Article pubs.acs.org/IECR
Plantwide Control of Biodiesel Production from Waste Cooking Oil Using Integrated Framework of Simulation and Heuristics Dipesh S. Patle,† Z. Ahmad,*,† and G. P. Rangaiah‡ †
School of Chemical Engineering, Engineering Campus, Universiti Sains Malaysia, Nibong Tebal, Penang, Malaysia 14300 Department of Chemical & Biomolecular Engineering, National University of Singapore, Engineering Drive 4, Singapore 117585
‡
S Supporting Information *
ABSTRACT: This article describes the systematic plantwide control (PWC) design of an ecofriendly process for biodiesel production from waste cooking oil (WCO) feedstock. A simulation model was developed to produce biodiesel from WCO that reduces both production costs and adverse environmental impacts. An effective PWC system is crucial for the safe, smooth, and economical operation of a biodiesel plant. Hence, a PWC system was developed for a homogeneously catalyzed biodiesel process using the integrated framework of simulation and heuristics (IFSH). The main merits of the IFSH methodology are effective use of rigorous process simulators and heuristics in developing a PWC system and simplicity of application. Finally, the performance of the developed control system was assessed in terms of settling time, a dynamic economic index based on the deviation from the production target (DPT), and the overall total variation (TV) in manipulated variables. These performance assessments and the results of dynamic simulations showed that the developed PWC system is stable, effective, and robust in the presence of several disturbances and that biodiesel quality can be maintained despite these disturbances. This is the first study to develop a complete PWC system for a homogeneously catalyzed two-step biodiesel production from WCO. reactors,8 and rotating packed-bed reactors.9 Mohammadshirazi et al.10 performed energy and economic analyses of biodiesel production from WCO by transesterification. Among these alternatives, a homogeneous alkali-catalyzed transesterification process is commonly used in the industry.11,12 Note that the studies described thus far focused on process development and, therefore, did not address control. Design of plantwide control (PWC) refers to the development of a control system for a complete chemical plant having interconnected unit operations with mass and/or energy recycles. The presence of recycles changes the process dynamics by introducing an integrating effect, which can lead to the snowball effect (i.e., high sensitivity of recycles to disturbances). A good control system distributes the effects of disturbance(s) to unit operations involved to avoid severe snowball effects. The design of PWC systems for safe, stable, and efficient operation of complex chemical processes has been studied extensively.13−19 Methodologies in these studies can be categorized as heuristics, optimization, mathematical, and mixed approaches.20 Rigorous mathematical- and optimization-based methodologies are complicated and require extensive computations, especially for complex chemical processes. Additionally, their solutions depend on the degree of detail used. Therefore, heuristic-based methodologies seem to be attractive because they are less complicated and easy to implement. Luyben et al.18 proposed a comprehensive ninestep procedure based on the relative importance of control and operational objectives. However, heuristic-based methods
1. INTRODUCTION Finite resources and environmental issues are likely to limit the use of fossil fuels in the future. Consequently, researchers have started exploring alternative sources of fuel. In the search for alternative fuels, biodiesel made from vegetable oil, animal fat, or waste cooking oil is seen as a potential alternative to diesel. Among several methods for producing biodiesel are pyrolysis, microemulsification, and transesterification. Transesterification can be carried out in the presence of alkali, acid, or enzyme catalyst. The reaction can also occur in the absence of a catalyst under supercritical conditions.1 Use of a homogeneous alkali catalyst in the transesterification process requires neutralization of the catalyst. Also, this process is affected by free fatty acids (FFAs) in the feedstock such as in waste cooking oil (WCO), which leads to saponification and consequently reduces the yield and also results in difficult product separation. Therefore, Canakci and Van Gerpen2 proposed a two-step process, in which the esterification of FFAs is carried out using an acid catalyst in the first step and then transesterification is performed to produce biodiesel from the treated oil using an alkali catalyst. Zhang et al.3 proposed four different methods for producing biodiesel: an alkali-catalyzed process using pure oil, an alkalicatalyzed process using WCO, an acid-catalyzed process using WCO, and an acid-catalyzed process using hexane extraction. Zhang et al.4 then carried out an economic study and concluded that the acid-catalyzed process using WCO is most cost-effective. West et al.5 concluded that the process using heterogeneous acid catalyst is more beneficial than the other process alternatives. Hass et al.6 analyzed the capital and operating costs of an alkali-catalyzed biodiesel process. Other research on the development of novel reactor designs for biodiesel production includes membrane reactors,7 gas−liquid © 2014 American Chemical Society
Received: Revised: Accepted: Published: 14408
June 12, 2014 August 10, 2014 August 24, 2014 August 25, 2014 dx.doi.org/10.1021/ie5023699 | Ind. Eng. Chem. Res. 2014, 53, 14408−14418
Industrial & Engineering Chemistry Research
Article
the complex homogeneously catalyzed two-step production of biodiesel from WCO. The present article is organized as follows: The biodiesel process design is described in the next section. Section 3 describes the application of the chosen PWC methodology to the complex biodiesel process with four liquid recycles. The performance of the designed control system in terms of settling time, a dynamic economic index based on the deviation from the production target (DPT), and overall total variation (TV) in manipulated variables, as defined in Appendix 1, is assessed in section 4. Finally, the article concludes by outlining the findings of this study.
