Вы находитесь на странице: 1из 11

Computers and Chemical Engineering 34 (2010) 620630

Contents lists available at ScienceDirect

Computers and Chemical Engineering


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

Planning and scheduling of salt harvest in solar evaporation ponds


Karla Ossandn a , Patricio Pinto b , Luis A. Cisternas a,c,
a

Departamento de Ingenieria Quimica, Universidad de Antofagasta, Casilla 170, Antofagasta, Chile SKM, Santiago, Chile c Centro de Investigacin Cientco y Tecnolgico para la Minera, CICITEM, Chile
b

a r t i c l e

i n f o

a b s t r a c t
Several chemicals are produced from brines by solar crystallization using solar ponds. The process of salt harvest consists on mechanically retiring the salts precipitated in the solar evaporation ponds and to leave them in their respective stockpile. In an industrial operation several ponds are used for the salt crystallization, and hence the harvest planning can be a nontrivial task. Therefore, the objective of this work is to plan the feeding ow to each of the solar ponds, the manipulation of brine solution between ponds, brine and solids inventories in each pond that maximizes the production and the harvest periods. All this having as input data the evaporation rate, feed concentration, pond concentration range, and the operational initial conditions. The presented model corresponds to a MINLP, which includes the mass balances in each pond, equilibrium conditions, and planning and operational restrictions. The problem was solved in two steps: rstly, the maximization of the salt harvest was determined and then, using this maximum harvest, the maximum availability of the contractor was determined. Several cases have been studied, including: ternary (NaNO3 KNO3 H2 O) and quaternary systems (KClKNO3 K2 SO4 H2 O), pond systems with 3 and 4 ponds, and considering 12 and 26 operation periods per year. 2009 Elsevier Ltd. All rights reserved.

Article history: Received 26 March 2009 Received in revised form 24 September 2009 Accepted 29 September 2009 Available online 7 October 2009 Keywords: Solar evaporation ponds Crystallization Solar salt harvest

1. Introduction The solar evaporation ponds are used for producing salts or generating heat energy (Lior & Bakish, 2001). The solutions processed in ponds for the production of salts are obtained from different origins like seawater, natural lakes, underground brines and mining solutions. The products obtained from brines include KCl, K2 SO4 , Na2 SO4 , Li2 SO4 , and H3 BO4 (Butts, 2001; Flotz, 1993). A list of salts that are obtained by solar evaporation are shown in Table 1. The operation of solar pond is favourable if a big area and arid climate are available, reason why the production capacity and efciency depends on the location of the ponds. At the present time, solar evaporation pond systems are designed and operated by combining art and science. However, due to the increasing demand of some salts, space limitations and increased operation and capital costs, a more scientic focus for the design and operation of solar ponds is being required. For example, a thermodynamic model has been developed to study the behaviour of solid liquid equilibrium in aqueous electrolyte systems (Kwok, Ng, Taboada, & Cisternas, 2008; Song & Yan, 2003). A model to optimize the process has been created together with experimental simulations that determine the best pond height

Corresponding author at: Departamento de Ingenieria Quimica, Universidad de Antofagasta, Casilla 170, Antofagasta, Chile. Tel.: +56 55 637323. E-mail address: lcisternas@uantof.cl (L.A. Cisternas). 0098-1354/$ see front matter 2009 Elsevier Ltd. All rights reserved. doi:10.1016/j.compchemeng.2009.09.008

depending on the climatic conditions (Murthy, Ramakrishna, & Pandey, 2003), and a model to determine the optimum conditions for the solar evaporation of Dead Sea brines (Bloch, Farkas, & Spiegler, 1951) has been developed. Also a methodology has been developed based on experimental data to increase the lithium production from residual brine (Hamzaoui, Mnif, Hammi, & Rokbani, 2003). Efforts to develop new ways to graphically represent the phase behaviour have been undertaken and applied to simulate brine evaporation, and to foresee the solid phases deposed during the evaporation process (Hammi, Musso, Mnif, & Rokbani, 2003a, 2003b, 2004; Musso, 2003). The use of hyperplanes to simulate the equilibrium conditions in multicomponent systems were analyzed by Pressly and Ng (1999), and then Cistermas, Cangana, Aravena, and Vargas (2007) have demonstrated that their approach is sufcient to simulate solar evaporation ponds. However, few efforts have been carried out to improve planning (what has to be done) and scheduling (when this has to be done) of salt harvest in solar pond systems to maximize the salt production and minimize the cost. The objective of this work consists of determining by means of planning and scheduling, the maximum precipitation of salts in solar pond systems, and to distribute its harvest in such a way to leverage the maximum use of the contractor in a year period horizon. With this objective, a mathematical model is developed whose results indicate the planning and scheduling of brine ows between ponds, the brine and solid inventory in each particular pond and when ponds harvest must be done. All of this having as data the

K. Ossandn et al. / Computers and Chemical Engineering 34 (2010) 620630

621

Nomenclature Sets I K L LI(i) N T {i/i I, is an evaporation solar pond} {k/k K, is a chemical specie} {l/l L, is a precipitated salt} {l/l L, is a precipitated salt in pond i, i I} {n/n N, is an hyperplane equation} {t/t T, is an operation period}

Binary variables F yi,j,t equal to 1 if there is owrate from pond i to pond j, and 0 if there is no owrate from pond i to pond j P yi,t equal to 1 if the pond i is in operation in the period t, and takes the value 0 if the pond i is in harvest in the period t

HislU HTiU MWk U an,i,k bn,i L xi,t,k


U xi,t,k

Parameters Ai area of the pond i CM maximum production of salts obtained from the rst objective function Ei,t volumetric rate of evaporation per unit area of pond i in period t Us total mass owrate upper bound for the output FTi,t stream of pond i in period t HisU upper bound for the solid precipitate height in pond i HislL lower bound for the brine height in pond i upper bound for the brine height in pond i upper bound for total height for the pond i molecular weight of the specie k big M parameter coefcient of the hyperplane equation n in pond i for the specie k coefcient of the hyperplane equation n in pond i lower bound for the mass fraction of each specie k in the pond i and period t upper bound for the mass fraction of each specie k in the pond i and period t factor which adjusts production to a minimum value in Eq. (33) water density solid density in pond i brine density in the pond i stoichiometric coefcient of specie k in dissolution of salt l

