Initial Validation of a Reverse Osmosis Simulator Luis Palacin, Fernando Tadeo, Johanna Salazar, Cesar de Prada University of Valladolid, Dpt. of Systems Engineering and Automatic Control, 47011 Valladolid, Spain
[email protected] Abstract Reverse Osmosis (RO) is the most common technique to produce drinkable water in arid regions. The desalination plants are usually designed for a short number of constant operation points. A dynamic strategy of control would help to decrease the operation expense and the equipments cost. However, the lack of dynamic simulation tools for this kind of plants, avoids the optimal design, which would use a continuously changeable operation point. For this, a new dynamic simulation library of RO plants was presented elsewhere. Now, the present paper deals with the validation of that library, using experimental data from a real RO plant.
1. Introduction Reverse Osmosis (RO) is known to be an effective technique in arid regions to produce drinkable water from brackish and sea water, [1]. This is because RO plants need less energy, investment cost, space and maintenance than other alternative desalination processes, [2], hence, it is the preferred desalination technique worldwide, [3], [4]. Typically, RO plants are operated at a constant operation point, producing a constant flow of drinkable water. However, the water consumption suffers strong variations during a day, especially if the desalination plant fulfils water to a small town. The variable difference between the water production and the water demand is compensated by big storage tanks at the end of the production line. The size of these tanks would be easily decrease if a changeable control strategy would be taken into account. Besides, the changeable control strategy would decrease too the cost of the energy consumption, because the variations in the prize of electricity during a day. In the case of RO plants, where the energy is fulfilled by solar panels or wind turbines, the changeable control strategy would decrease the size of the batteries, with the following improve of the operation expense. The most common reason to not operate the plant using a changeable control strategy is because only static simulation tools
978-1-4244-6850-8/10/$26.00 ©2010 IEEE
where used during the design stage. In the following years, with the increase in the use of dynamic simulators, the strategy control of RO plants would be strongly improved. In fact, a better operation strategy of a RO plant was shown in [5-7], where the changeable water production was varied along the day, adjusting production and demand, and taking into account the optimization of energy consumption. Besides, a Model Predictive Control was tested for a simulated plant in [8]. Finally, an integrated design of a RO desalination plant was carried out in [9], with an important decrease in capital expenditure. For this, a dynamic simulator of RO plants was needed. Thus, in [10] the dynamic library of RO plants was presented, used in the integrated design and in the predictive control. A detailed description of the math model can be found in [11], and the upgrade of the library in [12]. More examples of advanced control strategies in desalination plants, which require good dynamic simulation tools, can be seen in [13-16]. This paper deals with the validation of this dynamic library. Initially, the values of the different parameters of the math models had been taken from the specialized bibliography [10]. Now, most of these parameters are estimated using experiments in a real RO plant.
2. Reverse Osmosis Plant The RO plant used for validation of the Reverse Osmosis Library, is shown in Fig. 1, and a simplified diagram of the plant is shown in Fig. 2. First, a pump (B1) pumps brackish water from a well to the supply tank (T1). From this tank, water is pumped to a set of filters and chemical additions. The high pressure pump (B2) increases its pressure to a value above the osmotic pressure so that the pressurized water can pass then through the RO membrane rack. The difference in pressure between each side of the membranes produces a flow of clean water through the membranes. Finally, this clean water is stored in other tank (T2), which supplies water to the consumers. Reverse osmosis is a separation process that uses high pressure to force a solvent (water) through a semi permeable membrane that retains the solute (salt) on one side. The clean water flow is called “permeate” and needs a remineralization before consumed, and the rejected water flow is called “retentate” or “concentrate”. Typical values of permeate flow are 45% of the feed flow for sea water, and until 80% for brackish water. An in-depth description of the components of an RO plant could be seen in [17].
numerical methods capable of processing complex systems represented by DAE equations and discrete events. Moreover, free versions are available for research and teaching (http://www.ecosimpro.com).
supply tank (T1)
P1
supply pump (B1) hypochlorite sodium well / sea pH control
sequestrant
bisulphite
Figure 1. Pilot desalination plant. The RO plant which is utilized in this paper has been design to produce 800 L/h of permeate water from a flow of 2000 L/h of brackish water. A central problem during operation is the decrease in performance of the membranes, due to deposits (silt, scale, organic components, etc). Thus, in order to prevent precipitation and eliminate microorganisms, a pretreatment is needed. It normally consists of filtration and the addition of chemical products. In addition, periodic cleanings are scheduled to reduce the amount of deposits. The final part of the process is a post-treatment, to make the water drinkable, by the addition of chlorine and a remineralization. The RO process requires a high pressure on the feed side of the membrane: close to 30bar bar for the plant utilized in this validation (and up to 80 bar for extreme cases with very salty seawater). This pressure is generated by the high pressure pump (B2), (that, in this case, it is a positive displacement pump), that consumes most of the electrical energy needed by the plant. One typical strategy of energy recovery, that is commonly utilized in desalination plants several times bigger than the plant in hand, is the use of pressure chambers at the outlet of the concentrate flow, which recover most of the pressure of the concentrate flow.