largely depend on experience and do not extract the advantages of rigorous process simulators during the design of a control structure. For example, rigorous process simulators can be effectively used to aid in decision making in the selection of suitable pairings of controlled variables (CVs) and manipulated variables (MVs) keeping in mind the PWC perspective, such as product quality, throughput, and inventory control. The integrated framework of simulation and heuristics (IFSH) methodology14,15,20 makes effective use of rigorous process simulators. Another PWC methodology is economic plantwide control.21−23 The IFSH methodology is attractive because it is easy to implement and involves minimal computations other than process simulation. The PWC of many industrial processes has been studied so far.20 Recently, a few articles on the PWC of biodiesel processes have been reported in the literature. Shen et al.24 explored the design and control of a biodiesel process with phase segregation and recycle in the reactor system. They proposed a decentralized control structure and satisfactorily tested for a 20% change in production rate with a settling time of less than 10 h. They obtained a 20% reduction in the total annual cost and a 26% reduction in the energy cost using internally recycled glycerol. However, the limitations of their work include the facts that the oil was represented by single triglyceride and not by its detailed composition; the detailed kinetics of the reaction was not used; and pure oil was considered, which is not attractive because its high cost and limited availability. Cheng et al.25 studied the plantwide design and control of a biodiesel process using a sugar catalyst that could accommodate disturbances of ±20% in the production rate and ±5% in the FFA content. They reported that the process response was somewhat sluggish because of the large holdup in the reactors. Zhang et al.26 developed PWC for biodiesel production from pure vegetable oil that does not require the esterification section needed when using WCO as a feedstock to convert the free fatty acids (FFAs) into biodiesel. The above literature review shows that the PWC of a twostep (viz., esterification and transesterification) homogeneously catalyzed biodiesel production from WCO has not been studied so far. Additionally, detailed constituents of WCO and detailed reaction kinetics were not considered in the previous studies on biodiesel. Therefore, this study considers WCO as a feedstock, for which esterification should be carried out to pretreat FFAs, which otherwise can lead to saponification. In this study, a biodiesel process model is developed considering the detailed constituents of WCO as well as the detailed reaction kinetics. A PWC system for the complete biodiesel plant, including esterification and transesterification of the WCO, was developed. Optimum design conditions were taken from our recent work,27 in which two process alternatives were developed and optimized for conflicting objectives (namely, maximum profit, minimum heat duty, and minimum organic waste) using the elitist nondominated sorting genetic algorithm. Then, the better process was determined based on economic and environmental objectives. The limited availability and high cost of fresh oil limit its use in the production of biodiesel. The advantages of using WCO for biodiesel production are two-fold: the waste oil is used effectively, and the biodiesel production costs are reduced. Additionally, the open disposal of WCO can lead to environmental issues if it is not properly utilized. In summary, the novelty of the present study is the development and evaluation of a PWC system for
2. BIODIESEL PROCESS DESIGN Based on the potential availability of WCO in Malaysia, this study assumes a biodiesel plant capacity of 120 kt/year. The process, along with its design and operating parameters, was taken from our previous work.27 Both steady-state and dynamic simulation models of the biodiesel process were developed using Aspen Plus V8.0 and Aspen Plus Dynamics (APD) V8.0, respectively. For reliable biodiesel process modeling, it is crucial to capture the nonideal phase equilibrium, as the biodiesel process involves highly nonideal components. Shen et al.24 used the universal functional activity coefficient (UNIFAC) model for the calculation of vapor−liquid equilibrium (VLE) and liquid−liquid equilibrium (LLE) envelopes in the transesterification system. They reported that the binary azeotrope of glycerol−methanol and the LLE envelope predicted by the UNIFAC model are in agreement with the experimental data. Dortmund-modified UNIFAC was used in this study to predict the physical properties of the considered components, as it has been used successfully.27−30 The physical properties of the vapor phase were determined by the Soave−Redlich−Kwong equation of state (SRK EOS). Model parameters for thermophysical properties such as heat capacity, liquid molar volume, vapor pressure, heat of formation, heat of vaporization, and liquid viscosity for tri-, di-, and monoglycerides (TGs, DGs, and MGs, respectively) were taken from the biodiesel databank of Aspen Plus. Methods for the development of these thermophysical property models are available.31−33 The required critical pressure (Pc), critical temperature (Tc), and acentric