Table 1 Salts obtained in solar evaporation ponds. Mineral name Anhydrite Arcanite Bischote Carnalite Darapskite Epsomite Glaserite Halite Kainite Kieserite Leonite Mirabilite Niter Nitratine Schoenite Silvinite Silvite Tenardite Molecular formula CaSO4 K2 SO4 MgCl2 6H2 O MgCl2 KCl6H2 O NaNO3 Na2 SO4 H2 O MgSO4 7H2 O 3K2 SO4 Na2 SO4 NaCl KClMgSO4 3H2 O MgSO4 H2 O K2 SO4 MgSO4 4H2 O Na2 SO4 10H2 O KNO3 NaNO3 K2 SO4 MgSO4 6H2 O KCl + NaCl KCl Na2 SO4

evaporation rate, concentration operation range, concentration of the feed and the initial pond conditions. 2. Mathematical model In this section, a MINLP model to optimize the planning and scheduling of the salt production and harvest in solar evaporation pond systems is presented. The problem consists of determining a maximum production of precipitated salts in solar evaporation ponds and to distribute their harvest in a such way of leveraging the maximum use for the contractor carrying out the task. From this viewpoint, the problem has two objectives which are solved in a hierarchical manner. It also looked for determining the distribution of ow transfers among ponds. The solution strategy consists of using mathematical programming based on a superstructure of the solar evaporation process. The superstructure of the solar evaporation pond system is built by including a mixer in the input of each pond and a divider in the output as shown in Fig. 1, and by allowing the transfer of brines among all the ponds or among those the designer wants to consider. This means that all ponds could be connected. Each brine pond can be sent to a single pond, and therefore stream division is not allowed, however a pond may receive brine of several ponds. 2.1. Pond modeling The pond model should include the mass balances and the solidliquid equilibrium conditions between the precipitated solids and the brine. To develop this model, each pond is represented as shown in Fig. 2. Then, mass balances for each component was developed for the liquid phase and the solid phase separately

H2 O s i sl i k,l

Positive continuous variables CSi,t,k salt harvest in the pond i in the time period t for the component k mass owrate of specie k from pond i to pond j in Fi,j,t,k the period t A Fi,t,k k specie mass owrate of feed brine for the pond i in period t e Fi,t,k input mass owrate in the pond i in time period t for the specie k s output mass owrate in the pond i in time period t Fi,t,k for the specie k A FTi,t total mass owrate of feed stream to pond i in period t s FTi,t total brine mass owrate for the pond i in period t s Ii,t,k pond i solid inventory in period t for the specie k.
sl Ii,t,k sl ITi,t ri,t,l

pond i brine inventory in period t for the specie k total brine inventory for the pond i in period t molar rate at which the salt l precipitates for the pond i and period t mass fraction of feed brine for the specie k mass fraction of specie k in pond i and period t

A xk xi,t,k

Fig. 1. Solar evaporation pond representation in the superstructure.

622

K. Ossandn et al. / Computers and Chemical Engineering 34 (2010) 620630

where MWk is the molecular weight of the specie k, k,l is the stoichiometric coefcient of specie k in dissolution of salt l, ri,t,l is the molar rate at which the salt l precipitates. The mass balance for species k in the solid phase or precipitate when a pond is operating, is given by the following expression: MWk
l LI(i) k,l ri,t,l s s + Ii,(t1),k = Ii,t,k

i I, t T, k K.

(4)

Eq. (4) shows that precipitation of the species k in all the salts l s precitated plus the initial solid inventory (Ii,(t1),k ) must be equal s to the nal solid inventory (Ii,t,k ) for each period t in pond i. When the pond is harvested, it is necessary to remove the entire inventory of brine in the preceding period, then the brine mass ow output must be equal to the inventory of the previous period, i.e.,
Fig. 2. Solar evaporation pond representation for modeling.
sl s Ii,(t1),k = Fi,t,k

i I, t T, k K.

(5)

considering only the ow of salt precipitated as a mass transfer from one phase to another. In order to simplify the problem, losses by ltration and/or occlusion were not considered in the mass balance. However, these phenomena can be easily included. Each pond has two possible states in every period: operation or harvest, then its modeling is carried out using a disjunction in the following way: Operation f (x) = 0 Ax + By C Harvest g(x) = 0 Dx + Ey F . (1)

Now, the inventory of solid precipitated in the previous period should be harvested, then the mass balance for the solid when the pond is harvested is the following:
s Ii,(t1),k = CSi,t,k

i I, t T, k K,

(6)

where CSi,t,k represents the salt harvest in the pond i in the time period t for the component k. For the brine output mass ow rate and brine inventory of each pond, there are relationships between the total mass and the mass of each specie, either when the pond is in operation or in harvest, which are shown in Eqs. (7)(10):
s FTi,t = kK sl ITi,t = kK s s Fi,t,k = xi,t,k FTi,t sl sl Ii,t,k = xi,t,k ITi,t sl Ii,t,k s Fi,t,k

When the system is operating there is no harvest, and therefore the solid phase increases its inventory of salts, whereas the liquid phase may or may not receive or send brine to other ponds, which is why its brine inventory can increase or decrease. Moreover, when a pond is under harvest, the entire inventory of brine is sent to another pond, brine input is removed, and the salt inventory reduced to zero. Equations in the disjunction (1) represent the mass balances, equilibrium conditions, and restrictions of operation conditions, scheduling and planning, as explained below. The water mass balance of a pond in operation is given by
e Fi,t,H
2

i I, t T, i I, t T, i I, t T, k K, i I, t T, k K,

(7)

