Research Article 12 Control of the Fractionator Top Pressure for a Delayed Coking Unit in Khartoum Refinery

Research Article 12 Control of the Fractionator Top Pressure for a Delayed Coking Unit in Khartoum Refinery Salah Eldeen F..Hegazi 1, Gurashi Abdallah Gasmelseed 2, Mohammed M.Bukhari 3 1 Department of Chemical Engineering,Faculty of Engineering,Jazan University,Saudi Arabia. Email :salahhegazi2003@yahoo.com 2 Department of Chemical Engineering, Faculty of Engineering, University of science &technology -Sudan- Email :gurashigar@hotmail.com 3 Department of Chemical Engineering, Faculty of Engineering, University of Bengahzi Libya (Received November 02, 2014; Accepted February 28, 2015) Abstract -The control of the fractionator top Pressure for delayed coking unit in Khartoum refinery, by keeping it in certain limits is essential to adjust disturbances which lead to deviation from the desired values. The aim of this paper is to select the suitable controller and the appropriate tuning method. The control loop is designed depending upon controllability and performance. The column top pressure cascaded with a reflux rate. The mathematical model for the loop was determined using MATLAB toolbox system identification by designing Graphical User Interface (GUI). The same was used to identify the transfer function for each loop. Having identified the transfer functions, the loop was closed and the characteristic equation was obtained from the overall transfer function using Routh-Hurwitz together with the direct substitution method to get the ultimate gain and ultimate period, from which the adjustable parameters were determined using Z-N criterion. These were used to investigate the offset upon a unit step change in the set point using proportional only (P), proportional integral (PI) and proportional integral derivative controller (PID).The controller that gave the minimum offset is found to be proportional integral derivative controller ( PID) and has been selected. The procedure is repeated using Root-Locus and Bode criterion. In each case the controller that gave the lowest offset was selected. The three methods of Routh, Roots Locus and Bode were compared according to their offset and found to be within good accuracy. However, for good initial estimate of the adjustable parameters, it was recommended to apply proportional gain (k c =2.5, integral time ( =0.75min ) and derivative time( for system tuning, with respect to the PID controller. Having closed the loops in conventional system, the same was transformed to digital computer control. The analysis of the offset investigation and system stability were performed on (Z-domains). The results are in agreement with conventional analysis. Index terms: Delayed coking, Roots Locus, Bode, tuning, conventional. I. INTRODUCTION Control of Delayed coking unit Coker Many oil refineries have nowadays integrated or are planning to develop a fairly complex unit called Delayed Coking Unit (DCU). The purpose of DCU is upgrading petroleum residue by thermal cracking[1] Most aspects of the DCU have been addressed previously modeling and simulation optimization process control [2]; [3];[4],[5;[6];[7],[8];[9]. Delayed coking is one of the most difficult refinery units to operate and control. The unit takes vacuum residue (fresh feed), heats it and injects it into the main fractionator bottom, where it is mixed with an internal reflux recycle of heavy cracked material. The total fresh and recycled feed is then heated in the coker furnace to a high cracking temperature. Coke remains in the drums and is periodically removed. That is the main reason for the coker being a difficult unit to operate. Twice daily filled coke drums are switched off for coke removal and empty drums are connected. The drum that was just filled then goes through a cycle of steaming out, cooling, opening, coke removal, closing, steaming and pressure testing, heating and finally reconnecting to the furnace and fractionator. [10] Al-fula crude is pumped from the surge tanks unit through several heat train exchangers and into the fractionator. The flow rate of coker feed is controlled upstream of the exchangers by the flow controller leading to the fractionating tower. Coker feed is preheated through three heat exchangers prior to entering the fractionator. Controlling the temperature of the Product quality is controlled by throttling the amount of reflux pumped back to the tower at various control points.[11] Problem Statement During the operation of the delayed coking unit in Khartoum refinery, the pressure at the top of coking tower is raise in severe leads to breakout the valve and lack of control. The adjustment of the operating top pressure of the delayed coking unit in its desired value by