factor (ω) for the SRK EOS were estimated by the Gani group contribution method.29 Details of the constituents of the oil were taken from the Aspen Plus Biodiesel Model.29 The composition of oil given in this model was adjusted to include 6% FFAs. Esterification and transesterification are represented by 10 and 96 reactions, respectively.27 In our previous work,27 two process alternatives for biodiesel production from WCO were optimized for multiple objectives by the elitist nondominated sorting genetic algorithm (NSGA-II), implemented in MS Excel using Visual Basic for Applications (VBA) that was linked with Aspen Plus. The better process in terms of higher profit and lower environmental impact is considered in this work. The main features of the selected process are that (1) it uses one reactor for esterification; (2) it uses three reactors for transesterification with intermediate phase separators; and (3) biodiesel−glycerol separation occurs first, followed by methanol separation and washing of biodiesel. This scheme prevents backward transesterification reactions, as methanol is present until biodiesel and glycerol are separated, and the recovered methanol contains a very small amount of water, which avoids energy-intensive methanol−water separation and facilitates 14409
dx.doi.org/10.1021/ie5023699 | Ind. Eng. Chem. Res. 2014, 53, 14408−14418
Industrial & Engineering Chemistry Research
Article
Figure 1. Simplified process diagram for biodiesel production from WCO (ID, internal diameter; RR, reflux ratio).
methanol reuse in the process (see Figure 1). Further, the chosen process is close to that used in industry.34,35 The optimal values of design parameters such as reactor temperatures, residence times, and the locations of the feed trays in the distillation columns obtained in our previous work27 were used in the present study. The distillation columns were operated under a vacuum and used cooling water in the condensers. Running distillation columns under a vacuum can impact the control strategy
significantly because the pressure on the trays can change, thereby affecting the tray temperatures. Further details about the process can be found in Patle et al.27 The surge capacities in the reactors, reflux drums, and column bases were sized to provide 5 min of holdup when at the 50% level. The steadystate conditions in the dynamic simulation using APD were found to be consistent with those in the steady-state simulation using Aspen Plus. Figure 1 presents the process flow diagram 14410
dx.doi.org/10.1021/ie5023699 | Ind. Eng. Chem. Res. 2014, 53, 14408−14418
Industrial & Engineering Chemistry Research
Article
for the production of biodiesel from WCO and also includes the values of selected design and operating parameter.
Table 1. Anticipated Disturbances in the Biodiesel Process and Their Effects on Product Flow Rate, Fresh Methanol, Recycled Methanol, and Overall Conversion
3. DESIGN OF A PWC SYSTEM In this work, a PWC system was designed based on the IFSH methodology proposed by Murthy Konda et al.36 This systematic and hierarchical methodology has eight levels, where, in addition to heuristics, both steady-state and dynamic process models are effectively utilized in decision making during the design of a control system. For example, decisions based on heuristics are corroborated using dynamic simulations. The main advantage of the IFSH methodology is that it overcomes an over-reliance on heuristics to design a decentralized regulatory control system. Each level of IFSH and its application to the chosen biodiesel process are described below. Level 1.1. Define PWC Objectives. In this level, PWC objectives are formulated from the operational point of view. Classically, these include production rate, product purity, process/equipment constraints, safety concerns, and environmental regulations. In case of any conflict between plantwide objectives and unitwise objectives, the former should be given priority. For the biodiesel process, the PWC objectives are as follows: (1) obtain a stable production rate under normal operation with rapid and smooth performance in the presence of disturbances; (2) achieve product quality according to European standard EN 14214 (biodiesel > 96.5%); (3) maintain distillation temperatures below 250 and 150 °C for the columns processing biodiesel and glycerol, respectively, to avoid thermal degradation; and (4) maintain the required ratio of methanol to oil (6:1 molar ratio under normal conditions) and the required methanol split fraction to the three transesterification reactors (RTRANS1/RTRANS2/RTRANS3 = 0.77:0.12:0.11 under normal conditions), to obtain biodiesel quality meeting EN standards. Note that different sets of PWC objectives might result in different control systems. Level 1.2. Determine the Number of Control Degrees of Freedom (CDOF). Murthy Konda et al.37 proposed a method for determining the number of control degrees of freedom (CDOF) based on the number of restraints that is easy to apply and effective. As deduced by Murthy Konda et al.,37 CDOF is found by subtracting the sum of the numbers of restraints and redundancies from the total number of streams. The number of restraints of a unit is defined as the total number of independent and overall material balances having no associated inventory. The number of redundancies