A multi-scale intelligent design method for reverse osmosis membrane elements and processes for high-flux membranes

Abstract

The present invention discloses a multi-scale intelligent design method for reverse osmosis membrane elements and processes for high-flux membranes, comprising the following steps: establishing a three-dimensional multi-physics field coupled CFD model of the reverse osmosis membrane element; optimizing the geometric parameters of the water inlet screen unit at a submillimeter scale using a hybrid intelligent optimization algorithm combining Bayesian optimization with pattern search; characterizing the hydrodynamics and mass transfer characteristics through three-dimensional multi-physics field simulation based on the optimization results; establishing a meter-scale industrial system model, and optimizing the process parameters and membrane parameters using a genetic algorithm. The present invention can suppress the fluid resistance caused by the improvement of mass transfer capacity, thereby reducing energy consumption.

Description

Technical Field

[0001] The present invention relates to the technical field of water treatment equipment, and in particular to a multi-scale intelligent design method for reverse osmosis membrane elements and processes for high-flux membranes. Background Art

[0002] In the prior art, the flow channels of the reverse osmosis membrane are mostly meshed diamond-shaped flow channels, with a structure such as Figure 1 As shown in the figure, during operation, water to be treated enters the flow channel at the water inlet. After being treated by the reverse osmosis membrane, the impurity-laden concentrate flows through the concentrate channel and is discharged from the water production end. The diamond mesh screen changes the hydrodynamic conditions on the membrane surface, reducing the amount of retained material near the membrane surface and increasing the mass transfer rate, thereby reducing the rate of membrane fouling. However, the diamond mesh screen structure's ability to improve the mass transfer coefficient is limited. Furthermore, as the mass transfer capacity increases, the pressure drop within the channel increases rapidly, requiring more energy. Summary of the Invention

[0003] Purpose of the invention: The purpose of the present invention is to provide a multi-scale intelligent design method for reverse osmosis membrane elements and processes for high-flux membranes to solve the problems existing in the background technology.

[0004] Technical solution: The multi-scale intelligent design method for reverse osmosis membrane elements and processes for high-flux membranes described in the present invention includes the following steps:

[0005] (1) Establish a three-dimensional multi-physics field coupled CFD model of reverse osmosis membrane elements;

[0006] (2) At the submillimeter scale, a hybrid intelligent optimization algorithm combining Bayesian optimization with pattern search is used to optimize the geometric parameters of the water inlet screen unit;

[0007] (3) Based on the optimization results, characterize the hydrodynamics and mass transfer characteristics through three-dimensional multi-physics simulation;

[0008] (4) Establish a meter-scale industrial system model and use genetic algorithms to optimize the design of process parameters and membrane parameters.

[0009] Furthermore, the reverse osmosis membrane element includes: a reverse osmosis unit and a water production central tube; a plurality of pure water diversion ports are provided on the tube wall of the water production central tube, and the reverse osmosis unit is wound around the circumference of the water production central tube; the reverse osmosis unit includes a water inlet spacer, a reverse osmosis membrane and a water production spacer stacked in sequence, wherein the water inlet spacer is composed of a plurality of closely arranged V-like structure subunits.

[0010] Furthermore, the cross-sectional shape of the V-like structural subunit is defined by a parabolic function, a cosine function or an absolute value function; the V-like structural subunit generates eddies in the water inlet channel to improve the mass transfer coefficient and suppress concentration polarization.

[0011] Furthermore, a three-dimensional multi-physics coupled CFD model is constructed based on the Navier-Stokes equations and convection-diffusion.

[0012] Furthermore, the hybrid intelligent optimization algorithm is specifically as follows: Bayesian optimization is used as a global search to generate an initial optimal solution; with the initial optimal solution as the starting point, a pattern search algorithm is used to perform local fine optimization;

[0013] Furthermore, the Bayesian optimization adopts the expected improvement function EI as the acquisition function; the pattern search algorithm is specifically as follows: a positive definite expansion set is used to define the search direction; the solution is updated by iteratively performing search and polling operations, and terminated when the grid size is smaller than a threshold.