selection the best techniques of control is the aim of this paper. Paper objectives General Objectives To develop a mathematical model using graphical user inter-phase (GUI). In order to test the performance of delayed coking unit at Khartoum. Refinery. To control of the process using controller that gives the best performance. To compare between the performance and best selection tuning method. Specific Objectives To determine the transfer function from model using system identification. To determine the optimum parameters; Kc, τ I and τ D. To tune the feedback controllers used in delayed coking unit. To compare the accuracy of Bode criteria, Root Locus method and Routh test. II. MATERIALS AND METHODS System Identification System Identification is the art and science of building mathematical models from measured input-output data. To examine the measured data, and create new data sets from the original one by various preprocessing.then the models will be estimated using the obtained data. Identification: get the system transfer function Get the closed loop characteristic equation of the system. Investigate offset for the system by calculation using Routh to get ultimate gain (Ku),ultimate period (Pu),, usepropotional gain (Kc,),integral time ( τ ) and derivative time (τ ) from Z-N table. 1. Get the system performance using P,PI and PID, upon a unit step change in set point. 2. 5-Repeat step1 to step3 for other loops. 3. Compare the performance 4. Select the controller that gives the best performance. 5. Repeat using Bode Method and Root Locus method. Select for each the controller that gives the best performance. III. RESULTS AND DISCUSSION Control of the column top pressure The top pressure of the fractionator of the delayed coking unit (DCU) in Khartoum Refinery is the primary control loop and the reflux flow rate is the secondary control loop. The MATLAB tool box system identification by the Graphical User Interface was used to determine the model and the transfer functions of the cascade control loops. 13 Different tuning methods are used for investigating the system stability. Frequency Dynamic System Analysis of the top pressure Data for pressure control Table (3.1): Operating records for Column top pressure and reflux rate Pressure, MPa 0.093 68 0.095 68 0.0109 74 0.0111 73 0.011 73 0.0114 74 0.0113 74 0.0114 74 Reflux flow(t/h) Transfer functions of the Top pressure and the reflux rate To keep the pressure of the column at designed value through controlling the reflux rate of the column, the MATLAB software supported with GUI (system identification) the transfer function was obtained. The transfer function in z-domain is: The same is transformed into S-domain: The System Performance for Closed Loop: P-Controller For closed loop transfer function using P- controller only: Routh-Hurwitz Method The characteristic equation is (4) (3) +(6.7 + 0.35 K) + (25-4.5 k) S +33.6 +5.05k=0 Routh Array : S3 1 25 4.5 k S 2 6.7+0.35k 33.6+5.05k S1 (25-4.5k) -

S0 33.6+5.05k 25-4.5 k - =0 The ultimate gain Ku = 4.1 The characteristics equation is +(6.7 + 0.35 K) + (25-4.5 k) S +33.6 +5.05k=0 (5) putting s=i.w(6) -I 14 w=4.3 rad/sec, The ultimate period Pu=2π/ω=1.46 min Using Ziegler Table; Kc=2.1 for P-controller only. The step response curve is illustrated infigure(3.1) below Fig.(3.1):The final value of closed loop response of the system using P-controller only The Offset for Closed Loop Transfer Function The Transfer function is: ) The offset = 0.24 1 = -0.76 S 2 6.7+0.35k 33.6+5.05k S1 (25-4.5k) - S0 33.6+5.05k τ 25-4.5 k - =0 PI-Controller For closed loop transfer function using PI-controller Using Z-N table; Kc=1.85 and =1.2

3.8 The Offset for Closed Loop Transfer Function R(t)=1.0(14) R(s) = The offset =0.28-1= -0.72 The step response curve is illustrated below τ Fig.(3.2): The final value of closedresponse of the system using PI-Controller r(t)=1.0 (19) PID-Controller For closed loop transfer function using PID-controller R(s) = (20) From Z-N table; Kc=2.5,, The Offset for closed loop transfer function for PID controller The step response curve is illustrated in Figure (3.3),The offset =1-1= 0.0

16 Fig.(3.3): The final value of closed response of the system using PID-Controller using Routh test. Fig.(3.4): The final values for different type of controllers using Routh test Fig.(3.5): The final values and overshoots for different type of controllers using Routh test.

