Dey’s group and other researchers conducted extensive research to optimize the PID controller parameters using optimizationalgorithms [9–16, 18, 25, 26]. In addition, the current work tunes the PID parameters using the Z-N method, which can be used inseveral industrial applications. In addition, as a future work, this tuned model can be represented as a State space form.

† Corresponding author. E-mail address: [email protected]

Published by World Academic Press, World Academic Union

28 S. Das & et al.: Control action simulation for industrial boiler performance analysis

(proportional gain, integral gain, derivative gain) and for the betterment of the designed control system forthat particular plant.

The oil-fired boiler has rated power 160MW. In 1969, data acquired during experiments established thebase for the system recognition. Both physics and empirical methods were used for the production of theboiler-turbine dynamic model presentation[3]. Control loop tuning is performed via adjusting its control pa-rameters, namely integral gain/reset, gain/proportional band and the derivative gain/rate to the optimum valuesfor the expected control response. Commonly, it is compulsory to guarantee response stability as well as non-oscillated process for any arrangement of set points and process conditions[2]. In modern technical society,automatic control system is used for better production quality and quantity. In the automated industry someparameters such as the pressure, flow rate, temperature, level and pH should be measured and controlledby process controller for manufacturing the end product with ensured safety. Process control mechanism isrequired by the manufacturers in order to gain: less cost, reduced variability, protection and high quality prod-ucts. The boiler-turbine model used was designed with perfection[1]. There are several types for tuning pro-cesses such as the Z-N (Ziegler-Nichols) method, C-C (Cohen-Coon) method, PID tuning[27] software and themanual tuning. Traditional and recent technology based tuning methods are available for PID tuning. There aremanual tuning including Ziegler-Nichols tuning[19], Software tools, Cohen-Coon, GA (Genetic Algorithm),PSO (Particle Swarm Optimization) and BFO (Bacterial Forging Optimization). The Ziegler-Nichols methodfor closed/ opened loop systems and the Cohen-Coon tuning methods for open loop systems are the mostpopular methods. The model is a third order, non-linear MIMO system given by Astrom and Eklund[4]. Thetest data are used to achieve basic parameter estimation that was involved in the model proposed in [5]. Thisled to a second order non-linear system of differential equations with fuel flow and control valve setting.The system output was consisted of the control variables, drum pressure and the power output. Morton andPrice restructured the boiler model to comprise drum water level deviation by presenting evaporation ratedynamics[7]. This model was extended to comprise the power given Bell and Astrom et al.[1] in order to createa third order non-linear MIMO system. This extended model was engaged with control valve position, fuelflow, and feed water flow as control inputs. Meanwhile, outputs included the drum pressure, power output,and drum water level deviation. The experimental results verified that the evaporation equation presence andthe fluid density dynamics in Bell and Astrom’s model produced a practical depiction of the drum water leveldynamics. Even though, the model is of low order, it was adept to illustrate some of the complex dynamicslinked with the real plant[4]. Tuning methods were applied to control of process variables in a proper way[6].The analysis of traditional PID, fuzzy logic and PID-Fuzzy has been performed using MATLAB version 2013and simulink software. The obtained results depicted small overshoot and fast response as compared to PIDcontroller[21]. Hemalatha et al. demonstrated the necessity of Boiler level control using Labview [8]. Multipleprocess control variables were controlled using Labview for complex systems[22]. Karuppiah et al. proposedboiler temperature controlling system using GSM (Global System for Mobile Communications) system[17]. Inthis paper, the capability investigation of the inclusive process simulation model with a prediction of the tran-sient response of a triple-pressure heat recovery steam generator with reheater to the start-up and shutdownprocesses of a heavy-duty gas turbine was introduced[20]. The presentation of advanced predictive controlmethods in thermal processes have been developed in this paper[23]. In this paper the proposal of engineeringfriendly control strategy to handle the several problems in fluidized bed combustor is given[24].