(8) (9) (10)

where

sl + Ii,(t1),H O

2O

s = Fi,t,H

+ Ei,t Ai O MWH2 O

H2 O

sl + Ii,t,H

2O

+
e where Fi,t,H sl Ii,t,H s and Fi,t,H

H2 O,l ri,t,l

i I, t T,

(2)

l LI(i)
2O 2O

represent the input and output mass


2O

sl ow rate of water in the pond i in period t, while Ii,(t1),H


2O

and

represent the water inventory in pond i, in periods (t 1)

and t respectively. The evaporation is represented by the expression Ei,t Ai H2 O , where Ei,t is volumetric rate of evaporation per unit area of pond i in period t, Ai is the area of the pond i, and H2 O is the water density. Molecular water presents in hydrated salts is expressed as MWH2 O H2 O,l ri,t,l , where MWH2 O is the molecl LI(i) ular weight of the water, H2 O,l is the stoichiometric coefcient of the water in dissolution of salt l, ri,t,l is the molar rate at which the salt l precipitates. The mass balance for the other species present in the brine of pond i in period t is given by
e sl s sl Fi,t,k + Ii,(t1),k = Fi,t,k + Ii,t,k +

MWk
l LI(i)

k,l ri,t,l

i I, t T, k K.
(3)

e s sl sl Fi,t,k , Fi,t,k , Ii,(t1),k , and Ii,t,k have similar meaning as in Eq. (2) but now for specie k. Of course, there is not evaporation for the other species. Salt precipitation is expressed as MWk k,l ri,t,l , l LI

sl t, ITi,t is the total brine inventory for the pond i in period t, and xi,t,k is the mass fraction of specie k in pond i and period t. The representation of the phase equilibrium behaviour in solar evaporation ponds can be a complex task because: (1) the brine and precipitated solids are not in equilibrium since the latter are not in contact with the brine, this is mainly due to the absence of mixing in the pond. (2) Evaporation occurs on the surface of the brine, it is there where supersaturation and precipitation of salts occurs. Precipitated salts sediment forming the solid phase at the bottom of the pond. Crystals are born on the pond surface and grow as sediment. (3) The pond does not work isothermally, with a temperature and concentration prole from surface to bottom of the brine. (4) Ambient temperature and evaporation rate change during the day. (5) Brines are multicomponent solutions usually with some species not found in databases of thermodynamic models. The phase equilibrium modeling with models such as UNIQUAC will increase the complexity of the mathematical model, but it does not offer a very accurate representation of the phase phenomenon in the pond. An alternative is to represent the equilibrium of the pond by using hyperplanes. Pressly and Ng (1999) proposed the use of hyperplanes to represent the phase equilibrium. Cistermas, Cangana, Aravena, and Vargas (2007) showed that hyperplanes can be a good approach to represent the equilibrium conditions for the modeling of solar evaporation ponds. The hyperplane is represented by a system of linear equations, which depends on the number of salts that precipitate in the pond,

s FTi,t

is the total brine mass ow rate for the pond i in period

K. Ossandn et al. / Computers and Chemical Engineering 34 (2010) 620630

623

i.e., an,i,k xi,t,k = bn,i


kK

n N, i I, t T,

(11)

The inequality shown below indicates that the volume of brine inventory should be between the maximum and minimum brine volumes allowed in the pond, which are given by the product between the pond area and brine height allowed. This is, Ai HislL
sl ITi,t sl i

where an,i,k and bn,i are constant parameters. Eq. (11) represents the isothermal equilibrium conditions for the pond i. In addition the sum of the species mass fraction is included in the model by the following expression: xi,t,k = 1 i I, t T.
kK

Ai HislU

i I, t T,

(15)

(12)

In practice, the solar evaporation ponds operate in ranges of concentrations to control the precipitation of salts, so it is necessary to dene limits on the concentrations of each component for the operation and harvest:
L U xi,t,k xi,t,k xi,t,k

sl height and i is the brine density in the pond i. In the same way, the volume of solid inventory in the pond must be limited so that the maximum height of the solid inventory does not exceed the permitted limit:

where HislL and HislU are the lower and upper bounds for the brine

i I, t T, k K,

(13)

Is k K i,t,k s i

Ai HisU

i I, t T,

(16)

U L where xi,t,k and xi,t,k are the lower and upper bounds for the mass fraction of each specie k, in the pond i and period t. The transfer of brine between ponds is limited either because the transfer is by gravity or the capabilities of the pumps: s Us FTi,t FTi,t Us FTi,t

where, HisU is the upper bound for the solid precipitate height and s is the solid density in pond i. i Also, the pond i has a maximum total height, HTiU , then, I sl k K i,t,k sl i + Is k K i,t,k s i Ai HTiU

i I, t T,

(14)

is the total mass ow rate upper bound for the output where stream of pond i in period t. Inventories of brine and solid must be bounded within the maximum and minimum heights allowed in the ponds. Therefore, it is necessary to include inequalities that limit the volume of solids in the brine ponds.

i I, t T,

(17)

where HTiU is the upper bound for total height for the pond i. Based on the restrictions described above, the modeling disjunctions for the pond, either in operational or harvest status, are given by

e sl s sl F MWH2 O i,t,H2 O + Ii,(t1),H2 O = Fi,t,H2 O + Ei,t Ai H2 O + Ii,t,H2 O l LI(i) e sl s sl Fi,t,k + Ii,(t1),k = Fi,t,k + Ii,t,k + MWk k,l ri,t,l l LI(i) s s MWk k,l ri,t,l + Ii,(t1),k = Ii,t,k l LI(i) s s FTi,t = Fi,t,k kK sl sl ITi,t = Ii,t,k kK s s Fi,t,k = xi,t,k FTi,t sl sl Ii,t,k = xi,t,k ITi,t an,i,k xi,t,k = bn,i kK xi,t,k = 1 kK U L xi,t,k xi,t,k xi,t,k s Us FTi,t FTi,t sl ITi,t Ai HislL Ai HislU sl i Is k K i,t,k Ai HisU s i I sl Is k K i,t,k k K i,t,k U
i I, t T, k K.
sl i