in a process refers to the number of variables that need not be manipulated. In the chosen biodiesel process (Figure 3), there are a total of 106 streams (including mass and energy streams), the number of restraints is 27, and the number of redundancies is 12. Hence, CDOF = 106 − 27 − 12 = 67. Level 2.1. Identify and Analyze Plantwide Disturbances. Plantwide disturbances can pose tough challenges during normal operation of a plant. Cognizance of possible disturbances in the process is of vital importance in the development of a control scheme and controller tuning. For this purpose, a steady-state simulation model can be used to investigate the effects of possible disturbances. Table 1 lists the anticipated disturbances during biodiesel plant operation and their effects on important quantities. PWC studies in the literature24−26 have used disturbance magnitudes between ±5% and ±20% in feed flow rate. In this work, we considered up to +20% and −30% changes in WCO as the availability of WCO is
disturbance no. with details D1 (+10% in WCO flow rate) D2 (−10% in WCO flow rate) D3 (−10% in preexponential factor of transesterification reactions, due to catalyst deactivation) D4 (+5% in WCO flow rate and D3 simultaneously) D5 (D2 and D3 simultaneously) D6 (+20% in WCO flow rate) D7 (−30% in WCO flow rate)
Δ(product flow rate) (%)
Δ(fresh methanol flow rate) (%)
Δ(recycle methanol flow rate) (%)
Δ(overall conversion) (%)
10.002
10.453
9.85
0
−10.005
−9.447
−10.22
0
0
0
0
∼0.1
5.021
4.63
4.87
∼0.1
−10.011
−11.74
−10.34
∼0.1
19.4
19.5
0
−30.2
0
20.023 −30.013
−29.74
uncertain. Also, a −10% change in kinetics, which could arise as a result of errors in the kinetics estimations, is considered. It can be seen from Table 1 that the change in WCO leads to a nearly proportionate variation in the recycle streams and product flow rates and that the overall conversion of WCO is almost unaffected. The performance of the designed PWC system in the face of the disturbances in Table 1 is presented and discussed in section 4. Level 2.2. Set Performance and Tuning Criteria. In this level, settling time is selected as the criterion for performance evaluation of the control system. As a large number of control loops are involved in this process, each controller has to be tuned once the control loops are established. Flow, level, and pressure controllers were tuned based on the guidelines given by Luyben.38 The remaining controllers were tuned using the built-in Autotuner in Aspen Plus Dynamics (APD). Controllers with time lags, such as temperature and composition control loops, were tuned using the closed-loop autotune variation method in APD.39 The Tyreus−Luyben (TL) method24,40 was used to determine the controller tuning parameters in such loops. The controllers without time lags were tuned using the open-loop tuning method in APD; the Cohen and Coon (CC) method was used to tune these control loops. Table S1 in the Supporting Information summarizes the tuning methods used for all control loops. Level 3.1. Production Rate Manipulator Selection. Selection of a suitable throughput manipulator (TPM) is critical for a process to respond to variations rapidly and smoothly. This step deals with the selection of a manipulated variable for varying throughput. Internal variables, such as reactor operating parameters on this path, are preferred over external variables (i.e., fixed flow followed by on-demand option). In a fixed-flow strategy, the feed flow rate is set and controlled at a particular value. On the contrary, in an ondemand strategy, the production rate is directly controlled at the desired value. The decision regarding the TPM can be made using a steady-state simulator. A process variable having a larger steady-state gain should be the natural choice. In the biodiesel process, the reactors have to be run under optimal conditions to maximize the product formation; these optimal conditions are fixed by optimization in this case. Therefore, 14411
dx.doi.org/10.1021/ie5023699 | Ind. Eng. Chem. Res. 2014, 53, 14408−14418
Industrial & Engineering Chemistry Research
Article
limit, which is given as the remote set point for the respective controller. Respective Ratios of Feeds. Fresh methanol is manipulated to maintain the desired ratio of methanol using controllers RC100 and RC200. The molar ratio of methanol to FFA is 10:1 in RFFA, and the methanol ratio to TG, DG, and MG is 6:1 in RTRANS1, RTRANS2, and RTRANS3, during normal operation. Similarly, ratio controllers RC101 and RC201 are also implemented to maintain the ratios of sodium hydroxide (1 wt %) and sulfuric acid (10 wt %), respectively. Methanol Split Ratios for the Continuous Stirred-Tank Reactors (CSTRs). Controllers SP200 and SP201 are employed to maintain the split fractions (RTRANS1/RTRANS2/ RTRANS3 = 0.77:0.12:0.11 under normal conditions). CSTR Temperatures. To maintain the WCO conversion, the reactor temperatures have to be maintained at their desired values. APD (V8.0) provides different options for reactor temperature control such as constant duty, constant temperature, condensing, evaporating, dynamics, and log mean temperature difference (LMTD). One of these options can be chosen depending on the application and the extent of realistic results sought from simulations. For example, the constant-duty model is the most