In the current work, the boiler-turbine model is used based on the work that previously done by R. Dimeoet al.[3]. It is a authentic work and a State Space matrix is proposed i.e. ABCD matrix [3] from which themathematical model is obtained through MATLAB 2013 programming. Using simulator, tuning of the modelis performed and those tuned responses can be seen in the next phase of the research work. The structure of theremaining sections is as follows. Section 2 represented the state space model in mathematical model approachfor further analysis of the system. The models are tuned using Z-N tuning method and 3D plot is obtained fortheir robustness parameters in section 3. In section 4, the rank of observability and controllability is obtainedwith discrete domain analysis. Finally, the conclusion is presented in section 5.

WJMS email for contribution: [email protected]

World Journal of Modelling and Simulation, Vol. 13 (2017) No. 1, pp. 27-36 29

2 Control tuning aspect towards existing boiler model

The State space is a mathematical representation of set of values which are linearly independent for allpossible states of the system. This mathematical model has been chosen based on Dimeo et al.[3], where theState Space matrix is presented below. The linear approximation to the system is given in Eqs. (1) and (2) asfollows.

sdx

dt= Ax+Bu, (1)

y = Cx+Du, (2)

where x = x− x0, y = y − y0, u = u− u0 and the linear system matrices are given by:

A =

−2.509× 10−3

6.94× 10−2

−6.69× 10−3

00−0.1000

, (3)

B =

0.900

−0.349− 0.1514.1550−1.3891.659

, (4)

C =

016.34× 10−3

001004.71× 10−3

, (5)

D =

000.253

00000.512− 0.014

. (6)

Afterward, the state space mathematical model (Eqs. (3), (4), (5) (6)) is transformed into transfer func-tional approach for the analysis of the system. It is a MIMO system, where each input three outputs areobtained as in Fig. 1. So, from three sets of inputs nine models are observed for tuning purpose of the system.

Fig. 1: State space to mathematical model conversion presentation for a MIMO system

The contribution of the current work is to perform the analysis with Z-N tuned mathematical modelsfor the better efficiency of the boiler system, while the previous related work[2] was done with the un-tunedprocess. MATLAB (version 2013) software is executed on nine sets of models. Thus, Fig. 2(a, b and c) areobtained to illustrate the unit step response graphs of the proposed function before tuning, which input to theMIMO system.

WJMS email for subscription: [email protected]


(a) Step response of T.FI/P1O/P1

(b) Step response of T.FI/P1O/P2

(c) Step response of T.FI/P1O/P3

Fig. 2: Before tuning: step function response of the MIMO system for the 1st input

Thus, Fig. 3(a, b and c) are obtained to illustrate the unit step response graphs of the proposed functionbefore tuning, which input to the MIMO system.

Thus, Fig. 4(a, b and c) are obtained to illustrate the unit step response graphs of the proposed functionbefore tuning, which input to the MIMO system.

The preceding step function graphical presentation is tuned in the next section as the second phase of theproposed work.

3 Tuning methodology introducing simulation effect

Applying Z-N tuning method on the step responses of the systems the mathematical model as demon-strated in Fig. 5.

The Z-N tuning method has been applied on the nine sets of mathematical model. Mainly, software toolis used for tuning purpose. After applying Z-N tuning method the graphical representation is demonstrated inFig. 6(a, b and c). After PID tuning, the obtained step response graphs are illustrated as follows.

After applying Z-N tuning method the graphical representation is demonstrated in Fig. 7(a, b and c).After PID tuning, the obtained step response graphs are illustrated as follows.

After applying Z-N tuning method the graphical representation is demonstrated in Fig. 8(a, b and c).After PID tuning, the obtained step response graphs are illustrated as follows.

The optimal tuned parameter Kp (proportional gain) Ki (integral gain) Kd (derivative gain) values of thePID controller using the Z-N method tuning method are depicted in Tab. 1 for the proper adjustment of thedesigned controller system for the boiler-turbine plant. The rise time, settling time, peak and overshoot valuesare shown in Tab. 2, which characterize the optimal parameters.

