An optimization method based on convection-dispersion theory

Through the convection diffusion optimization algorithm, the dynamic soil-water potential and diffusion coefficient mechanism are used to solve the problem of low convergence and low efficiency of cluster intelligent algorithms in complex optimization problems, and high-precision and rapid optimization are achieved, which is suitable for a variety of optimization and engineering applications.

CN113806992BActive Publication Date: 2024-11-22任长江
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202110838632.5
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2021-07-23
Publication Date
2024-11-22
Estimated Expiration
2041-07-23

AI Technical Summary

Technical Problem

The existing cluster intelligent algorithms have problems such as difficulty in convergence analysis, lack of theoretical guidance on parameter selection, low algorithm accuracy and running time efficiency when solving complex optimization problems, especially in the nonlinear multi-parameter continuous threshold optimization problem.

Method used

A convective diffusion optimization algorithm (CDEOA) is proposed to use the search mechanism of dynamic soil and water potential weight, dynamic mechanical diffusion coefficient and dynamic molecular diffusion coefficient, and use the fitness value and extreme value of particle population to update the position, combining the driving force of soil and water potential and Mechanical and molecular diffusion mechanisms achieve rapid and precise optimization.

Benefits of technology

It has achieved high-precision and rapid convergence optimization effects, is suitable for mathematical problems such as combination optimization, constraint optimization, and multi-objective optimization, and has achieved good applications in actual engineering fields such as power systems and data clustering.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN113806992B_ABST
    Figure CN113806992B_ABST
Patent Text Reader

Abstract

The present application provides an optimization algorithm based on the convection-diffusion theory, which relates to the field of artificial intelligence optimization algorithms. The method includes five mechanisms: one is the overall migration movement of particles under the action of the hydraulic gradient of soil water potential. The second is the global search mechanism. Under the action of mechanical dispersion, particles tend to approach the optimal position of the group; the third is the local search mechanism. Under the action of mechanical dispersion, particles tend to approach the individual historical optimal position. The fourth is the soil water potential weight update mechanism. In this stage, the algorithm adjusts the size of the soil water potential with a dynamic weight mechanism. As the number of iterations gradually increases, the algorithm gradually approaches the optimal value. The fifth is the particle mutation mechanism. The algorithm controls the self-mutation of particles with the mutation probability P, enhances the diversity of the particle group, and avoids the algorithm from falling into local search too early and stagnating. Through the above search mechanism, the algorithm gradually approaches the optimal value after multiple iterations. When the algorithm meets the search stop condition, the algorithm stops searching and returns the optimal value.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present application relates to the field of engineering technology optimization, and in particular to a convection-dispersion optimization method. Background Art

[0002] In response to complex optimization problems in the field of engineering technology, many scholars choose to seek solutions from realistic models of nature, thus opening up the research of nature-inspired computing. These algorithms have good performance in adaptability, self-learning ability, robustness and efficiency. In 1993, Dario et al. studied mobile robot systems and proposed the concept of swarm intelligence, pointing out that simple ordered systems can produce non-trivial intelligent behaviors, that is, the collective wisdom generated by a group of simple individuals. Swarm intelligence has brought into play all the advantages of clusters, self-organization, decentralized control, high robustness, flexibility and low consumption, and can still provide optimal solutions when facing large-scale complex problems. The earliest swarm intelligence algorithms, including ant colony optimization algorithms and particle swarm optimization algorithms, have received widespread attention and application, and have been very successful in many fields, such as the classic traveling salesman problem. However, facing the same problem, different swarm intelligence algorithms have different performances. The relationship between the modeling method of the problem and the efficiency of the algorithm solution; the balance between the accuracy and breadth of the algorithm search are still very challenging.

[0003] Swarm intelligence algorithms currently face the following problems: First, there is a lack of a convergence analysis framework applicable to all algorithms, such as the convergence analysis method of particle swarm optimization or ant colony optimization algorithms, but it is difficult to generalize it to other algorithms. Second, for specific problems, the selection of algorithm parameters such as population size and learning rate can only rely on experience and experimental testing, without deterministic theoretical guidance. Third, for special and complex problems, the current algorithms have problems such as low algorithm accuracy and low running time efficiency. For example, the ant colony algorithm has excellent convergence when solving discrete TSP problems, but its convergence and accuracy are poor for nonlinear multi-parameter continuous threshold optimization problems. Therefore, studying swarm intelligence algorithms that can be generalized is a difficulty and hot spot in current artificial intelligence research.