idealistic, as it does not reflect a consideration of the heat-transfer area or temperature differential. In this study, the LMTD option was used to yield more realistic results. The overall heat-transfer coefficient (U) was assumed to be 0.4 kW/(m2 K) for both cooling water (at 30 °C for heat removal) and hot water (at 80 °C for heat addition). The heat capacity was assumed to be 4.2 kJ/(kg K) for both cooling water and hot water. The areas (A) for heat transfer, based on one-half of the lateral surface areas of the cylindrical reactors, were 24.07, 43.9, 40.98, and 40.8 m2 for RFFA, RTRANS1, RTRANS2, and RTRANS3, respectively. Therefore, their UA values were 9.63, 17.56, 16.39, and 16.31 kW/K, respectively. For the initial steady state, the heat duties found from the APD simulation were 0.32, 0.68, −0.97, and 0.64 GJ/h for RFFA, RTRANS1, RTRANS2, and RTRANS3, respectively. The corresponding utility outlet temperatures were 70, 65, 45, and 65 °C for these reactors. Accordingly, the required utility flow rate were 7619, 10794, 15397, and 10159 kg/h for RFFA, RTRANS1, RTRANS2, and RTRANS3, respectively. The effective relative gain array (ERGA),41 described in Appendix 2, was employed to determine the suitable controlled variable (CV)−manipulated variable (MV) pairings for the reactors and distillation columns. Xiong et al.41 presented a dynamic loop pairing criterion for decentralized control of multivariable processes, where the control loop pairing procedures of the relative gain array (RGA) were extended to develop the new method ERGA, which reflects dynamic loop interactions under finite bandwidth control. In addition to steady-state gain, ERGA also utilizes the bandwidth information on the process open-loop transfer function. Level 4.2. Selection of Manipulators for Less Severe Controlled Variables. Primarily, this step determines the level and pressure control loops. Levels along the primary process path should be stable and properly controlled, as they are often integrating. To achieve a self-consistent level control, the control of levels before the TPM should be in the direction opposite to the flow, whereas the level controls after the TPM should be in the direction of the flow. Therefore, in this study, the levels in the primary process path should be controlled in the direction of the flow to obtain a self-consistent control structure for the selected TPM fixed (feed) flow. Although a
these parameters will not be used as the TPM, and the feed (WCO) flow rate is identified as the next best choice for the TPM. Level 3.2. Product Quality Manipulator Selection. Selection of a manipulated variable for product purity is addressed in this level. Maintaining biodiesel purity is among important objectives when developing a control system, because the biodiesel product must meet EN/ASTM standards. Accordingly, biodiesel purity and impurity levels, such as the levels of TGs, DGs, and MGs in the final product, should be monitored and subsequently controlled. Therefore, proper reaction conditions should be maintained to convert WCO completely so that its components do not end up in the final product. The critical variables for this step include the methanol-to-oil molar ratio, reactor temperature, and catalyst. The desired ratio of methanol to oil was maintained using ratio controllers RC100 and RC200 through the cascade loop to maintain FFA and TG impurities , respectively, in the final product below the permissible limit (see Figure 3). The two inputs to RC100 are (1) the amount of FFAs and (2) the amount of methanol in the stream entering RFFA. The amount of FFAs remaining in the biodiesel determines the set point for RC100 (i.e., the set point for the molar ratio of methanol to FFAs will change when the FFA content in the final product violates the permissible limit). Accordingly, fresh methanol (stream MEOH) to the esterification section is manipulated by the RC100 controller. Similarly, the two inputs to RC200 are (1) the amount of TG and (2) the amount of methanol in the stream entering the splitter. The amount of TG remaining in the biodiesel determines the set point for RC200 (i.e., the set point for RC200 will change when the TG in the final product violates the permissible limit). Accordingly, fresh methanol (stream MEOH-2) is manipulated by the RC200 controller. Methanol impurity in the biodiesel product is controlled by manipulating the wash water flow rate (controller CC200 in Figure 3). Also, as glycerol is a valuable byproduct from the transesterification section in biodiesel production, its purity should be maintained at the desired value. A cascade loop is implemented to manipulate the reboiler duty of FRAC-4 to maintain the glycerol purity. Additionally, the control system is designed in such a way that the maximum temperature in FRAC-4 will not go beyond 150 °C to avoid glycerol decomposition. In the esterification section, glycrol (stream GLY-IN) is fed in small quantities to achieve the separation of unreacted sulfuric acid in the heavy phase from the phase separator W-1. Later, the glycerol is recovered through FRAC-1, R-CAO, S-1, and F-1, and it can be either recycled or mixed with the byproduct glycerol stream GLY-OUT. In this study, the recovered glycerol is not recycled because of its small quantity. The set point of controller FC101 (for the glycerol flow rate) is determined by the amount of sulfuric acid in the lighter phase. Level 4.1. Selection of Manipulators for More Severe Controlled Variables. In this level, process constraints related to process stability, equipment, operation, safety, and the environment are addressed. These are crucial to control as they can lead to severe operational ramifications. The important constraints in this process are discussed next. Column Temperatures, TFRAC‑1 and TFRAC‑4 ≤ 150 °C and TFRAC‑2 and TFRAC‑3 ≤ 250 °C. These temperatures are allowed to vary within acceptable limits. The controller becomes active when the controlled variable (i.e., temperature) violates the 14412