[0014] Furthermore, the optimization objective function of step (2) is:

[0015] ;

[0016] in, APLR It represents the ratio of the axial pressure drop per meter between the designed membrane module and the commercial membrane module. ; and represent the mass transfer coefficients of the designed membrane module and the commercial membrane module, respectively; Used to control the trade-off between fluid resistance and mass transfer capacity, As a penalty factor to punish excessive pressure drop.

[0017] Furthermore, the meter-scale industrial system model is described by one-dimensional differential algebraic equations.

[0018] Furthermore, step (4) is as follows:

[0019] The formula for a one-dimensional differential algebraic equation is as follows:

[0020] ;

[0021] Where k is the number of stages of the seawater reverse osmosis system, Q is the flow rate, is the transmembrane pressure, is the water flux, is the salinity in the main body of the retentate, is the salinity in the main body of the permeate, is the salinity at the membrane wall, is the dimensionless length, is the membrane area of ​​the kth stage, is the number of membrane elements in each pressure vessel in the kth stage, and In order to optimize the length of the membrane in the module perpendicular and parallel to the feeding direction, i.e. the x and y axis, is the water permeability coefficient, is the reflection coefficient, is the osmotic pressure coefficient, is the intrinsic retention rate of the membrane;

[0022] The flow rate is expressed as in, For the k Number of pressure vessels of each level, The number of water inlet mesh sheets for each membrane element, and are the feed channel height and feed channel porosity, respectively;

[0023] The membrane area of ​​the kth stage is expressed as ,in, is the length of the membrane parallel to the y-axis of the feed in the commercial module, is the membrane area of ​​commercial membrane elements;

[0024] The salinity in the bulk of the permeate and at the membrane wall is expressed as ,in, is the salt permeability coefficient, is the average unit mass transfer coefficient at the membrane wall;

[0025] The optimized process and membrane parameters at meter scale include: water permeability coefficient of membrane and salt permeability coefficient , Number of pressure vessels at each level , the number of membrane elements in each pressure vessel , transmembrane pressure at each level 、Number of spacers per membrane element ; And the optimization objective function of process and membrane parameter optimization design is:

[0026] ;

[0027] Including the annual cost of membrane and electricity cost per cubic meter of fresh water, is the required total membrane area, is the cost of the membrane per unit area, is the annual amortization factor, is the annual operating time, is the permeate flow rate, is the energy cost per kilowatt-hour, is the specific energy consumption.

[0028] An electronic device according to the present invention includes a memory and a processor, wherein the memory stores a computer program, and the processor implements the steps of any one of the methods when executing the program.

[0029] Beneficial effects: Compared with the prior art, the present invention has the following significant advantages: The present invention adopts a V-like structure instead of a mesh structure in the design of the submillimeter scale water inlet screen structure, and adopts the proposed hybrid intelligent optimization algorithm of Bayesian combined with pattern search to optimize the design of the geometric parameters of the membrane elements. The designed structure can not only greatly improve the mass transfer capacity of the reverse osmosis membrane element and improve the water production efficiency by generating eddies on the cross section of the water inlet channel, but also can suppress the fluid resistance caused by the improvement of the mass transfer capacity while significantly reducing the effects of concentration polarization and membrane fouling, thereby reducing energy consumption. At the same time, the present invention constructs a meter-scale system level model based on three-dimensional multi-physics field simulation, and further optimizes the design of the process and membrane parameters through genetic algorithms. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] Figure 1 The flow channel distribution diagram of the water inlet screen in the prior art;

[0031] Figure 2 Schematic diagram of the reverse osmosis membrane element structure of the present invention;

[0032] Figure 3 for Figure 2 Flow channel distribution diagram of the spacer of the reverse osmosis membrane element;

[0033] Figure 4 Schematic diagram of the geometric design parameters and calculation domain of the water inlet screen unit of the present invention;

[0034] Figure 5 This is a flow chart of the geometric parameter optimization method of the water inlet screen unit of the present invention;

[0035] Figure 6 The invention provides a multi-scale hierarchical optimization design framework for reverse osmosis membrane elements and systems for high-flux membranes.