The obtained results in Tab. 2 are matched with the same illustrated in Figs. 6-8. As reported theT.FI/P3O/P3 had the superior robustness metrics, namely the rise time, settling time, peak and overshoot values.Fig. 9 illustrated the 3D plots of the above process characteristic parameters (robustness parameters).


World Journal of Modelling and Simulation, Vol. 13 (2017) No. 1, pp. 27- 36 31

(a) Step response of T.FI/P2O/P1 (b) Step response of T.FI/P2O/P2


Fig. 3: Before tuning: step function response of the MIMO system for the 2nd input

Table 1: Tuning Parameter

Kp ki kd

0.0002906 2.1882× 10−7 02.1882× 10−7 −1.5251× 10−9 −0.00081155−0.0058923 −3.2412× 10−9 44.334−0.007494 −5.6429× 10−5 0−0.00027214 −5.746× 10−8 0.1358−0.00098741 −3.8054× 10−8 −0.26886−0.017436 - 0.00013129 00.0039114 3.8496× 10−6 0.001209420.1143 0.024527 −34.4373

4 Controllability, observability and rank determination

The relationship between state and input is the Controllability and Observability gives the informationabout the capability of external measurement of the internal system state variables. The Controllability andobservability performance analysis of the system are calculated. State space modeling for the better entities ofthe system is necessary as well the controllability and observability, which determined in Tab. 3.

The information of the system input and output parameter towards better performance ranks are observedfrom the results in Tab. 4.

Since, both the controllability and observability matrices have the same rank value, so the system iscontrollable and observable as well. These results established the efficiency of the proposed system to selectthe optimal PID controller parameters. Finally, the previous stability analysis of the system in the continuousdomain is expressed in the discrete domain (‘z’ domain) as follows (Eqs. (7), (8), (9) (10)):

Step response of the

T.FI/P1O/P1 =9.2e−5z − 1.6e−5

z2 − 2z + 0.9999. (7)






Fig. 4: Before tuning: step function response of the MIMO system for the 3rd input

Fig. 5: State space to Mathematical model conversion for a MIMO system after PID tuning using the Z-Nmethod tuning method


T.FI/P1O/P2 =−6.618e−5z3 + 8.453e−5z2 + 2.636e−5z − 4.468e−5

z4 − 3.89z3 + 5.939z2 − 3.939z + 0.9797. (8)


T.FI/P2O/P1 =0.0007999z − 0.0002488z2 − 1.999z + 0.9995

. (9)







Fig. 6: After tuning: step function response of the MIMO system for the 1st input

(a) Step response of T.FI/P2O/P1 (b) Step response of T.FI/P2O/P2


Fig. 7: After tuning: step function response of the MIMO system for the 2nd input






Fig. 8: After tuning: step function response of the MIMO system for the 3rd input

Table 2: Robustness parameter

Step function Rise Time(sec) Settling Time(sec) Peak Over- shoot (%)T.FI/P1O/P1 3.23× 103 1.18× 104 1.14 13.8T.FI/P1O/P2 964 3.39× 103 1.1 9.51T.FI/P1O/P3 1.99× 104 6.51e× 104 1.06 5.74T.FI/P2O/P1 323 1.18× 103 1.14 13.8T.FI/P2O/P2 672 1.16× 104 1.06 6.11T.FI/P2O/P3 299 977 1.09 9.11T.FI/3O/P1 323 1.18× 103 1.14 13.8T.FI/P3O/P2 23.5 89.8 1.1 10.3T.FI/P3O/P3 7.73 27.5 1.05 5.27

(a) Settling time and Rise time (b) Settling Peak and Overshoot

Fig. 9: 3D plot of the robustness parameters

T.FI/P3O/P1 =0.0005403z − 0.0005395z2 − 1.999z + 0.9995

. (10)



Table 3: Robustness parameterStep function Controllability matrix Observability matrix