For the system performance using Routh test as a tuning method and proportional controller. Figure(3-1) illustrates the offset =-0.76.Using PI controller reduce the offset to -0.72 illustrated in Figure(3.2). Adding derivative controller to PI eliminates the offset to zero according to Figure(3.3).The overshoot=28.6%.due to elimination of the offset the is PID controller is selected. Controller Tuning Using Root-Locus Plots 17 To determine the ultimate gain and ultimate period from the root locus, this can be realized by the following methods: Using the close loop transfer function And OLTF is given as: The OLTF= (24) Draw the roots locus using MATLAB software Fig.(3.6): Root- Locus Analysis for controlling top pressure of the column Expand equation = (31) +(6.7 + 0.35 KC) + (25-4.5 kc) S +33.6 +5.05k=0 (25) Putting s=i.w (26) Equating the imaginary part to zero gives: - Equating the real part to zero gives - The equation becomes: (28) Substituting into equation (28) =4.1 The offset =0.24-1= -0.76 The step response curve is illustrated Fig.(3.7) Using Ziegler Table; Kc=2.1 for P-Controller. The transfer function is: r(t)=1.0 (30)

18 Fig.(3.7):The final value of closed loop response of the system using P-controller only. PI-Controller For closed loop transfer function using PI-controller τ Then; Kc=1.85 and =1.2 3.15 The Offset for Closed Loop Transfer Function r(t)=1.0 (35) The offset =0.28-1= -0.72 The step response curve is illustrated in Figure(3.8) Fig.(3.8):The final value of closed loop response of the system using PI-Controller PID-Controller The Offset for Closed Loop Transfer Function for For closed loop transfer function using PID-controller PID controller r(t)=1.0 (40) Taking Kc=2.5,τ τ, R(s)= (41)

The offset =1-1= 0.0 The step response curve is illustrated in figure below Fig.(3.9): The final value of closed loop response of the system using PID-Controller Fig.(3.10): The final values for different type of controllers using Root Locus method. Fig.(3.11):The overshoot of closed loop response of the system for different type of controllers using Root-Locus method.

For the system performance using Root Locus as tuning method and proportional controller. Figure.(3.7) illustrates the offset = -0.76. Using PI controller reduce the offset to -0.72 illustrated in Figure.(3.8) Adding a derivative controller to PI eliminates the offset to zero according to figure(4-9).the overshoot using PID is 28.6% illustrated in 20 Figure.(3.11).Due to elimination of the offset the best selection is PID 3.19 Controller Tunning Using Bode Method The OLTF= Using MATLAB commands for plotting bode diagram as illustrated in Figure.(3-12) below Figure (3.12):Bode diagram for controlling toppressure of the column The results of MATLAB are: (45) =2.43 min M= (47) Log AR=log kc+log (48) Log AR=log kc+log AR=1; Ku=3 Using Ziegler Table ;Kc=1.5 for P-Controller. 3.17 The Offset for Closed Loop Transfer Function for P-Controller only: The Transfer function is: r(t)=1.0 (50) R(s)= (51) The offset =0.184-1= -0.186 The step response curve is illustrated in figure below

21 Fig.(3.13): The final value of closed loop response for the system using P-controller only PI-Controller For closed loop transfer function using PI-controller τ Using from Z-N table; Kc=1.35 and =2.03 4-22 The Offset for Closed Loop Transfer Function (55) The offset =0.1682-1= -0.8318 The step response curve is illustrated in Figure(3.14) Fig. (3.14): The final value of closed loop response of the system using PI-Controller PID-Controller For closed loop transfer function using PID-controller, (61) Taking Kc=1, (60) 3.24The Offset for closed loop transfer function for PID controller The offset =1-1= 0.0 The step response curve is illustrated in Figure(3.15)

22 Fig. (3.15): The final value of closed loop response of the system using PID-Controller Fig. 3.16: The final values of closed loop response of the system using different types of controller applying Bode criterion. Fig.(3.17): The overshoot of closed loop response of the system using different types of controllers applying Bode criterion.