[0036] Figure 7 Iterative graph of the objective function for optimizing the geometric parameters of the water inlet screen unit of the present invention;

[0037] Figure 8 This is a schematic diagram of the results of the geometric parameter optimization of the water inlet screen unit of the present invention;

[0038] in, Figure 1-Figure 3 Among them, 1. Reverse osmosis unit; 11. Water inlet screen; 12. Reverse osmosis membrane; 2. Water production center pipe; 21. Diversion port. DETAILED DESCRIPTION

[0039] The technical solution of the present invention will be further described below with reference to the accompanying drawings.

[0040] like Figure 2As shown, an embodiment of the present invention provides a multi-scale intelligent design method for reverse osmosis membrane elements and processes for high-flux membranes, comprising the following steps:

[0041] (1) A three-dimensional multi-physics field coupled CFD model of the reverse osmosis membrane element is established; specifically, the reverse osmosis membrane element includes a reverse osmosis unit 1 and a water production center pipe 2. A plurality of pure water diversion ports 21 are provided on the wall of the water production center pipe, which are evenly distributed on the outer wall of the water production center pipe 2. The reverse osmosis unit 1 is wound around the circumference of the water production center pipe 2. The reverse osmosis unit includes an inlet screen 11, a reverse osmosis membrane 12 and a water production screen stacked in sequence, and the structure of the inlet screen 11 is composed of a plurality of neatly and tightly arranged V-shaped sub-units.

[0042] Among them, the V-type structure refers to a structure whose cross section is composed of basic functions such as parabola, trigonometric function, absolute value function, etc. The water inlet screen includes the following design parameters: sub-unit width W1, sub-unit thickness H1, half of the sub-unit height H2, the distance between adjacent sub-units in the x-axis direction Ws, and the distance between adjacent sub-units in the y-axis direction Ls. In addition, the design parameters also include the basic functions f ( x ) parameters. The width W1 of the subunit is 0.1-1 mm; the thickness H1 of the subunit is 0.01-0.1 mm; half the height H2 of the subunit is 0.3-1 mm; the distance Ws between adjacent subunits in the x-axis direction is 0.2-1 mm; and the distance Ls between adjacent subunits in the y-axis direction is 0.4-2 mm.

[0043] The Navier-Stokes equations are used to describe the motion of viscous incompressible fluids and are expressed as:

[0044] ;

[0045] The convection-diffusion equation is used to describe the mass transfer process and is expressed as:

[0046] ;

[0047] Where, represents the fluid density, represents the fluid velocity, Indicates the fluid pressure, represents the identity matrix, represents the fluid dynamic viscosity, represents the gradient operator, represents the diffusion coefficient tensor, Indicates the fluid concentration.

[0048] (2) At the submillimeter scale, a hybrid intelligent optimization algorithm combining Bayesian optimization with pattern search is used to optimize the geometric parameters of the water inlet screen unit; specifically, the hybrid intelligent optimization method includes global optimization and local optimization. Among them, the global optimization adopts the Bayesian optimization algorithm, which includes the following steps: obtaining sampling points by randomly sampling the target model, and using the sampling points to train the Gaussian process model; determining the next evaluation point by calculating the acquisition function; after evaluating the new point, updating the posterior distribution of the Gaussian process according to the new observation results; returning to train the Gaussian process model using the sampling points and repeating this cycle until the convergence criterion is reached. Among them, the convergence criterion refers to a criterion composed of the maximum number of evaluations, the posterior uncertainty criterion, and the objective function convergence criterion. Among them, the posterior uncertainty criterion means that if the variance of the model near the optimal solution is small enough, it means that the model has been sufficiently certain that the current solution is optimal, and it can be stopped at this time; the objective function convergence criterion means that if the change of the optimal target value is less than a preset threshold over a period of time or multiple iterations, the algorithm is considered to have converged and can be stopped.