P yi,t

H2 O,l ri,t,l

s i

Ai HTi

P yi,t s I sl i,(t1),k = Fi,t,k s Ii,(t1),k = CSi,t,k s FT s = Fi,t,k i,t k sl sl = ITi,t Ii,t,k k s s F = xi,t,k FT i,t i,t,k sl sl Ii,t,k = xi,t,k ITi,t an,i,k xi,t,k = bn,i k xi,t,k = 1 k L U xi,t,k xi,t,k xi,t,k Us s FTi,t FTi,t

(18)

624

K. Ossandn et al. / Computers and Chemical Engineering 34 (2010) 620630

The disjunction of the above equation can be written in terms of binary variables using the big M method. This is,
e sl s Fi,t,k + Ii,(t1),k = Fi,t,k + Ei,t Ai P H2 O yi,t sl + Ii,t,k + l LI(i) e sl s sl Fi,t,k + Ii,(t1),k = Fi,t,k + Ii,t,k + l LI(i)

MWk

k,l ri,t,l

i I, t T, k K

Mixers at the entrance of each pond are modeled by considering that mass ow into a pond must be equal to the sum of mass ows from other ponds and the brine feeding to the pond system:
e Fi,t,k = j I,i = j / e where Fi,t,k is the input mass ow rate in the pond i in time period A Fi,j,t,k + Fi,t,k

MWk

k,l ri,t,l

i I, t T, k K

i I, t T, k K,

(22)

MWk
l LI(i) s FTi,t = kK sl ITi,t = kK

k,l ri,t,l

s Ii,(t1),k

s = Ii,t,k + CSi,t,k

i I, t T, k K

s Fi,t,k

i I, t T i I, t T i I, t T, k K i I, t T, k K n N, i I, t T

A t for the specie k and Fi,t,k is the k specie mass ow rate of brine feeding for the pond i in period t. Similar expressions to Eqs. (7)(10) are required for the feed stream: A FTi,t = kK A A A Fi,t,k = xk FTi,t A Fi,t,k

sl Ii,t,k

i I, t T, i I, t T, k K,

(23)

s s Fi,t,k = xi,t,k FTi,t sl sl Ii,t,k = xi,t,k ITi,t

(24)

an,i,k xi,t,k = bn,i


kK

xi,t,k = 1 i I, t T
kK U L xi,t,k xi,t,k xi,t,k s Us FTi,t FTi,t P yi,t Ai HislL s Ii,t,k

(19)

A period t, and xk is the mass fraction of brine feeding for the specie k.

A where FTi,t is the total mass ow rate of feed stream to pond i in

i I, t T, k K

2.3. Harvest planning For the planning of pond system, it is considered that there are discrete processes times over a period of 1 year horizon (Biegler, Grossmann, & Westerberg, 1997). These are dened considering that during that period there are constant operating conditions as the rate of evaporation and brine composition. Usually, the harvest lasts about a month, so if the process time is one month, it is not necessary to include restrictions on the planning process. Unfortunately, planning is usually carried out in 1- or 2-week periods, and therefore, it is necessary to include restrictions related to planning. As an example, lets consider that the harvest lasts four weeks and that the operational time periods are of one week. Then if the pond i is operating in the period (t 1) and within the next period t the pond i is harvesting, it must be in harvest during the (t + 1), (t + 2), and (t + 3) periods, completing four harvest periods. This is dened with the following expressions:
P P P P P yi,(t1) yi,t yi,(t+1) yi,(t+2) yi,(t+3)

i I, t T
sl ITi,t sl i P Ai HislU yi,t

i I, t T

kK s i

P Ai HisU yi,t

i I, t T
P Ai HTiU yi,t

I sl k K i,t,k sl i sl P Ii,t,k Uyi,t


s P Ii,t,k Uyi,t

Is k K i,t,k s i

i I, t T

i I, t T, k K i I, t T, k K

P CSi,t,k U(1 yi,t ) i I, t T, k K P ri,t,l Uyi,t l e P Fi,t,k Uyi,t

i I, t T

i I, t T, k K

P where yi,t is a binary variable that takes the value 1 if the pond i is in operation in the period t, and takes the value 0 if the pond i is in harvest in the period t.

i I, t T.

(25)

2.2. Mixer and spliter modeling In each pond, it is considered that there is a streams mixer at the input of the pond and a streams splitter at the pond output. The splitter allows the mass ow from a pond i to pond j, i.e., the pond output stream splitting from a pond to several ponds is not allowed. Then, for each pond: Fi,j,t,k =
F yi,j,t s Fi,t,k

This logical expression can be expressed with binary variables, following procedures described in the literature (Clocksin & Mellish, 1981; Raman & Grossmann, 1991), then Eq. (25) can be represented by the following system of inequalities:
P P P yi,(t1) yi,t + yi,(t+1) 1 P P P yi,(t1) yi,t + yi,(t+2) 1 P P P yi,(t1) yi,t + yi,(t+3) 1

i I, t T, i I, t T, i I, t T.

(26) (27) (28)

Fi,j,t,k = 0

F yi,j,t

i, j I, i = j, t T, k K, /

(20)

where Fi,j,t,k is the mass ow rate of specie k from pond i to pond j s in the period t and Fi,t,k is the output mass ow rate in the pond i in time period t for the specie k. Disjunction (20) can be written as
s F Fi,j,t,k Fi,t,k U(1 yi,j,t ) i, j I, i = j, t T, k K / s Fi,t,k =

Operationally, it is not possible to transfer brine simultaneously in the same period between two ponds. Then inequality (29) must be F F included, which restricts the binary variables yi,j,t and yj,i,t cannot have the value 1 in the same period, i.e., if there is brine transfer from pond i to pond j, there can be no brine transfer from the pond j to the pond i in the same period, and vice versa:
F F yi,j,t + yj,i,t 1

Fi,j,t,k

i I, t T, k K
(21)

i, j I, t T.