3. Dynamic Library of RO Plants The dynamic library of RO plants has been developed in the simulation environment EcosimPro©. EcosimPro© is a powerful modeling and simulation tool that follows advanced methodology for modeling and dynamic simulation. It provides an object oriented and non causal approach that allows creating new simulations interconnecting reusable component libraries. EcosimPro© is based on very powerful symbolic and
sand filter
P2
cartridge filter
Q2 Q1
P3
RO membranes
P4 high pressure pump (B2)
Q3 hypochlorite sodium
pH control
outlet tank (T2)
water demand
Figure 2. RO diagram. The dynamic library of RO plants is based on a set of components representing the different units of a RO plant. As sand filters, cartridge filters, different types of pumps, RO membranes, storage tanks, exchanger energy recoveries, valves, control systems (such as PIDs or PLCs), etc. Every component has been modeled using first principles and correlations from literature: Mass balances, energy balances, and physic-chemist equations are in the core of the models. Complex elements, such as sand filters or RO membranes, have been discretized by using the finite volume method. Fig. 3 corresponds to the graphical simulator of the real plant (shown in Fig. 1.), than is validated in the present paper. The simulator consists of two parallel membranes, one high pressure positive displacement
Figure 3. Graphical simulator of the pilot plant, made with the dynamic library. pump, one sand filter, one cartridge filter, the addition of several chemical components (hypochlorite, bisulphite, sequestrant, chloride acid, etc.), supply centrifugal pumps and storage tanks.
4. Validation of the Simulator of the Real Plant Although the evolution of all important variables in RO plants (flows, pressure, salt concentration, solids concentration, pH, temperature, etc.) is calculated by the dynamic library, the initial validation shown in this paper is done only for the hydraulic part of the library: flow and pressure. The pilot plant (figure 1), which is used in the parameter estimation and validation of the simulation, was specially designed for testing and experimentation and has a huge number of sensors, bigger than in a typical RO plant. All the main parameters in the RO plant can be measured in different points of the plant. Sensors are managed from an OMRON PLC, and the data loading is carried out by the use of OPC protocol. In order to realize a good parameter estimation, several experiments were carried out changing the operation pressure and the relation between the flows of permeate and concentrate. For the calculation of the variables of the system, the simulator of the pilot plant needs to know the value of some variables, or boundary conditions. The typical
boundary conditions are the input and output conditions of the system. In particularly, the simulator requires to know the value of the inlet pressure of the system (or the feed flow), the outlet pressure of the system, and the setpoint of the different PID controllers (or the manipulated variables of the PID controllers). In this case, the value of the inlet pressure (P1 in figures 2 and 3) was supplied to the simulator. Figure 4 shows the value of this variable during one experiment. The value of the outlet pressure is constant and equal to the atmospheric pressure, and the rest of boundary conditions are directly known because the own experimentation.
Figure 4. Pressure of the inlet of the system(P1).
Using these boundary variables, the simulator is able to predict the rest of variables along the system. In particularly, the calculated variables are the pressure in the outlet of the sand filter (P2), the pressure in the outlet of the cartridge filter (P3), the pressure in the outlet of the high pressure pump (P4), the feed flow (Q1), the permeate flow (Q2) and the retentate flow (Q3). The validation of the simulator is done by the minimization of a certain objective function that penalizes the different between the value of the calculated variables and the measured variables. The minimization can be easily solved using any non linear optimization algorithm, such as Sequential Quadratic Programming (SQP), which is available in the own EcosimPro©. Figures 5 to 11 show the value of the calculated values of the different variables (pressure and flow) during one experiment, and the measured values. Notice how both curves flow together in each figure. In particular, figure 5 shows the pressure after the sand filter (P2). This pressure is calculated by an empirical model that relates the pressure drop along the filter and depends basically on the inlet flow (Q1). Besides, the pressure drop in the cartridge follows approximately a mathematical equation (1), which is described deeper in [11].
J=
ε 3ΔP , 2 KH m μSi2 (1 − ε )
Figure 6. Pressure of the outlet of the cartridge filter (P3).
(1)
Figure 7. Pressure of the outlet of the high pressure pump (P4).
Figure 5. Pressure of the outlet of the sand filter (P2). Figure 6 shows the outlet pressure of the cartridge filter, calculated by the equation (1) and using as input pressure the value estimated in figure 5. The small errors in figure 5 star to increase and pile up along the system. Figure 7 shows the pressure after the high pressure pump (P4). This pressure is calculated as the result of the input pressure (figure 6) plus the increase of pressure supplied by the pump. This pressure
increase is easily calculated by the curve of the pump. The pump of the pilot plant is a positive displacement pump, and the supplied increase of pressure depends polynomially of the velocity of the pump. Figure 8 shows the relation between the water flow and the velocity of the pump. Notice how the measured data follows a polynomial curve. In this case, the average error between measured and estimated data is less than 5%. Next, figure 9 shows the feed flow (Q1) during the experiment. The feed flow is calculated by the energy balance between the inlet of the sand filter and the outlet of the high pressure pump. Experiments in the pilot plants help to estimate the pressure drop term in the energy balance, using a semi empirical equation. The permeate flow (Q2) is calculated using the equation (2), the permeability of the membranes [11]. Figure 10 shows the evolution during the experiment of this variable. Finally, the retentate flow (Q3) can be estimated as the difference between Q1 and Q2, and it is shown in figure 11.