dx.doi.org/10.1021/ie5023699 | Ind. Eng. Chem. Res. 2014, 53, 14408−14418
Industrial & Engineering Chemistry Research
Article
be analyzed by preparing a “Downs Drill” table (see Table S2, Supporting Information). Negligible accumulation suggests that the inventory is well-regulated. Level 7. Effects of Integration. The dynamics of the process should be studied for the anticipated disturbances, both with and without recycles closed; see Murthy Konda et al.36 for more details on this step. This was done by observing the overall accumulation profile of WCO in the complete plant and the effects on important parameters such as conversion and production rate. Figure 2 shows that the accumulation profiles
proportional- (P-) only controller might suffice to obtain a satisfactory performance, a proportional−integral (PI) controller is implemented in some level control loops to attain the tight control. A tight level control in reactors is important as it can affect reaction rates.38 Also, a proper level control in phase separators and in columns is necessary. In all distillation columns, the levels in the reflux drum and column base are controlled using the distillate flow and bottoms flow, respectively. This is in accordance with heuristics that the level should be controlled so that the disturbances are directed away from the primary process path. Also, based on heuristics, the control of the level in the reflux drum using the reflux flow rate is not appropriate, as the columns in the biodiesel process are operating at smaller reflux ratios. Additionally, this is also in agreement with Richardson’s rule, which states that an inventory variable should be controlled with the manipulated variable that exerts the largest effect on it within that unit. The liquid levels in the respective reflux drums are controlled using distillate flow rates (valves V-3, V-4, V-29, and V-31 in Figure 3). In APD simulation, the reflux flow rate is kept constant by default. There is no strict need to install the reflux pump, valve, and flow controller for simulation purposes. In the real physical system, these would have to be installed. The reflux flow can be manipulated according to the requirements (e.g., composition control by manipulating the reflux ratio or fixing the feed-to-reflux ratio). Also, we did not encounter any difficulty in disturbance rejections with keeping the reflux flow rate constant. Therefore, other options such as controlling the reflux ratio or reflux-to-feed ratio, were not explored. Also, the top stream from all distillation columns was almost pure methanol, whose separation was easy because of the significantly larger difference in relative volatilities. The liquid levels in reactors and phase separators are controlled using liquid outlets, as shown in Figure 3. The pressures in all distillation columns were controlled using the respective condenser duties; these were verified using dynamic simulations. These pairings were also in agreement with the pairings resulting from the ERGA analysis, described in the previous section. The pressure in the flash evaporator was controlled by manipulating the overhead vapor flow, and the level was controlled using bottom liquid flow rate. Level 5. Control of Unit Operations. This step, in particular, deals with the control of individual unit operations. It should be done before testing component material balances, as some component inventory loops can be implicitly satisfied in this level. Basic control of the most common processes is wellestablished, as described by Luyben.38 In the previous section, all level and pressure control loops were already determined. Temperature control loops in CSTRs and distillation columns were also determined in level 4.2. In the biodiesel process, dual composition controls are not required in any column because the top product is methanol, which is reused in the process. Simulation results indicate that the unit operations are well regulated. The pH of the outlet stream of the neutralization reactor was controlled using inlet calcium oxide in R-CAO and inlet phosphoric acid in both R-CAT and R-CAT2. Level 6. Check Component Material Balances. It is necessary to ensure that the component inventory is wellregulated. This means that the accumulation of the components in the entire plant with the control structure developed so far should be zero or negligible. The overall accumulation of all components in the plant should be calculated and observed. If required, unit-wise accumulation can be determined. This can
Figure 2. Profiles of WCO accumulation with and without recycle due to (a) disturbance D6 and (b) disturbance D7.