(29)

j I,i = j / F yi,j,t = 1 j I,i = j /

i I, t T

where
F yi,j,t =

During the harvest periods, the use of the contractor who performs the harvest is limited, which requires that one pond can be harvested only, therefore it should include a restriction requiring a single harvest pond at a time. The restriction is
P (1 yi,t ) 1 t T. iI

1 if there is owrate from pond i to pond j 0 if there is not owrate from pond i to pond j

(30)

K. Ossandn et al. / Computers and Chemical Engineering 34 (2010) 620630 Table 3 Common data in the examples. Parameter Water density Solid density Lower bound for brine inventory height Upper bound for brine inventory height Upper bound for solid inventory height Upper bound for the total pond height Horizon time Unit ton/m ton/m3 m m m m year
3

625

Value 1 1.5 0.2 1.1 0.8 1.1 1

Fig. 3. Daily rate of evaporation.

2.4. Objective functions The problem is to determine the maximum salt harvest and then plan the harvest in order to minimize the cost associated with this activity. The objective function that maximizes the harvesting of salt is Max
iI t T kK

the same for all examples. To determine the cost of pumping, it was assumed that the ponds are located in rows and with a height difference between ponds of 4 m, allowing, when possible, the ow of brine by gravity. Other ow of brine between ponds must be pumped at a cost as indicated in Table 2. MINLP model for each case was implemented and solved using the BARON/GAMS software. 3.1. Example 1: planning ponds of sodium nitrate 3.1.1. Data This example relates to the planning of operations and harvest of sodium nitrate from aqueous solutions containing Na+ , K+ and NO3 . The brine feed to the ponds has a constant composition in time, with 19% of NaNO3 , 2% of KNO3 and 79% of H2 O. The density of brine is considered constant and equal to 1.364 ton/m3 (Linke & Seidell, 1965). To model the equilibrium conditions the system operating at an average temperature of 25 C is considered, then it is possible to use the equilibrium data of the NaNO3 KNO3 H2 O system at 25 C available in the literature (Linke & Seidell, 1965) and are given in Table 4. NaNO3 is the only salt that is precipitated, then a model for the solidliquid equilibrium with NaNO3 as solid is necessary, which corresponds to the aqueous system saturated with NaNO3 from 0% KNO3 (point C, Table 4) to 19.1% of KNO3 (point B, Table 4). As indicated above, in this work, these equilibrium conditions are represented by a hyperplane, which in this case corresponds to the equation of a straight line connecting points B and C: 2.5132xNaNO3 + xKNO3 = 1.2038. (36)

CSi,t,k .

(31)

Once the maximum production of salt has been determined, this value is used as a restriction for the new objective function that minimizes the operation periods, i.e., a maximum amount of salt in a minimum period of operation and thus, earn higher harvest periods. The objective function that minimizes the period of operation is given by the following expression: Min
iI t T P yi,t .

(32)

The restriction associated with the maximum production of salts can be relaxed to search for more harvest periods. For example the following restriction can be used: CSi,t,k CM,
iI t T kK

(33)

where CM is the maximum production of salts obtained from the rst objective function and is a factor which adjusts production to a smaller value than the maximum. If is equal to 1, the maximum production of salt is included in the model. In this work, a value of equal to 0.9 is used. 3. Case studies This section presents the application of the model to three examples, two corresponding to a ternary system and another one for a quaternary system. In each example, several cases are studied by changing some of the conditions of the problem. The examples considered include data, such as the daily rate of evaporation shown in Fig. 3, the cost of pumping between ponds (see Table 2), and other variables displayed in Table 3, which are
Table 2 Pump cost between ponds in US$/ton. Pond i Pond j 1 1 2 3 4 0 8.72 104 1.74 103 2.62 103 2 0 0 8.72 104 1.74 103 3 0 0 0 8.72 104 4 0 0 0 0

Operational data are given in Table 5. The system has three solar evaporation ponds with an area of 40,000 m2 each. The operation programming considers 12 periods, each lasting a month. Each pond has an initial solution and precipitate (NaNO3 ) inventories. Details of inventory values as well as brine concentrations are given in Table 5. Also, Table 5 provides brine concentration ranges in which to operate each pond. Four cases were considered, being the case 1 as given in Table 5. Case 2 corresponds to case 1, but where permitted brine concentration range where the pond can operate, has been extended. Specically, the NaNO3 concentration range is 0.4030.479 for all ponds. On the contrary, in case 1 permitted concentrations of NaNO3 have different ranges for each pond (see Table 5). Case 3 corresponds to case 1, but an additional pond has been included. This new pond pond number 4 has the same characteristics as pond
Table 4 Equilibrium data for the NaNO3 KNO3 H2 O system at 25 C. Liquid phase composition, % w Key A B C NaNO3 0 40.3 47.9 KNO3 27.7 19.1 0 H2 O 72.3 40.6 52.1 KNO3 KNO3 + NaNO3 NaNO3 Solid phase

626 Table 5 Data for example 1.

K. Ossandn et al. / Computers and Chemical Engineering 34 (2010) 620630

Ponds Annual operation periods Length of operation periods Pond area (m2 ) Upper bound for output ow rate (Mton/period) Pond 1 Initial solution inventory (ton) Initial solid (NaNO3 ) inventory (ton) 47,100 14,204

3 12 1 month 40,000 800 Pond 2 47,900 13,324 0.450 0.073 0.477 0.4100.460 0.0050.190 0.3900.550 Pond 3 46,700 14,644 0.460 0.048 0.492 0.4200.480 0.0050.190 0.3900.550

Initial composition of solution inventory (mass fraction) 0.440 NaNO3 0.098 KNO3 0.462 H2 O Concentration range of solution allowed in each pond (mass fraction) 0.4000.450 NaNO3 0.0050.190 KNO3 0.3900.550 H2 O

Table 6 Mathematical and computational features of cases for example 1. Cases Continuous variables Binary variables Restrictions Iterations number CPU time (s)a CPU time (s)b Salt production (ton/year) Harvest periods
a b

Case 1 1,387 108 2,174 1 153.66 202.31 89,562 5

Case 2 1,387 108 2,174 1 192.91 149.31 99,350 10

Case 3 2,041 192 3,086 1 516.00 561.75 119,395 5

Case 4 2,986 237 4,789 1 1941.22 2362.77 94,763 9

CPU time for the objective function that maximizes the salt harvest. CPU time for the objective function that minimizes the operation times.