[0049] The acquisition function refers to the expected improvement function, which is expressed as:

[0050] ;

[0051] in, is the optimization variable consisting of the design parameters , is a Gaussian process in the variable The mean value at is the currently known optimal objective function value, is a small positive constant, is the cumulative distribution function of the standard normal distribution, is a Gaussian process in the variable The standard deviation of is the probability density function of the standard normal distribution. The variable with the largest final acquisition function value will be used as the next evaluation point.

[0052] The local optimization part adopts the pattern search algorithm, which includes the following steps: determining the initial solution, initial grid size and positive definite expansion set; evaluating the objective function value of the current point and performing a search operation. If a better value cannot be found, each subset in the positive definite expansion set is traversed and a polling operation is performed. Otherwise, the solution is set as the current position; if the polling operation does not find a better value, the grid size is reduced and the search step is performed again, and this cycle is repeated until the convergence criterion is reached. Otherwise, the solution is set as the current position; wherein the grid is a conceptual grid composed of solutions, and the algorithm will select a solution from it and perform the objective function calculation on the solution; the initial solution is set to the optimal solution after Bayesian optimization; the positive definite expansion set is set to A set of search directions, n represents the dimension of the design variable, expressed as In addition, the convergence criterion is composed of the objective function convergence criterion, the minimum tolerance of the grid size, and the maximum number of iterations. Among them, the minimum tolerance of the grid size means that when the grid size is reduced to a very small threshold, the algorithm is considered to have converged and can be stopped.

[0053] The optimization objective function of the geometric parameter optimization design of the water inlet screen unit is:

[0054] ;

[0055] in, APLR It represents the ratio of the axial pressure drop per meter between the designed membrane module and the commercial membrane module. ; and represent the mass transfer coefficients of the designed membrane module and the commercial membrane module, respectively; Used to control the trade-off between fluid resistance and mass transfer capacity, As a penalty factor to punish excessive pressure drop.

[0056] (3) Based on the optimization results, the hydrodynamic and mass transfer characteristics are characterized through three-dimensional multi-physics simulation. Specifically, within the set computational domain, the Darcy friction factor and Sherwood number are characterized for different Reynolds numbers. Due to the periodic structure of the inlet screen, the computational domain is constructed as five continuous subunits in the y-axis direction.

[0057] Specifically, the Darcy friction factor is calculated by the following formula:

[0058] ;

[0059] in, is the fluid density, is the average flow velocity at the inlet, is the length of the calculation domain. is the hydraulic diameter, is the pressure drop over the computational domain length.

[0060] The Sherwood number is calculated as follows:

[0061] ;

[0062] is the average mass transfer coefficient of the membrane wall, which is calculated by the following formula:

[0063] ;

[0064] in, is the width of the calculation domain. is the solute concentration in the bulk of the retentate, is the solute concentration at the membrane wall.

[0065] (4) Establish a meter-scale industrial system model and use genetic algorithms to optimize the design of process parameters and membrane parameters. The details are as follows: The meter-scale industrial system model consists of a one-dimensional differential algebraic equation:

[0066] ;

[0067] Where k is the number of stages of the seawater reverse osmosis system, Q is the flow rate, is the transmembrane pressure, is the water flux, is the salinity in the main body of the retentate, is the salinity in the main body of the permeate, is the salinity at the membrane wall, is the dimensionless length, is the membrane area of ​​the kth stage, is the number of membrane elements in each pressure vessel in the kth stage, and To optimize the length of the membrane in the module perpendicular and parallel to the feed direction (x, y axis direction), is the water permeability coefficient, is the reflection coefficient, is the osmotic pressure coefficient, is the intrinsic retention rate of the membrane.

[0068] Specifically, the flow rate can be expressed as in, For the k Number of pressure vessels of each level, The number of water inlet mesh sheets for each membrane element, and are the feed channel height and feed channel porosity, respectively.

[0069] The membrane area of ​​the kth stage can be expressed as ,in, is the length of the membrane parallel to the feed direction (y-axis direction) in the commercial module, is the membrane area of ​​commercial membrane elements.