For the system performance using Bode criterion as a tuning method and proportional controller. Figure(3-13) illustrates the offset = -0.8318.Using PI controller reduce the offset to -0.186 illustrated in Figure(3-14). Adding a derivative controller to PI eliminates the offset to zero according to figure(3.15).the overshoot illustrated in Figure (3.17) is 0% According to elimination of the offset the best selection is PID using Bode method compared to Routh and Root Locus method. IV. CONCLUSION AND RECOMMENDATION The delayed coking unit in Khartoum Refinery is considered as one of the important unit for treatment of the heavy crude oil (Al-fula crude ).The Graphical User Interphase is used in identification and analysis the control system. The dynamic performance of the system is investigated,the selection of the best mode of controller is selected (PID-Controller). Different tuning methods were used. These are : Ziegler, Bode, Nyquist, Root Locus and Routh test. From the control point of view, the delayed coking process is a solution to the problem of decreasing residual fuel demand. It also generates a variety of fuels and in some cases a considerable amount of high quality coke, while eliminating environmentally unfriendly streams that often involve a disposal cost. Implementing advanced process control and optimization on a coking plant is quite a difficult task but the results could be remarkable: energy savings, maximized throughput, decreased CO emissions and improved yields while increasing the overall profit of the refinery. Further work has to be done on the following: 1-Maximizing hot residue from crude distillation unit to delayed coker. 2-Maximizing LPG production and fuel gas utilization. 3-Implementation of advanced process control. 4-Application of override control is essential for controlling the base level. ACKNOWLEDGEMENTS The authors wish to thank the Faculty of Graduation Studies in Gezira University, and Khartoum Refinery Company for their help and support, this Paper. 5. References [1] HessamVakilalroayaeietal, Dynamic behavior of coke drum process safety valves during blocked outlet condition in the refinery delayed coking unit,2012 [2] Elliott, J.D., "Optimize coker operations", Hydrocarbon Processing, vol.82,9,2003, p.85-90 23 [3] G Zahedi, A Lohi, Z Karami,A neural network approach for identification and modeling of delayed coking plant, International Journal of Chemical Reactor Engineering,2009 [4] Haseloff, V.& Friedman, Y.Z.&Goodhart, S.G., "Implementing coker advanced process control", Hydrocarbon Processing, vol. 86, 6, June 2007, p.99-130 [4] Chang, A.I.&Nishioka, K.&Yamakawa, T., "Advanced control project stabilizes delayed coker, increases throughput", Oil and Gas Journal, vol. 99, 34, 2001, p.52-56 [6] Depew, C.A.&Hashemi, M.H.& Davis, J., "Evaluation of alternative control strategies for delayed coker by dynamic simulation", Proceedings of the American Control Conference, 1988, p.240-246 [7]Haseloff, V.& Friedman, Y.Z.&Goodhart, S.G., "Implementing coker advanced process control", Hydrocarbon Processing, vol. 86, 6, June 2007, p.99-130 [8] Chen, Q.L.& Yin, Q.H.& Wang, S.P.& Hua, B., "Energy-use analysis and improvement for delayed coking units", Energy, vol. 29, 12 15, 2004, p.2225-2237 [9] Wang, C.& Chen, Q.& Hua, B., "Analysis of delayed coking process of different heat exchange and fractionation options", Petroleum Refinery Engineering, vol. 36, 2, 2006, p.19-22 [10] NurulHamizahBintiBaharan, ''Analysis of partial least square estimation process and control of distillation column process'', Faculty of Chemical Engineering and Natural Resources University College of Engineering and Technology Malaysia, November 2006 [11] Manual of Delayed Coking Unit at Khartoum Refinery, 2007.