[0004] In view of the problems of low algorithm accuracy and low running time efficiency of the above-mentioned swarm intelligence algorithm, the present invention proposes a convection-dispersion optimization algorithm (CDEOA) based on the mathematical theory of soil water and salt migration. CDEOA abstracts the migration and diffusion process of salt ions in the soil as an artificial intelligence algorithm proposed for finding an optimal mechanism. Summary of the invention

[0005] The present application provides a convection-diffusion optimization method, which uses dynamic soil water potential weight, dynamic mechanical diffusion coefficient and dynamic molecular diffusion coefficient search mechanism, and the algorithm has the characteristics of fast convergence and high accuracy. The method includes:

[0006] Initializing a particle group and the hydrodynamic diffusion coefficient, soil water potential and position of each particle in the particle group, and calculating the fitness value of each particle in the particle group;

[0007] For an individual particle, the fitness value of the particle is compared with the fitness value corresponding to the individual extreme value. When the fitness value of the particle is less than the fitness value corresponding to the individual extreme value of the particle, the position of the particle is taken as the current individual extreme value of the particle.

[0008] For a group of particles, the fitness value of the group of particles is compared with the fitness value corresponding to the global extreme value. When the fitness value of the group of particles is less than the fitness value corresponding to the global extreme value, the position of the particle is used as the current global extreme value.

[0009] updating the hydrodynamic diffusion coefficient, soil water potential and position of the particle according to the individual extreme value of the particle and the global extreme value;

[0010] In the current iteration, adaptively mutate the mass point at each position in the mass point group, and then return to the fitness value comparison step;

[0011] Before triggering the end condition, return to the step of comparing the fitness value of the particle with the individual extreme value of the particle, wherein the end condition is that the error between the position of each particle in the particle group and the target position is less than a preset threshold, or a preset number of iterations are completed.

[0012] Optionally, initializing the soil water potential, the hydrodynamic diffusion coefficient, and the particle position of each particle in the particle group, and calculating the fitness value of each particle in the particle group includes:

[0013] Initialize the soil water potential according to formula (1);

[0014] ψ0(i)=Rand*(ψ su (i)-ψ sl (i))+ψ sl (i) i = 1…m(1);

[0015] Initialize the hydrodynamic diffusion coefficient according to formula (2);

[0016] D lh0 (i)=ωψ0(i)+Rand*D h0(i)+Rand*D m0 (i) i=1…m(2)

[0017] Initialize the particle position according to formula (3);

[0018] x(i)=Rand*(U b (i)-L b (i))*Rand+L b (i) i = 1…m(3);

[0019] In the above formulas (1) to (3), Rand is a random number in the interval [0, 1]; m is the number of the salt ion particles; ψ0 is the initial value of the soil water potential of the particles; ψ smin is the minimum value of soil water potential at the particle; ψ smax is the maximum value of soil water potential at the particle; D lh0 is the initial value of the particle hydrodynamic diffusion coefficient; ω is the soil water potential weight coefficient; D h0 is the initial value of the particle mechanical dispersion; D m0 is the initial value of the particle molecule diffusion.

[0020] Optionally, the soil water potential (particle motion driving force mechanism) is calculated according to formula (1), and the hydrodynamic diffusion coefficient (global search mechanism and local search mechanism) is calculated according to formulas (2) and (3);

[0021] During the search process, the algorithm records the global optimal position and the individual optimal position, and obtains the global fitness minimum value and the individual fitness minimum value;

[0022] The individual optimal position searched for each individual particle is used as the individual particle extreme value, and the global optimal position searched for the group particle is used as the group particle global extreme value;

[0023] Driving force mechanism of particle motion: Under the action of soil water potential, the particle group searches in the search space according to formula (4):

[0024] ψ s =ψ m +ψ p +ψ g +ψ o (4)

[0025] In the above formula, ψ s is the total soil water potential; ψ m is the matrix potential, ψ p is the pressure potential, ψ g is the gravitational potential, ψ o is the penetration potential; in order to reduce the algorithm parameters, ψ s Instead of the overall soil water potential, ψs The faster the particle moves in the search space, the slower the particle moves in the search space;

[0026] Global search mechanism, the particle performs global search according to formula (5) and records the global optimal position information according to formula (6);

[0027] D h =λν μ (5);