J w = A (Δp − Δπ ),
(2)
hydraulic part of the simulator are presented, confirming the validity of this part of the dynamic library.
Figure 8. High pressure pump curve.
Figure 11. Retentate flow (Q3). Acknowledgements The authors wish to express their gratitude to the Spanish MICINN for its support through project DPI2006-13593, as well as the EU for its support through Sixth Framework Program (Reference FP62004-INCO-MPC-3), and SETA S.L. and the rest of the groups working in the OPEN-GAIN project for feedback and comments. References Figure 9. Feed flow (Q1)
Figure 10. Permeate flow (Q2) Conclusions The importance of dynamic simulation tools for the design of RO plants has been discussed. In order to improve the design and control of this kind of plants, a dynamic library of RO plants previously developed, has been checked, and initial results for a preliminary validation based on experiments carried out on the real plant are presented here. More precisely, tests of the
[1] C. Fritzmann, J. Löwenberg, T. Wintgens, and T. Melin,“State-of-the-art of Reverse Osmosis Desalination”, Desalination, Vol. 216, pp. 1-76, 2007. [2] A. Gambier, A. Krasnik, and E. Badreddin, “Modeling of a Simple Reverse Osmosis Desalination Plant for Advanced Control Purposes”, American Control Conference, New York, USA, 2007. [3] R. Baker, “Membrane Technology and Applications”, Wiley, West Sussex, England, UK, 2004. [4] M. Wilf, “The Guidebook to Membrane Desalination Technology”, Balaban Desalination Publication, L'Aquila, Italy, 2007. [5] J. Salazar, F. Tadeo, and C. de Prada, “Control System for Reverse Osmosis Plants and Renewable Energy Generation”, International Conference on Renewable Energies and Power Quality, Granada, Spain, 2010. [6] L. Palacin, F. Tadeo, and C. de Prada, “Operation of Desalination Plants Using Hybrid Control”, 2nd Maghreb Conference on Desalination and Water Treatment, Hammamet, Tunisia, 2009.
[7] L. Palacin, C. de Prada, F. Tadeo, and J. Salazar, “Operation of Medium-size Reverse Osmosis Plants with Optimal Energy Consumption”, IFAC International Symposium on Dynamics and Control of Process Systems, Leuven, Belgium, 2010. [8] D. Sarabia, F. Capraro, L.F.S. Larsen and C. de Prada, “Hybrid NMPC of supermarket display cases”, Control Engineering Practice, Vol. 17, pp. 428-441, 2009. [9] L. Palacin, F. Tadeo, and C. de Prada, “Integrated Design Using Dynamic Simulation of Reverse Osmosis Plants”, International Conference on Industrial Engineering and Engineering Management, Hong Kong, China, 2009. [10] L. Palacin, C. de Prada, S. Syafiie, and F. Tadeo, “Library for Dynamic Simulation of Reverse Osmosis Plants”, 20th European Modeling and Simulation Symposium, Briatico, Italy, 2008.7 [11] S. Syafiie, L. Palacin, C. de Prada, and F. Tadeo, “Membrane Modeling for Simulation and Control of Reverse Osmosis in Desalination Plants”, Control 2008, Manchester, UK, 2008. [12] L. Palacin, F. Tadeo, H. Elfil, and C. de Prada “New Dynamic Library of Reverse Osmosis Plants with Fault Simulations”, 2nd Maghreb Conference on Desalination and Water Treatment, Hammamet, Tunisia, 2009. [13] I. Alatiqi, H. Ettouney, and H. El-Dessouky, “A Short-term, Pattern-based Model for Water Demand Forecasting”, Journal of Hydroinformatics, Vol. 9, pp. 39-50, 2007. [14] A. Gambier, and E. Badreddin, “Dynamic Modelling of MSF Plants for Automatic Control and Simulation Purposes: a Survey”, Desalination, Vol. 166, pp. 191-204, 2004. [15] G. Bacelli, J.C. Gilloteaux, and J. Ringwood, “A Predictive Controller for a Heaving Buoy Producing Potable Water”, European Control Conference 2009, Budapest, Hungary, 2009. [16] A. Zafra-Cabeza, M.A. Ridao, and E.F. Camacho, “Optimization in Reverse Osmosis Plants through Risk Mitigation”, European Control Conference 2009, Budapest, Hungary, 2009. [17] N. M. Al-Bastaki, and A. Abbas, “Modeling an Industrial Reverse Osmosis Unit”, Desalination, Vol. 126, pp. 33-39, 1999.