[0070] The salinity in the main body of the permeate and at the membrane wall can be expressed as ,in, is the salt permeability coefficient, is the average unit mass transfer coefficient at the membrane wall.

[0071] The optimized process and membrane parameters at meter scale include: and salt permeability coefficient , Number of pressure vessels at each level , the number of membrane elements in each pressure vessel , transmembrane pressure at each level 、Number of spacers per membrane element The optimization objective function of process and membrane parameter optimization design is:

[0072] ;

[0073] Including the annual cost of membrane and electricity cost per cubic meter of fresh water, is the required total membrane area, is the cost of the membrane per unit area, is the annual amortization factor, is the annual operating time, is the permeate flow rate, is the energy cost per kilowatt-hour, is the specific energy consumption;

[0074] Specific embodiment: Cosine function, parabola function and absolute value function are used to represent the V-like structure, which are mathematically expressed as follows:

[0075] ;

[0076] Table 1 Different trade-off factors The geometric design parameters of the water inlet screen and the corresponding calculation domain size are optimized using the hybrid intelligent optimization algorithm.

[0077] ;

[0078] Based on the conditions of inlet salinity of 35000 ppm, water recovery rate of 50%, pump efficiency of 85%, energy recovery efficiency of 95%, and concentration polarization limited to no more than 1.2, the membrane cost per unit area is calculated. , optimize the design of process and membrane parameters of meter-scale industrial system models (taking a two-stage reverse osmosis system as an example).

[0079] The specific energy consumption of the two-stage reverse osmosis system is calculated by the following formula:

[0080]

[0081] in, and Represent the flow rate of the first-stage inlet, first-stage outlet and second-stage outlet respectively, and Represent the transmembrane pressure at the first stage inlet, first stage outlet and second stage outlet respectively, Represent the hydraulic pressures at the second stage inlet and the second stage outlet, respectively. Indicates the pump efficiency of the seawater reverse osmosis system, Indicates energy recovery efficiency.

[0082] Since limiting the maximum concentration polarization factor can effectively alleviate membrane fouling and scaling, it is crucial for high permeability seawater reverse osmosis systems. , average permeation flux , Salinity of concentrated water at each stage outlet and the average inlet velocity , and the average salinity of the permeate Parameters such as RO and RO should also be within the engineering parameter range. At the same time, it is necessary to avoid negative driving force in the membrane assembly (i.e., the osmotic pressure difference exceeds the operating pressure). For multi-stage RO system (the stage number is set in this example ), when optimizing the process and membrane parameters, the constraints are summarized as follows:

[0083] ;

[0084] The concentration polarization factor is calculated by the following formula:

[0085]

[0086] First stage inlet transmembrane pressure Calculate using the following formula:

[0087]

[0088] Table 2 Design parameters of the industrial system model and their corresponding value ranges

[0089] ;

[0090] The optimization vector composed of system design parameters is obtained by genetic algorithm. The optimization design was carried out and the optimization results shown in Table 3 were obtained.

[0091] Table 3 Optimal design parameters after optimization and numerical simulation results

[0092] .

Claims

1. A multi-scale intelligent design method for reverse osmosis membrane elements for high-flux membranes, characterized in that: The following steps are involved: (1) Establish a three-dimensional multi-physics field coupled CFD model of reverse osmosis membrane elements; (2) At the submillimeter scale, a hybrid intelligent optimization algorithm combining Bayesian optimization with pattern search is used to optimize the geometric parameters of the water inlet screen unit. The hybrid intelligent optimization algorithm is as follows: Bayesian optimization is used as a global search to generate the initial optimal solution; starting from the initial optimal solution, a pattern search algorithm is used for local fine optimization; Bayesian optimization uses the expected improvement function EI as the acquisition function; the pattern search algorithm is as follows: a positive definite expansion set is used to define the search direction; the solution is updated by iteratively performing search and polling operations, and terminated when the grid size is less than a threshold; the optimization objective function is: ； in, APLR It represents the ratio of the axial pressure drop per meter between the designed membrane module and the commercial membrane module. ; and represent the mass transfer coefficients of the designed membrane module and the commercial membrane module, respectively; Used to control the trade-off between fluid resistance and mass transfer capacity, As a penalty factor to punish excessive pressure drop; (3) Based on the optimization results, characterize the hydrodynamics and mass transfer characteristics through three-dimensional multi-physics simulation; (4) Establish a meter-scale industrial system model and use genetic algorithms to optimize the design of process parameters and membrane parameters.