T.FI/P1O/P1

„0.0003 0.0000190.0019 0

« „0.0003 0.00190.0019 0

«

T.FI/P1O/P2

[email protected] −0.1025 0.0103 −0.0010

0 1.0000 −0.1025 0.01030 0 1.0000 −0.10250 0 0 1.0000

1CCA0BB@−0.0001 −0.0025 −0.0004 0.0000−0.0023 −0.0004 0.0000 00.0001 0.0000 0 00.0000 0.0000 0 1.0000

1CCAT.FI/P1O/P3 NA NA

T.FI/P2O/P1

„1.0000 −0.0025

0 1.0000

« „0.0026 0.01380.0138 0

«T.FI/P2O/P2 NA NAT.FI/P2O/P3 NA NA

T.FI/P3O/P1

„1.0000 −0.0025

0 1.0000

« „0.0027 0.00000.0000 0

«

T.FI/P3O/P2

[email protected] −0.0025 0.0103 −0.0010

0 1.0004 −0.1025 0.01030 0 1.0000 −0.10250 0 0 1.0000

1CCAT.FI/P3O/P3 NA NA

Table 4: Rank of controllability and Observability matrix

Step function Controllability rank Observability rankT.FI/P1O/P1 2 2T.FI/P1O/P2 4 4T.FI/P1O/P3 NA NAT.FI/P2O/P1 2 2T.FI/P2O/P2 NA NAT.FI/P2O/P3 NA NAT.FI/3O/P1 2 2T.FI/P3O/P2 4 4T.FI/P3O/P3 NA NA


T.FI/P3O/P2 =0.000545z3 − 0.001247z2 + 0.0009891z − 0.0002874

z4 − 3.98z3 + 5.939z2 − 3.939z + 0.9797. (11)

From the previous processes including tuning, controllability, observability and z-domain analysis, it isestablished that the MIMO boiler system achieved superior performance.

5 Conclusion

The proposed study depicted an interdisciplinary industry oriented work with control system. This gener-alized model can be defined for any type of control system based power plant operation. During the runtime ofthe process the designed system can be implemented and the process parameters are continuously monitoredfor better performance. In the current study, the State space to transfer functional approach was presented.

In future, from the obtained MIMO system further stability analysis can be done comparing with theprevious unbounded system [3]. Utilizing other different types of advanced methods of tuning such as GeneticAlgorithm, Particle Swarm Optimization, Bacterial Forging, FBADE etc. the designed control system can becriticized. In addition, using other several advanced and more efficient stability analysis methods such as Jurystability, Lyapunov stability method for non-linear system analysis can be attempted for a better investigationprocess about the boiler-turbine power plant system efficacy.

References

[1] R. Bell, K. Astram. Dynamic models for boiler-turbine-alternator units: Data logs and parameter estimation for a160mw unit. 1987.



[2] S. Das, A. Chakraborty, et al. Study on different tuning approach with incorporation of simulation aspect for z-n(ziegler-nichols) rules. Nuclear Physics B, 2012, 2(1-2): 2250–3153.

[3] R. Dimeo, K. Y. Lee. Boiler-turbine control system design using a genetic algorithm. IEEE Transactions on EnergyConversion, 1995, 10(4): 752–759.

[4] K. Eklund. A simplified non-linear model of a drum boiler-turbine unit. International Journal of Control, 1972,16(1): 145–169.

[5] K. Eklund. The controllability of steam output, pressure and water level in drum boilers. Steam at Work: Industrialand Marine Steam Plant, 1977, 75–84.

[6] Z. Gao, Y. Huang, J. Han. An alternative paradigm for control system design. in: Decision and Control, 2001.Proceedings of the IEEE Conference on, 2001, 4578–4585 vol.5.

[7] S. Ghosal, R. Darbar, et al. Application of Swarm Intelligence Computation Techniques in PID Controller Tuning:A Review. Springer Berlin Heidelberg, 2012.

[8] B. Hemalatha, J. A. Vimala, N. Natarajan. Boiler level control using labview. International Journal of ComputerApplications, 2010, 11(17): 85–88.