[0028] D h =λν μ (g.best-x i )(6);

[0029] Among them, D h is the mechanical diffusion coefficient, λ is the mechanical dispersivity, ν is the average pore flow velocity, μ is the mechanical dispersion index, and g.best is the global extreme value; x i is the position of the i-th particle;

[0030] Molecular diffusion-local search mechanism, the particle performs local search according to formula (7) and records the local optimal position information according to formula (8);

[0031] D m =D w e βθ (7);

[0032] D m =D w αe βθ (p.best-x i )(8);

[0033] Among them, D m is the molecular diffusion coefficient, D w is the diffusion coefficient of the solute in free water, θ is the soil moisture content; α is the empirical coefficient, β is the empirical molecular diffusion index, p.best is the individual extreme value, x i is the position of the i-th particle.

[0034] Optionally, updating the hydrodynamic diffusion coefficient, soil water potential and position of the particle according to the individual extreme value of the particle and the global extreme value includes:

[0035] Update the hydrodynamic diffusion coefficient based on the mechanical diffusion coefficient and the molecular diffusion coefficient;

[0036] updating the soil water potential according to the updated hydrodynamic diffusion coefficient;

[0037] The position of the particle is updated according to the updated soil water potential.

[0038] Optionally, the hydrodynamic diffusion coefficient is updated based on the mechanical diffusion coefficient and the molecular diffusion coefficient, including:

[0039] Update the hydrodynamic dispersion coefficient according to equations (9), (10) and (11);

[0040] D h (i, t) = Rand*λν μ·log(t / MIT) (g.best t -x i )(9);

[0041]

[0042] D(i, t) = D h (i,t)+D m (i, t)(11);

[0043] Among them, D h (i, t) represents the mechanical diffusion coefficient of the i-th particle in the t-th iteration; D m (i, t) represents the molecular diffusion coefficient of the i-th particle in the t-th iteration; D(i, t) represents the hydrodynamic diffusion coefficient of the i-th particle in the t-th iteration. g.best t represents the global extreme value in the tth iteration; p.best is the current individual extreme value. Rand is a random number between [0, 1], x i represents the i-th particle, MIT is the maximum number of iterations, θ(x i , t) is the soil moisture content of the ith particle in the tth iteration; β is the empirical molecular diffusion index, μ is the mechanical dispersion index, λ is the mechanical dispersivity, ν is the average pore flow velocity, and α is the empirical coefficient;

[0044] The soil moisture content in the above formula (11) is calculated according to the following formula (12):

[0045]

[0046]

[0047] (12) In the formula, θ a is the initial moisture content; θ b is the upper boundary moisture content; D is the saturated diffusivity; K is the saturated hydraulic conductivity; t is the current iteration number; x i is the current position of the i-th particle.

[0048] Optionally, updating the soil water potential according to the updated hydrodynamic diffusion coefficient includes:

[0049] Use equation (7) to update soil water potential;

[0050] ψ s (i, t) = ω(Rand*(ψ su -ψ sl )+ψ sl )+D(i, t)(13);

[0051] Among them, ψ su represents the maximum value of soil water potential, ψ sl Represents the minimum value of soil water potential, and ω represents the soil water potential weight coefficient of the i-th particle at the current moment.

[0052] Optionally, the method further comprises:

[0053] The soil water potential weight updating mechanism is adopted to update the soil water potential weight coefficient; wherein, the soil water potential weight updating mechanism is shown in formula (14);

[0054] ω=(1+log(t / MIT) / log(MIT)(14);

[0055] MIT is the maximum number of iterations, and ω is the soil water potential weight coefficient.

[0056] Optionally, the method further comprises:

[0057] The obtained mechanical diffusion coefficient recording the global extreme value, the molecular diffusion coefficient recording the individual extreme value, the obtained updated soil water potential, and the obtained soil water potential weight update mechanism are integrated to obtain the calculation formula (15) of the soil water potential of the group particle at any number of iterations;

[0058]

[0059] ψ s (i, t+1) represents the soil water potential of the i-th particle at time t+1.

[0060] Optionally, updating the position of the mass point according to the updated soil water potential includes:

[0061] Update the position of the mass point using equation (16);

[0062] x(i, t+1)=x(i, t)+ψ s (i, t)(16);

[0063] Among them, x(i, t+1) represents the position of the i-th particle at t+1 iterations, and x(i, t) represents the position of the i-th particle at t iterations.