3 in case 1. In practice, it is usual that the operation programming of solar evaporation ponds is made up in time periods less than one month. So that in case 4, 26 annual periods (2 weeks) were considered, therefore the upper bound for output ow rate (Mton/period) was changed from 800 to 400. All the other conditions are the same as case 1. 3.1.2. Results Resolution of the cases was achieved using a Intel 2.33 GHz Core 2 Quad processor. The main features of the mathematical models, its convergence and computational results are shown in Table 6. The CPU time required to achieve convergence to an optimal solution, depends on the size of the problem, then the cases 1 and 2 requires less than four minutes, while for cases 3 and 4, which includes more variables and restrictions the CPU time is between 8 and 40 min. The planning of the salt harvest and the distribution of brine between ponds during a year for the case 1, is shown in Fig. 4. Symbols P1, P2 and P3 represent ponds 1, 2 and 3 respectively. The brine feed to the pond system is identied in Fig. 4, in each pond and

period, as operation with feed. It is noted that all ponds are fed with fresh brine, and in some periods all the ponds are fed simultaneously. The brine transfer between ponds is identied with the symbol of the destination pond. For example, in period 1, brine is transferred from pond 3 to pond 1. Regarding the distribution of brine ows, it is observed that there are periods where not all ponds have outows, for example in period 1, pond 1 does not distribute the brine ow to other pond and does not have any fresh feed, while in period 2 it distributes its outow into pond 3 and is fed with unsaturated brine (fresh feed). These results suggest that one way to get maximum salt harvest is to manage the brine ows between ponds for brine concentrations that allow the maximum precipitation of salts. Clearly in Fig. 4, it can be seen that the number of harvest periods is 5 months, corresponding to months of March, April, October, November and December, a reasonable time considering that the number of ponds is small. It can also be seen that months of harvest do not correspond to months in which evaporation is minimal (winter, i.e., from June to September), as usual in practice. Several explanation are possible, rstly, it must be considered that the planning period is one year and the model tends to harvest yearend in order to maximize production. Secondly, restrictions on the brine composition may limit planning options. These topics will be analyzed later. An interesting analysis to observe is the variation of the height of salts in each pond to determine an average height for harvest. Fig. 5 shows the prole height of salts for each pond. Variation of the solid height, shows that the harvesting is done without reaching the maximum salt height permitted in the ponds (0.8 m). The reason is that it has reached the maximum allowable total height (salts + brines). It also shows that there are periods where there is no variation in salts height, this indicates that there is no precipitation. It should also be noted that the pond 1 cannot be harvested

Fig. 4. Planning harvest and distribution of brine ows during 1 year for case 1 (example 1).

Fig. 5. Variation in salt height of case 1 (example 1).

K. Ossandn et al. / Computers and Chemical Engineering 34 (2010) 620630

627

Fig. 6. Planning harvest and distribution of brine ows during 1 year for case 2 (example 1).

Fig. 8. Planning harvest and distribution of brine ows during 1 year for case 4 (example 1). Table 7 Equilibrium data for the system KClKNO3 K2 SO4 H2 O at 25 C (5). Liquid phase composition (wt%) Keys A B C D E F G KCl 21.6 0 26 21.9 0 0 26.4 KNO3 14.6 25.4 0 14.6 0 27.5 0 K2 SO4 0.9 4 1.1 0 10.75 0 0 H2 O 62.9 70.6 72.9 63.5 89.25 72.5 73.6 KCl + KNO3 + K2 SO4 KNO3 + K2 SO4 KCl + K2 SO4 KCl + KNO3 K2 SO4 KNO3 KCl Solid phase

in period 12 since pond 3 is in harvesting and harvest is not feasible for 2 ponds on the same period simultaneously. This restriction can be eliminated, but it will mean to have operational capacity to perform simultaneous harvests. For case 2, the planning of the salt harvest and the distribution of brine between ponds during a year is shown in Fig. 6, which clearly indicates that the number of harvest periods corresponds to 10 periods from March to December. These are periods of low and high evaporation. The production of salts is 99,350 ton/year, which is higher than in case 1 (89,562 ton/year). These results show that restrictions in concentration ranges of evaporation ponds affect the planning and production of salts. For case 3, the planning of salt harvest and the distribution of brine between ponds during a year is shown in Fig. 7, which clearly show that the harvest is made in March, April, May, July and September. The largest harvest is done in winter time (July) probably because there are more ponds available and therefore more operational alternatives. However, ponds are not harvested in the end of the year, although there are solid inventories available. This may be due to restrictions in brine concentrations. As in previous cases, there are some periods in which ponds are either fed with fresh brine and/or with brine from other ponds, or they are not fed at all. As in previous cases, salt height does not reach maximum height allowed (0.8 m) because the maximum total height of the pond is obtained rst. The above is common for all cases, i.e., the harvest is done when the maximum height of the pond is reached. For case 4, the planning of the salt harvest and the distribution of brine between ponds during a year is shown in Fig. 8, and the variation in salt height is shown in Fig. 9. Fig. 8 clearly indicates that the number of harvest periods correspond to 9 periods of 2 weeks, which equals 4.5 months of harvest. There are three periods where major harvest are done; March to April for pond 2, JuneJuly for pond 3 and December for pond 1. In this case, unlike case 1, there is harvest in the winter period, with a slight increase in salt harvest. However, inventories at the end of the year are small compromising future production. By using periods of 2 weeks instead of 1-month periods, the model can better represent reality, as several of the