[9] K. Jagatheesan, B. Anand, et al. Artificial intelligence in performance analysis of load frequency control inthermal-wind-hydro power systems. International Journal of Advanced Computer Science & Applications, 2015,In-Press(7).

[10] K. Jagatheesan, B. Anand, N. Dey, A. S. Ashour. Conventional controller based agc of multi-area hydro-thermalpower systems. in: International Conference on Computer Communication and Informatics, 2016, 169–179.

[11] K. Jagatheesan, B. Anand, et al. Proportional-integral-derivative (pid) controller equipped lfc of multi-area inter-connected power system. in: V International Conference Industrial Engineering and Environmental Protection,2015.

[12] K. Jagatheesan, B. Anand, et al. Load frequency control of hydro-hydro system with fuzzy logic controller consid-ering non-linearity. in: World Conference on Soft Computing, 2016.

[13] K. Jagatheesan, B. Anand, et al. Application of flower pollination algorithm in load frequency control of multi-areainterconnected power system with non-linearity. Neural Computing & Applications, 2016, In-press: 1–14.

[14] J. Kaliannan, B. Anand, et al. Load frequency control of multi-area interconnected thermal power system: Artificialintelligence based approach. International Journal of Automation & Control, 2016.

[15] J. Kaliannan, B. Anand, et al. Dynamic performance analysis of agc of multi-area power system consideringproportional-integral-derivative conroller with different cost functions. in: International Conference on Electrical,Electronics, and Optimization Techniques, 2016.

[16] J. Kaliannan, A. Baskaran, et al. Particle swarm optimization based parameters optimization of pid controller forload frequency control of multi-area reheat thermal power systems. International Journal of Advanced IntelligenceParadigms, 2015, In-press.

[17] T. Karuppiah, V. Sivasankaran, et al. Embedded system based industrial power plant boiler automation using gsmtechnology. International Journal of Advanced Research in Computer and Communication Engineering, 2013,2(8).

[18] K.Jagatheesan, B.Anand, et al. Effect of non-linearity and boiler dynamics in automatic generation control of multi-area thermal power system with proportional-integral-derivative and ant colony optimization technique. in: Recentadvances in nonlinear dynamics and synchronization with selected applications, Springer International Publishing,2016.

[19] T. R. Kurfess. Getting in tune with ziegler-nichols. Control Engineering, 2007, 54(2): 28–28.[20] N. Mertens, F. Alobaid, et al. Dynamic simulation of a triple-pressure combined-cycle plant: Hot start-up and

shutdown. Fuel, 2015, 167: 135–148.[21] R. Shakya, K. Rajanwal, et al. Optimization and designing of pid, fuzzy & pid-fuzzy controller. International

Journal of Scientific & Engineering Research, 2014, 5(1).[22] P. Srinivas. Modelling and simulation of complex control systems using labview. International Journal of Control

Theory & Computer Modeling, 2012, 2(4): 1–19.[23] G. Stojanovski, M. Stankovski. Predictive Control of Complex Industrial Thermal Processes. Springer International

Publishing, 2016.[24] L. Sun, D. Li, K. Y. Lee. Enhanced decentralized pi control for fluidized bed combustor via advanced disturbance

observer. Control Engineering Practice, 2015, 42: 128–139.[25] S. Tyagi, K. K. Bharadwaj. A particle swarm optimization approach to fuzzy case-based reasoning in the framework

of collaborative filtering. International Journal of Rough Sets & Data Analysis, 2014, 1(1): 48–64.[26] M. J. Walsh, J. Barton, et al. An antiwindup approach to power controller switching in an ambient healthcare

network. International Journal of Ambient Computing & Intelligence, 2011, 3(2): 35–55.[27] J. Zhong. PID controller tuning: A short tutorial. Mechanical Engineering, Purdue University Spring, 2006 K.

Elissa.


Documents

Control action simulation for industrial boiler ... · Control action simulation for industrial boiler ... a practical depiction of the drum water level dynamics. ... simulation for