Annals of Nuclear Energy 72 (2014) 461–466

Contents lists available at ScienceDirect

Annals of Nuclear Energy

journal homepage: www.elsevier .com/locate /anucene

Technical note

Fuzzy logic control for improved pressurizer systems in nuclear powerplants

http://dx.doi.org/10.1016/j.anucene.2014.05.0240306-4549/� 2014 Elsevier Ltd. All rights reserved.

⇑ Corresponding author. Tel.: +1 1 905 721 8668x5497; fax: +1 1 905 721 3046.E-mail address: Hossam.gabbar@uoit.ca (H.A. Gabbar).

Chris Brown, Hossam A. Gabbar ⇑University of Ontario Institute of Technology, 2000 Simcoe Street North, Oshawa L1H7K4 ON, Canada

a r t i c l e i n f o

Article history:Received 28 April 2014Accepted 16 May 2014

Keywords:CANDU stationsFuzzy logic controlPressurizer systemNPP performance improvement

a b s t r a c t

The pressurizer system in a CANDU nuclear power plant is responsible for maintaining the pressure of theprimary heat transport system to ensure the plant is operated within its safe operating envelope. Theinventory control for the pressurizer system use a combination of level sensors, feed valves and bleedvalves to ensure that there is adequate room in the pressurizer to accommodate any swell or shrinkagein the PHT system. The Darlington Nuclear Generating Station (DNGS) in Ontario, Canada currently uses aproportional controller for the bleed and feed valves to regulate the pressurizer inventory control whichcan result in large coolant level overshoot along with excessive settling times. The purpose of this paper isto develop a fuzzy controller to regulate the pressurizer inventory control and compare its performanceto the current proportional controller used at DNGS. The simulation of the pressurizer inventory controlsystem shows the fuzzy controller performs better than the proportional controller in terms of settlingtime and overshoot.

� 2014 Elsevier Ltd. All rights reserved.

1. Introduction

The primary heat transport (PHT) system for the CANDUnuclear power plant (NPP) is responsible for removing heat pro-duced by fission and fission products and transferring that heatto the boilers to produce steam for the turbines (Heat TransportSystem, 2014). The heat produced by the fuel must be removedto prevent damage to the fuel bundles and pellets to ensurecontainment is maintained. The pressurizer system is part of thePHT system and has two main functions; to maintain the coolantlevel in the core at a safe level and to maintain the pressure ofthe PHT system. Fig. 1 shows how the pressurizer fits into theprimary heat transport system for a CANDU nuclear power plant.

The pressurizer system uses a collection of valves and heaters toincrease and decrease the coolant level and pressure in thepressure vessel. Normal operating conditions at DNGS has thePHT system held at a pressure of 9.89 MPa(g) and approximately300 �C to keep the coolant in a saturated state and prevent thecoolant from boiling.

The pressurizer plays a critical role in safe plant operation, theaccident at the Three Mile Island (TMI) nuclear power plant onMarch 28th, 1979 can attest to that (Three Mile Island, 2014). Atransient occurred on the Unit 2 reactor which resulted in theautomatic actuation of the shutdown system and the control rods

were dropped into the core and stopped the fission process. As aresult of the transient, the temperature and pressure of the PHTcoolant began to rise. To accommodate the increase in pressure,the pressurizer pressure relief valve operated to reduce the pres-sure in the system. At this point, the system had operated properly.However, when the pressure in the PHT had been reduced, thepressure relief valve failed to close, and among other problems, thiseventually resulted in a severe loss of coolant to the fuel in thereactor and a partial meltdown.

The importance of the pressurizer for the PHT system cannot beoverlooked. It is for this reason that the shortcomings in the pres-surizer control system should be remedied by replacing the basicproportional controller with a fuzzy logic controller to maintain asafe and reliable system (Bhatt et al., 2009).

2. Pressurizer system

The pressurizer system for the PHT system is vital to ensure suf-ficient coolant for the fuel and accommodate the volume change ofthe PHT coolant from zero to full power (ZHANG et al., 2012; Liuet al., 2014; YE et al., 2010). The coolant (D2O) is kept in a saturatedstate with part of the pressurizer filled with liquid coolant and theremainder filled with coolant vapour. The vapour space of the pres-surizer allows for PHT pressure transients. The pressurizer itself isa cylindrical pressure vessel with pressure and level sensors tomonitor the pressurizer state.