variables such as brine ow and concentration varies during each operation period. 3.2. Example 2: planning ponds of sodium nitrate with cyclic schedule 3.2.1. Data The purpose of this example is to analyze the effect of a cyclic planning. Data in this example are the same as the previous one, but restrictions have been included so that the initial and nal inventories are similar. Concentrations of the brine being fed to each pond, the evaporation rates and commodity prices change every year. However, cyclic planning will allow better long-term gains if the values of the variables listed above are relatively stable. Then brine feed to the ponds has a composition of 19% of NaNO3 , 2% of KNO3 and 79% of H2 O. NaNO3 is the only salt that is precipitated. The system has solar evaporation ponds with an area of 40,000 m2 each. The operation programming considers 12 periods, each lasting a month, and the upper bound for output ow rate is 800 Mton/period. The NaNO3 concentration range is 0.4030.479 for all ponds, the concentration range for the other species are the same as in Table 5. Two cases were considered, being case 1 a system with 3 ponds and case 2 a system with 4 ponds.

Fig. 7. Planning harvest and distribution of brine ows during 1 year for case 3 (example 1).

Fig. 9. Variation in salt height of case 4 (example 1).

628 Table 8 Data for example 3.

K. Ossandn et al. / Computers and Chemical Engineering 34 (2010) 620630

Ponds Annual operation periods Length of operation periods Pond area (m2 ) Upper bound for output ow rate (Mton/period) Pond 1 Initial solution inventory (ton) Initial solid (K2 SO4 ) inventory (ton) 41,850 14,205

3 12 1 month 40,000 800 Pond 2 43,540 12,114 0.080 0.054 0.071 0.795 0.0500.200 00.146 0.0090.1075 0.6290.8925 Pond 3 41,500 14,639 0.090 0.061 0.066 0.783 0.0700.216 00.146 0.0090.1075 0.6290.8925

Initial composition of solution inventory (mass fraction) KCl 0.070 0.047 KNO3 0.076 K2 SO4 0.807 H2 O Concentration range of solution allowed in each pond (mass fraction) KCl 0.0010.190 00.146 KNO3 0.0090.1075 K2 SO4 0.6290.8925 H2 O

Fig. 12. Planning harvest and distribution of brine ows during 1 year for case 2 (example 2). Fig. 10. Planning harvest and distribution of brine ows during 1 year for case 1 (example 2).

3.3. Example 3: planning ponds of potassium sulphate 3.3.1. Data This example applies the method to a quaternary system and analyzes the effect of restrictions on the range of brine concentrations of each pond on the harvest planning. In this example, the production of K2 SO4 from brines containing K+ , Cl , NO3 and SO4 operating at 25 C was studied. Then the system of interest corresponds to the quaternary system KClKNO3 K2 SO4 H2 O at 25 C, the equilibrium data are given in Table 7 and shown in Fig. 14. The molar composition of the brine fed to the ponds, which is constant in time, is 1% KCl, 1% KNO3 , 19% K2 SO4 and 79% of H2 O. The density of the brine is considered equal to 1.212 ton/m3 and corresponds approximately to the density of a saturated solution of this quaternary system (Linke & Seidell, 1965). For simplicity, it was considered that the annual evaporation rate, the pumping

3.2.2. Results The problem was solved in an Intel 2.33 GHz Core 2 Quad processor, using BARON/GAMS. The problems have 1387 and 2041 continuous variables, 108 and 192 binary variables and 2183 and 3090 constraints for cases 1 and 2 respectively. Times of convergence were 188.47 and 1199.77 s for cases 1 and 2 respectively to maximize the harvesting of salt, and 231.38 and 1090.41 s for cases 1 and 2 respectively to minimize the operation periods. The salt harvest planning and the brine distribution between ponds during a year are shown in Figs. 10 and 12 for cases 1 and 2 respectively. Figs. 11 and 13 show salt height variations for the same cases. The maximum annual production of salt is 61,115 and 192,000 ton for cases 1 and 2 respectively (Figs. 11 and 13).

Fig. 11. Variation in salt height of case 1 (example 2).

Fig. 13. Variation in salt height of case 2 (example 2).

K. Ossandn et al. / Computers and Chemical Engineering 34 (2010) 620630

629

Fig. 16. Salt height variations for case 1 example 3.

Fig. 14. Phase diagram for the KClKNO3 K2 SO4 H2 O system at 25 C.

costs, horizon time, and operational parameters are the same as in example 1. The solar evaporation system consists of 3 ponds with 40,000 m2 each, each pond with a maximum output mass ow rate of 800 Mton/period, with periods of 1 month. Then, 12 annual operating periods are considered. Other operational data, such as the pond initial conditions, are given in Table 8. To represent the solidliquid equilibrium, a hyperplane was used (Pressly & Ng, 1999). As the solid phase is K2 SO4 , the surface of interest, which represents the solution saturated in K2 SO4 , corresponds to ABEC in Fig. 14. As the surface cannot be represented by a single hyperplane, it is necessary a set of hyperplanes. Let us call them sub-hyperplanes, but in fact they correspond to a hyperplane. The ABEC surface can be represented by two sub-hyperplanes, one including the ABE points and the other including the ACE points. Considering the initial brine concentration and the operational conditions, it is necessary to consider the sub-hyperplane dened by the ACE points only. Then the sub-hyperplane ACE is represented as (Cistermas & Montenegro, 2006): 2.956xKCl + xKNO3 + 7.965xK2 SO4 = 85.621. (37)

Fig. 17. Harvest planning and distribution of brine ows over 1 year for case 2 example 3.