of WCO with and without recycles closed did not deviate too much. From the dynamic simulations, no significant change was noticed in terms of the settling time of the biodiesel flow rate when the recycles were closed. This suggests that the plant dynamics were not significantly affected when the recycles were closed or opened. The conversion and product flow rate were found to be unaffected after the recycles were closed. This was mainly due to the parameters affecting these variables, such as temperature and methanol-to-oil ratio in the CSTRs, which were already set for control in the previous steps. Also, as found in level 2.1, the change in WCO led to proportionate variations in the recycle streams. Therefore, it can be safely assumed that the effects of integration were not severe and, hence, further modification was not necessary in the developed control system. Level 8. Enhance Control System Performance with Remaining CDOF. If required, the remaining CDOF can be further utilized to improve the performance of the developed control structure. For the complete biodiesel plant, the developed control system appeared to be satisfactory, so no modifications were warranted. The resulting control structure from the eight-level IFSH methodology used 52 of the available 67 CDOF, and it is shown in two plots for clarity: panels a and b of Figure 3 present the control structure for the esterification and transesterification sections, respectively. The percentage opening of control valves was set to about 50% for nominal operation. However, valve openings can marginally deviate from the set value, as was found by Zhang et al.26 In this process, the valve openings deviated between 49% and 51%. This was due to the pressure-driven simulation (i.e., the simulation model based on a pressure-flow solver, where pressure depends on upstream conditions) used in this study.
4. PERFORMANCE ASSESSMENT OF THE CONTROL SYSTEM The final control system for the complete biodiesel plant consists of 52 control loops (Figure 3). The details of the CV− MV pairings, the reasons for the pairings, and the tuning parameters are presented in Table S1 of the Supporting 14413
dx.doi.org/10.1021/ie5023699 | Ind. Eng. Chem. Res. 2014, 53, 14408−14418
Industrial & Engineering Chemistry Research
Article
Figure 3. PWC system designed for biodiesel production from WCO: (a) esterification section, (b) transesterification section. Note that reflux flow rates are fixed and their control is not shown.
processes. However, the main difference in the control structure aside from the inventory loops is the quality controllers, namely, FFA and triglyceride impurity, which
Information. This control structure cannot be compared directly with those of previous studies24,25 on the control of biodiesel processes, because they used different production 14414
dx.doi.org/10.1021/ie5023699 | Ind. Eng. Chem. Res. 2014, 53, 14408−14418
Industrial & Engineering Chemistry Research
Article
determine the remote set points of ratio controllers RC100 and RC200, respectively, and the methanol impurity in biodiesel product by manipulating the water flow rate. Previous studies24,26 on the PWC of biodiesel production used pure vegetable oil, which does not require an esterification section, whereas this section is needed when WCO is used as the feedstock to convert FFAs into biodiesel. Therefore, the resulting control structure in the present work is different (i.e., controllers in the top/esterification section and temperature controllers for RTRANS1, RTRANS2, and RTRANS3 in Figure 3). In this work, we included temperature controllers for the CSTRs as their temperatures were found to be varying. Other control loops in the transesterification section were similar to those presented by Zhang et al.26 Also, the components and kinetics used in the present work were more detailed than those used in earlier studies.24,26 Figure 4 depicts the transient profiles of the biodiesel production rate for disturbances D1, D2, D6, and D7. It can be
Table 2. Performance Characteristics of the PWC System Designed for the Complete Biodiesel Process performance criteria disturbance no.
settling time (h)
DPT (kg)
overall TV
D1 D2 D3 D4 D5 D6 D7
6.4 6.3 2.2 7.4 7.2 8.7 13.2
2123 2042 318 2223 2133 4073 7556
445 429 63 426 429 649 1104
WCO). This settling time was required for the PWC system because of the greater amount of time required in the phase separators and the larger holdup in the reactors. The DPT was the smallest for disturbance D3, wherein a −10% change was introduced in the pre-exponential factor for reactions converting WCO to biodiesel. The DPTs for disturbances D1−D5 were comparable (Table 2). As expected, the DPT was the largest for D7, followed by D6 because of the magnitude of the change in WCO. In general, an increase in the magnitude of the change in the throughput was found to lead to an increase in the DPT. Note that a smaller DPT is desired, which indicates that the PWC system is efficient in achieving the new production rate target, so that the economic losses are smaller. The overall TV in the manipulated variables was calculated by summing the TVs of all 52 individual controllers, to obtain the TV from a plantwide perspective. All of the manipulated variables were expressed in percentages; for example, the change in reboiler duty was taken as the percentage change in the reboiler duty (and not the absolute change in GJ/h). This is essential because computing the overall TV would be inconsistent if the manipulated inputs were expressed in their actual units; for example, flow rate has units of kg/h, whereas reboiler duty has units of GJ/h. As for the DPT results, D3 was found to have the lowest overall TV, whereas D7 had the largest. As in case of DPT, the increased magnitude of the change in throughput led to the increased overall TV (i.e., increased control effort). The profiles of WCO accumulation in the presence of disturbances D1, D2, D6, and D7 are illustrated in Figure 5. For
Figure 4. Transient profiles of the biodiesel production rate in the presence of selected disturbances.