Nomenclature

AcronymsDNGS Darlington Nuclear Generating StationNPP Nuclear Power Plant

PHT Primary Heat TransportTMI Three Mile Island

462 C. Brown, H.A. Gabbar / Annals of Nuclear Energy 72 (2014) 461–466

2.1. Design specifications

The pressurizer has two primary functions; to maintain the PHTcoolant at the appropriate pressure and to manage coolantinventory changes in the PHT due to coolant swell and shrinkage.There are two distinct control systems for the pressurizer; the firstto manage the coolant inventory and the second to maintain thepressure of the PHT coolant.

Fig. 1. CANDU primary heat transport system (CAN

The pressurizer inventory is controlled by throttling the dualfeed valves to increase coolant levels by taking coolant from theD2O storage tank and throttling the dual bleed valves to decreasecoolant levels by sending coolant to the bleed condenser. The cool-ant inventory is maintained at a set point dependent on the currentreactor power; as power increases, the coolant level is increased.The pressurizer coolant level is required to be kept within certainsafety limits. The coolant level must be kept well above the

DU training centre – heat transport system).

Fig. 2. Simplified pressurizer system model.

C. Brown, H.A. Gabbar / Annals of Nuclear Energy 72 (2014) 461–466 463

pressurizer heaters so that they do not damage themselves, butwell below the upper limit of the coolant level to allow for properregulation of the PHT inventory due to swell and shrinkage. SeeFig. 2 for a simplified pressurizer system model. The control systemfor the pressurizer inventory is a basic proportional controllergiven the pressurizer level. Improving this control system will bethe primary focus of this paper.

The pressurizer pressure control system is responsible formaintaining the coolant at the correct pressure and temperatureto keep the coolant in a saturated state as well as preheating

Fig. 3. Pressurizer c

Fig. 4. Proportional controller for CA

the PHT coolant on unit startup (Guo et al., 2011). The systempressure is controlled through the use of six heaters installed inthe base of the pressurizer to heat the coolant and increasepressure and dual steam bleed valves at the top of the pressurizerto release coolant steam to the bleed condenser and decreasepressure (Yongling et al., 2012; Oliveira and Almeida, 2011; Jinet al., 2011). The control system for the pressurizer pressure is aproportional integrative controller which is not covered in thispaper (Szederkényi et al., 2008; Kwon et al., 1997; Fazekas et al.,2007; YE et al., 2010).

ontrol system.

NDU pressurizer level control.

Fig. 5. Fuzzy logic controller pressurizer block diagram.

Fig. 6. Pressurizer level membership function.

Fig. 7. Pressurizer level rate of change membership function.

Fig. 8. Bleed and feed valve membership function.

464 C. Brown, H.A. Gabbar / Annals of Nuclear Energy 72 (2014) 461–466

2.2. Darlington Nuclear Generating Station pressurizer proportionalcontroller

The design for the pressurizer inventory controller at the DNGSuses a proportional controller using the level error between thecalculated set point and the current coolant level (See Fig. 3). Asa result of using only a proportional controller, it results in a largeovershoot and slow settling time. Fig. 4 shows the pressurizer levelresponse with a set point change from 6.1 m to 6.4 m. There is alarge overshoot to approximately 6.75 m almost instantly as theset point is increased and then a settling time of over an hour.

In the case of the pressurizer inventory, a large overshootdecreases the pressurizer’s ability to accommodate volume andpressure changes in the PHT coolant. This would have an effecton the system’s response to a transient to maintain coolant inven-tory and pressure.

3. Proposed fuzzy controller system design

3.1. Fuzzy logic control design

The fuzzy logic controller has four parts to its design; fuzzifica-tion, rule base, inference engine and defuzzification (Oliveira andAlmeida, 2011). For the pressurizer inventory fuzzy controller, Iam using the coolant level error and the rate of level change asmy inputs and the feed and bleed valve control as my output.Fig. 5 shows the block diagram of the proposed fuzzy logic control-ler adding the input value for the rate of change of the pressurizerlevel.

I am assuming a pressurizer with a coolant level ranging fromzero meters (0 m) to ten meters (10 m). Based on that, the pressur-izer level input variable fuzzification is as follows (See Fig. 6) witha dead zone of ±0.03 m (as seen by the red line in the center):

Following simulator tests, I determined that the rate of changeis between ±0.2 m per cycle. The membership function for the levelrate of change is as follows (See Fig. 7):

The outputs of the controller manipulate the feed and bleedvalves to increase and decrease the coolant inventory, respectively.The feed and bleed valve have the same membership function (SeeFig. 8), giving the valves a range from 0% to 100% open.