2. The multi-scale intelligent design method for a reverse osmosis membrane element for a high-flux membrane according to claim 1, characterized in that: The reverse osmosis membrane element comprises: a reverse osmosis unit (1) and a water production center pipe (2); a plurality of pure water diversion ports (21) are provided on the wall of the water production center pipe (2); the reverse osmosis unit (1) is wound around the circumference of the water production center pipe (2); the reverse osmosis unit (1) comprises a water inlet spacer (11), a reverse osmosis membrane (12) and a water production spacer stacked in sequence, wherein the water inlet spacer (11) is composed of a plurality of closely arranged V-shaped structure subunits.

3. The multi-scale intelligent design method for a reverse osmosis membrane element for a high-flux membrane according to claim 2, characterized in that: The cross-sectional shape of the V-like structural subunit is defined by a parabolic function, a cosine function or an absolute value function; the V-like structural subunit generates eddies in the water inlet channel to improve the mass transfer coefficient and suppress concentration polarization.

4. The multi-scale intelligent design method for a reverse osmosis membrane element for a high-flux membrane according to claim 1, characterized in that: The three-dimensional multi-physics coupled CFD model is built based on the Navier-Stokes equations and convection-diffusion.

5. The multi-scale intelligent design method for a reverse osmosis membrane element for a high-flux membrane according to claim 1, characterized in that: The meter-scale industrial system model is described by one-dimensional differential algebraic equations.

6. The multi-scale intelligent design method for a reverse osmosis membrane element for a high-flux membrane according to claim 5, characterized in that: The formula for a one-dimensional differential algebraic equation is as follows: ； Where k is the number of stages of the seawater reverse osmosis system, Q is the flow rate, is the transmembrane pressure, is the water flux, is the salinity in the main body of the retentate, is the salinity in the main body of the permeate, is the salinity at the membrane wall, is the dimensionless length, is the membrane area of ​​the kth stage, is the number of membrane elements in each pressure vessel in the kth stage, and In order to optimize the length of the membrane in the module perpendicular and parallel to the feeding direction, i.e. the x and y axis, is the water permeability coefficient, is the reflection coefficient, is the osmotic pressure coefficient, is the intrinsic retention rate of the membrane; The flow rate is expressed as in, For the k Number of pressure vessels of each level, The number of water inlet mesh sheets for each membrane element, and are the feed channel height and feed channel porosity, respectively; The membrane area of ​​the kth stage is expressed as ,in, is the length of the membrane parallel to the y-axis of the feed in the commercial module, is the membrane area of ​​commercial membrane elements; The salinity in the bulk of the permeate and at the membrane wall is expressed as ,in, is the salt permeability coefficient, is the average unit mass transfer coefficient at the membrane wall; The optimized process and membrane parameters at meter scale include: water permeability coefficient of membrane and salt permeability coefficient , Number of pressure vessels at each level , the number of membrane elements in each pressure vessel , transmembrane pressure at each level 、Number of spacers per membrane element ; And the optimization objective function of process and membrane parameter optimization design is: ； Including the annual cost of membrane and electricity cost per cubic meter of fresh water, is the required total membrane area, is the cost of the membrane per unit area, is the annual amortization factor, is the annual operating time, is the permeate flow rate, is the energy cost per kilowatt-hour, is the specific energy consumption.

7. An electronic device, characterized in that: The method comprises a memory and a processor, wherein the memory stores a computer program, and when the processor executes the program, the steps of the method according to any one of claims 1 to 6 are implemented.

Images