seen that the plant settled down smoothly to a new throughput. By and large, the change in WCO produced a proportionate change in the biodiesel production rate. The performance of the designed control system was assessed for disturbances D1− D7. Because of the large number of controllers, it is extremely difficult to analyze the performance of each and every control loop (say, in terms of integral square error). Moreover, the PWC system performance should be analyzed from a plantwide perspective. The criteria for the quantification of the control system’s performance should be reliable and easy, and essentially, they should describe several factors such as an economic index, smoothness, and stability. Vasudevan and Rangaiah42 proposed several criteria for the performance assessment of PWC systems. This work used some of these factors, namely, settling time (an indicator of smooth and safe operation of the plant), DPT (an indirect economic index), and overall TV in the manipulated variables (an indicator of the control efforts required for the PWC structure to attain stable operation) for the performance evaluation. For this, the plant was run for 5 h, after which the disturbances were introduced, one at a time. Table 2 presents the performance of the control system for disturbances D1−D7. For the complete biodiesel plant, having a capacity of 120 kt/year, the observed settling times of the biodiesel production rate in the presence of disturbances D1− D6 were reasonable compared to the settling time of less than 10 h for a biodiesel plant processing 10 kmol of oil/h (i.e., nearly 71 kt/year).24 A larger settling time of about 14.2 h was observed for the relatively large disturbance D7 (−30% in the
Figure 5. Profiles of WCO accumulation due to disturbances D1, D2, D6, and D7.
brevity, the accumulation of only WCO is shown here. However, the accumulations of all components in the complete plant and in the individual unit operations can be monitored to investigate whether the inventory loop needs to be modified. Figure 5 shows that the accumulations were greater for larger throughput changes. The accumulations of WCO for all disturbances eventually reached zero, which is important for 14415
dx.doi.org/10.1021/ie5023699 | Ind. Eng. Chem. Res. 2014, 53, 14408−14418
Industrial & Engineering Chemistry Research
Article
Figure 7 shows the performance of important control loops for selected disturbances. Other control loops (not shown for
safe and stable operation of the plant. The accumulation for D7 (i.e., for a −30% change in WCO) was highest and required the most time (around 20 h) to settle down to zero. Profiles of the TG impurity in biodiesel, the methanol impurity in biodiesel, and the biodiesel purity (i.e., wt % of biodiesel in the product) resulting from the selected disturbances are presented in Figure 6. Both the TG and
Figure 6. Profiles of TG impurity (wt %) in biodiesel, methanol impurity (wt %) in biodiesel, and biodiesel purity (wt %) in the presence of selected disturbances.
methanol impurities were observed to be below their permissible limits (i.e., TG impurity < 0.2 wt % and methanol impurity < 0.2 wt %), in accordance with the EN standards.43 This confirms that the proposed control system was effective in maintaining the product quality. A small increase in TG impurity can be observed in Figure 6 for the increased flow rate of WCO for D1 and D6 and vice versa for D5 and D7. A similar trend was observed for the FFA and methanol impurities. Consequently, a reverse trend was observed for biodiesel purity, but it was also kept under control (i.e., >96.5 wt %) for all disturbances.43 Table 3 compares the biodiesel quality for the large disturbances D6 and D7 against European standard EN
Figure 7. Performance of important control loops for selected disturbances: (a) level control in RFFA, (b) level control in RTRANS1, (c) first-phase level control in WASH-2, (d) secondphase level control in WASH-2, (e) level control in RTRANS2, (f) temperature control in RTRANS1.
brevity) were also found to work equally as well as those presented in Figure 7. Because of the large number of control loops present in this PWC scheme, individual control loops were not assessed in terms of conventional time-domain performance specifications such as overshoot, rise time, and integral error indices. Instead, as discussed above, comprehensive PWC performance assessment criteria were used. In any case, the performance of important control loops in terms of overshoot and rise time can be seen in Figure 7. As the control loops were able to maintain the controlled variables at their respective set points for the disturbances tested, it can be inferred that the control loops were properly paired and tuned. The settling times of the control loops shown in Figure 7, including the level controllers in the phase separators and reactor, justify the overall settling times of the biodiesel product given in Table 2. The observed settling times in the reactors were due to the larger holdup. Overall, the control loops in the developed PWC system performed satisfactorily in the presence of several disturbances.
Table 3. Comparison of Biodiesel Quality for D6 and D7 against European Standard EN 14214 parameter
EN 1421443
D6
D7
biodiesel (wt %) monoglycerides (wt %) diglycerides (wt %) triglycerides (wt %) glycerol (wt %) methanol (wt %) density (15 °C) (kg/m3)
>96.5