The system and controller have been designed in LabVIEW tooptimize the controller and compare the proportional logic con-troller to the fuzzy logic controller (See Fig. 9). The system followsthe same design as the block diagram (See Fig. 5) and clearly distin-guishes between the fuzzy logic controller and the systemsimulation.

3.2. Proposed algorithm

The rules created for the fuzzy controller using the pressurizerlevel and rate of change are shown in Table 1 below. Dependingon the error of the pressurizer level, the output is directed towardseither the bleed or feed valve.

4. Simulation and results

4.1. Simulation

The simulator used for analyzing the proposed fuzzy logic con-troller was created based on a rudimentary feed and bleed system(See Fig. 2). Using the LabVIEW Control Design and Simulationextension, a basic feed and bleed system was simulated with a tank

Fig. 9. LabVIEW pressurizer simulator and fuzzy controller block diagram.

Table 1Fuzzy controller rules.

Pressurizer level

Largenegative

Medium largenegative

Mediumnegative

Medium smallnegative

Smallnegative

Deadzone

Smallpositive

Medium smallpositive

Mediumpositive

Medium largepositive

Largepositive

Negative –Large

F F L ML SM 0 S S M L L

Negative –Medium

F F ML M SM 0 S SM ML L F

Negative –Minor

F F ML M S 0 SM SM ML L F

Rate of change Positive –Minor

F L ML SM SM 0 S M ML F F

Positive –Medium

F L ML SM S 0 SM M ML F F

Positive –Large

L L M S S 0 SM ML L F F

Bleed valve Feed valve

S – Small; SM – Small Medium; M – Medium; ML – Medium Large; L – Large; F – Full Open.

C. Brown, H.A. Gabbar / Annals of Nuclear Energy 72 (2014) 461–466 465

to test the proposed fuzzy controller versus a rudimentary propor-tional controller (See Fig. 9).

The simulation is not a perfect match to the field results due tothe simplification of the system, but it conveys the advantages ofthe fuzzy logic controller over the proportional controller as wellas its feasibility. The simulator allows for the introduction of noiseto the pressurizer level as well as manual setting of the pressurizerlevel set point.

Fig. 10. DNGS proportional controller step increase from 6.1 m to 6.4 m.

4.2. Tests and results

4.2.1. Minor set point changeThe proposed fuzzy controller was tested against the same set

point changes seen at the DNGS and shows a more stable responseto a set point change. For this test, a pressurizer inventory level setpoint change has been introduced from 6.1 m to 6.4 m with nonoise. The proportional controller response (See Fig. 10) shows alevel rising up to 6.59 m, giving an overshoot of 0.19 m and a largeoscillation before settling to the 6.4 m target. Testing the same set

point change from 6.1 m to 6.4 m with the proposed fuzzy control-ler and simulator, results in the pressurizer level rising to 6.4 mwith no overshoot (See Fig. 11). The settling time for the fuzzy con-troller is a significant improvement to the proportional controlleras the fuzzy controller settled to the new set point 4 times fasterthan the proportional controller.

Fig. 11. Fuzzy controller step increase from 6.1 m to 6.4 m (±0.05 m noise applied).

Fig. 12. Proportional controller response to major step change.

Fig. 13. Fuzzy logic controller response to major step change.

Fig. 14. Fuzzy logic controller response to small step change with ±0.05 noise.

466 C. Brown, H.A. Gabbar / Annals of Nuclear Energy 72 (2014) 461–466

4.2.2. Major set point changeThis test compares the response of the proportional controller

to the fuzzy controller for a large pressurizer inventory set pointchange from 2 m to 8 m. The proportional controller responseresults in an overshoot of 0.85 m (See Fig. 12) whereas the fuzzylogic controller is able to handle to large set point change betterand has no overshoot at all (See Fig. 13). Due to the proportionalcontroller overshoot and the simplicity of the controller, the

proportional controller has a level oscillation which increases thesettling time. The fuzzy logic controller is able to settle to thenew set point 4 times faster than the proportional controller.