extended. Specically, the KCl concentration range is 0.0010.216 for all ponds. On the contrary, in case 1 the permitted concentrations of KCl are 0.0010.190 for pond 1, 0.0500.200 for pond 2, and 0.0700.216 for pond 3. 3.3.2. Results The problem was solved in an Intel 2.33 GHz Core 2 Quad processor, using BARON-GAMS. The problem has 1753 continuous variables, 108 binary variables and 2713 constraints. Times of convergence were 677.08 and 863.03 s for cases 1 and 2 respectively to maximize the harvesting of salt, and 673.38 and 336.97 s for cases 1 and 2 respectively to minimize the operation periods. The salt harvest planning and the brine distribution between ponds during a year are shown in Figs. 15 and 17 for cases 1 and 2 respectively, whereas salt height variations are shown in Figs. 16 and 18 for the same cases. The maximum annual production of salt are 62,438 and 64,035 ton for cases 1 and 2 respectively, this means a small increase in production because the brine concentrations bounds were wider. However, harvest time increased signicantly. Namely, harvest periods are 5 and 10 months for cases 1 and 2 respectively, In case 1, the harvesting is done

Two cases were considered, being case 1 the same as given in Table 8. Case 2 corresponds to case 1, but where the permitted brine concentration range where the pond can operate, has been

Fig. 15. Harvest planning and distribution of brine ows over 1 year for case 1 example 3.

Fig. 18. Salt height variations for case 2 example 3.

630

K. Ossandn et al. / Computers and Chemical Engineering 34 (2010) 620630

during the months of April, May, July, August and October; in case 2 the harvest takes place from March to December. In the absence of cyclic planning status, ending inventories are small (Figs. 16 and 18). 4. Conclusions A MINLP model has been developed including disjunctive programming and hyperplane theory for the process planning for harvesting solar evaporation ponds. The model includes two objective functions: the rst one maximizes the salt production and the second one minimizes the operation periods. Although studied examples had considered a small number of ponds and components, along with extended periods of operation, the model to plan the harvest in the solar evaporation ponds is useful to identify conditions that may be effective in practice and can be an approach to optimize a part of the solar evaporation system. In the examples studied, ternary and quaternary systems with 3 and 4 ponds, the size planning problem consists of 13874397 continuous variables, 108420 binary variables and 21746792 constraints. The time for solving these problems is more sensible to the amount of annual periods of planning than to the number of components or number of ponds in the system. Convergence times indicate that in order to solve more complex systems, exploration of other optimization strategies will be desirable. The results of the presented examples have shown that several operation strategies can be used to maximize the production of salt and to increase the harvest periods. These strategies, which could be applicable in practice, include carrying out the harvest in summer months, feeding ponds in parallel, harvest when the maximum total height had been obtained, consider complex structures of brine transfer between ponds, among others. The results show that the harvest planning is sensitive to the brine concentration bounds in each pond. Planning with and without cyclicity has a signicant effect on results. The application to more complex systems, as quinary systems and production of various salts, and the use of new optimization strategies to minimize the times of problem solving relates to future work.

Acknowledgement The authors wish to thank CONICYT for nancial support, through Fondecyt Project 1060342. References
Biegler, L. T., Grossmann, I. E., & Westerberg, A. W. (1997). Systematic methods of chemical process design. New Jersey: Prentice Hall PTR. Bloch, M. R., Farkas, L., & Spiegler, K. S. (1951). Solar evaporation of salt brines. Industrial and Engineering Chemistry, 43, 15441553. Butts, E. (2001). Chemicals from brine (4th ed.). Kirk-Othmer encyclopedia of chemical technology. Cistermas, L. A., & Montenegro, M. (2006). Simulacin de Pozas de Evaporacin Solar con Hiperplanos. Parte I: Hiperplanos Ingeniera Qumica, 30, 2941. Cistermas, L. A., Cangana, J., Aravena, R., & Vargas, P. (2007). Simulacin de Pozas de Evaporacin Solar con Hiperplanos. Parte II: Pozas de Evaporacin Solar, Ingeniera Qumica, 31, 6169. Clocksin, W. F., & Mellish, C. S. (1981). Programming in prolog. New York: SpringerVerlag. Flotz, G. E. (1993). In J. J. McKetta (Ed.), Lithium and lithium compounds in inorganic chemicals handbook. New York: Marcel Dekker. Hammi, H., Musso, J., Mnif, A., & Rokbani, R. (2003a). Solubility phase diagrams coupled to computer science (DPAO). Part II: Applied to isothermal evaporation of Tunisian natural brines. Calphad, 27, 7177. Hammi, H., Musso, J., Mnif, A., & Rokbani, R. (2003b). Tunisian salt lakes evaporation studied by the DPAO method based on solubility phase diagrams. Desalination, 158, 215220. Hammi, H., Musso, J., Mnif, A., & Rokbani, R. (2004). Crystallization path of natural brine evaporation using the DPAO method. Desalination, 166, 205208. Hamzaoui, A. H., Mnif, A., Hammi, H., & Rokbani, R. (2003). Contribution to the lithium recovery from brine. Desalination, 158, 221224. Kwok, K. S., Ng, K. M., Taboada, M. E., & Cisternas, L. A. (2008). Thermodynamics of salt lake system: Representation, experiments, and visualization. AIChE Journal, 54, 707727. Linke, W. F, & Seidell, A. (1965). Solubilities of inorganic and metal organic compounds (4th ed.). Washington, DC: American Chemical Society. Lior, N., & Bakish, R. (2001). Supply and desalination (4th ed.). Kirk-Othmer encyclopedia of chemical technology. Murthy, Ramakrishna, G. R., & Pandey, K. P. (2003). Comparative performance evaluation of fertiliser solar pond under simulated conditions. Renewable Energy, 28, 455466. Musso, J. A. (2003). Solubility phase diagrams coupled to computer science (DPAO). Part I: Theory of the sequential representation. Calphad, 27, 6569. Pressly, T. G., & Ng, K. M. (1999). Process boundary approach to separations synthesis. AIChE Journal, 45, 19391952. Raman, R., & Grossmann, I. E. (1991). Relation between MILP modeling and logical inference for chemical process synthesis. Computers and Chemical Engineering, 15, 7384. Song, P., & Yan, Y. (2003). Thermodynamics and phase diagram of the salt lake brine system at 298.15 K. Computer Coupling of Phase Diagrams and Thermochemistry, 27, 343352.

Вам также может понравиться