4.2.3. Controller response to noiseAnalysis of the DNGS pressurizer inventory level signal shows a

noise of ±0.05 m. This test will use introduce a ±0.05 m noise signalwhile performing a minor pressurizer inventory level set pointchange from 6.1 m to 6.4 m as performed in Section 4.2.1. The sim-ulation results show that the fuzzy logic controller is able to prop-erly control the pressurizer inventory level with no loss of responsetime and settling time performance despite the introduced noise.(Fig. 14).

5. Conclusion

The proportional controller used at the Darlington NuclearGenerating Station is not optimal, resulting in large pressurizerinventory overshoots and long settling times on minor set pointchanges. In tests comparing the simulated proportional controllerand the proposed fuzzy logic controller, the fuzzy logic controllerwas shown to eradicate pressurizer inventory level overshootswhile decreasing the system settling time to 1=4 that of the propor-tional controller.

Fuzzy logic controllers are not widely used in nuclear powerstations, but definitely show that they have a place in being ableto increase system efficiency and decrease risk with a more robustcontroller when properly designed.

References

Bhatt, T.U., Madala, K.C., Shimjith, S.R., Tiwari, A.P., 2009. Application of fuzzy logiccontrol system for regulation of differential pressure in Liquid Zone ControlSystem. Ann. Nucl. Energy 36 (2009), 1412–1423.

Fazekas, C.s., Szederkényi, G., Hangos, K.M., 2007. A simple dynamic model of theprimary circuit in VVER plants for controller design purposes. Nucl. Eng. Des.237 (2007), 1071–1087.

Jianghua Guo, Xiaoli Sun, and Hao Wu, 2011. Research of Pressurizer Water LevelControl System Based on Fuzzy-PID Control. Fourth International Workshop onAdvanced Computational Intelligence Wuhan, Hubei, China; October 19–21,2011.

Heat Transport System. In CANDU Training Centre. Retrieved March 2014, fromhttps://canteach.candu.org/Content%20Library/19930204.pdf.

Ma Jin, Li Yongling, Huang Yu, Wang Bingshu, Chan Afang, 2011. Mechanism Modeland Simulation of Pressurizer in the Pressurized Water Reactor Nuclear PowerPlant. Proceedings of the 30th Chinese Control Conference.

Kee-Choon Kwon, Soon-Ja Song, Won-Man Park, Jang-Soo Lee, Jang-Yeol Kim, 1997.Development of the Test Simulator for Advanced Instrumentation and ControlResearch. IEEE Sixth Annual Human Factors Meeting.

Liu, C.Y., Yan, C.Q., Wang, J.J., 2014. Hybrid particle swarm optimization algorithmand its application in nuclear engineering. Ann. Nucl. Energy 64 (2014), 276–286.

Mauro V. Oliveira and José C. S. Almeida, 2011. Fuzzy Control Applied to NuclearPower Plant Pressurizer System. 2011 International Nuclear AtlanticConference.

Szederkényi G., Szabó Z., Bokor J. and Hangos K.M. 2008. Analysis of the networkedimplementation of the primary circuit pressurizer controller at a nuclear powerplant. 16th Mediterranean Conference on Control and Automation.

Three Mile Island. What went wrong. Retrieved March 2014 from http://www.threemileisland.org/science/what_went_wrong/.

Jian-Hua Ye, Jin-Ming Yi, Hua-Yan Ji, 2010. Research on Pressurizer Water LevelControl of Nuclear Reactor Based on RBF Neural Network and PID Controller.Proceedings of the Ninth International Conference on Machine Learning andCybernetics, Qingdao, 11–14 July, 2010.

Jian-Hua Ye, Jin-Ming Yi, Hua-Yan Ji, 2010. Research on Pressurizer Water LevelControl of Nuclear Reactor Based on RBF Neural Network and PID Controller.

Li Yongling, Ma Jin, Chan Afang, Huang Yu, Wang Bingshu, 2012. Mechanism Modelof Pressurizer in the Pressurized Water Reactor Nuclear Power Plant Based onPSO Algorithm. 2012 24th Chinese Control and Decision Conference (CCDC).

Zhang, Guo-Duo, Yang, Xu-Hong, Ye, Xiao-Long, Xu, Hang, Lu, Dong-Qing, Chen,Wen, 2012. Research on pressurizer water level control of pressurized waterreactor nuclear power station. Energy Procedia 16 (2012), 849–855.