[0064] Optionally, in the current iteration, adaptive mutation is performed on the particle at each position in the particle group, including:

[0065] Formula (17) is used to perform adaptive variation calculation on the mass points;

[0066]

[0067] Where m is the number of particles, b is l (i) represents the lower boundary of the position of the i-th particle, b u (i) is the upper boundary of the position of the ith particle, P is the mutation probability, and x(i, j) is the position of the ith particle in the jth iteration; the Ceil function means rounding the value in the brackets toward positive infinity. For example, if the number of particles is 10, the random number generated by Rand is 0.6186. Then i=ceil(10*0.6186)=7, that is, the 7th particle among the 10 particles is updated and mutated according to the upper and lower limit constraints. The meaning of formula (17) is that when the random number generated by Rand is greater than the mutation probability P, the particle position x(i, j) is mutated, that is, the diversity of the population is expanded; otherwise, when the random number generated by rand is less than the mutation probability P, the particle position x(i, j) is not mutated.

[0068] The specific effects of the present invention are: on the one hand, compared with other methods, the present invention solves optimization problems with higher accuracy and faster convergence speed. On the other hand, it provides an optional solution for engineering optimization problems. At present, most of the global artificial intelligence algorithms are proposed by foreign technicians, such as the classic genetic algorithm, particle swarm algorithm, ant colony algorithm, neural network algorithm, etc. Domestic technicians use the algorithms proposed by foreign technicians to solve engineering problems, or improve the algorithms proposed by foreign technicians. There are few original algorithm theories. The CDEOA algorithm proposed by the present invention is theoretically original, which increases the original technical contribution of domestic technicians in the field of artificial intelligence.

[0069] Based on the mathematical theory of soil water and salt migration, the present invention proposes a convection-dispersion optimization algorithm (CDEOA). CDEOA abstracts the migration and diffusion process of salt ions in soil as an artificial intelligence algorithm proposed for finding an optimal mechanism. Based on the advantages of the above-mentioned convection-dispersion optimization method, that is, the characteristics of fast convergence speed, short time and high precision of the above-mentioned convection-dispersion optimization method, the convection-dispersion optimization method can achieve good application prospects in mathematical problems such as combinatorial optimization, constrained optimization, multi-objective optimization, dynamic system optimization, and practical engineering fields: such as power systems, filter design, automatic control, data clustering, pattern recognition and image processing, chemical industry, machinery, communication, robotics, economy, bioinformatics, medicine, task allocation, TSP and other fields. BRIEF DESCRIPTION OF THE DRAWINGS

[0070] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings required for use in the description of the embodiments of the present application will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative labor.

[0071] Figure 1 is a flowchart of the steps of the convection-dispersion optimization method proposed in the embodiment of the present application;

[0072] Figure 2 is a flow chart of a convection-dispersion optimization method in an example of the present application;

[0073] Figure 3 This is a fitness change diagram of the Bohachevsky1# function solved by the CDEOA algorithm in an example of this application;

[0074] Figure 4 This is a fitness change diagram of the Bohachevsky2# function solved by the CDEOA algorithm in an example of this application;

[0075] Figure 5 This is a fitness change graph of the Bohachevsky3# function solved by the CDEOA algorithm in an example of this application. DETAILED DESCRIPTION

[0076] The following will be combined with the drawings in the embodiments of this application to clearly and completely describe the technical solutions in the embodiments of this application. Obviously, the described embodiments are part of the embodiments of this application, not all of them. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0077] Figure 1 is a flow chart of the steps of the convection-dispersion optimization method proposed in the embodiment of the present application, such as Figure 1 As shown, the steps are as follows:

[0078] Step S11: Initializing a particle group and the hydrodynamic diffusion coefficient, soil water potential and position of each particle in the particle group, and calculating the fitness value of each particle in the particle group;

[0079] The objective function value corresponding to the position coordinates of the particle at the current moment is used as the initial fitness value of the particle.

[0080] Step S12: for an individual particle, compare the fitness value of the individual particle with the fitness value corresponding to the historical extreme value of the individual particle, and when the fitness value of the particle is less than the fitness value corresponding to the historical extreme value, take the position of the particle as the current individual extreme value;

[0081] Step S13: for a group particle, compare the current fitness value of the group particle with the fitness value corresponding to the global extreme value of the group particle; when the fitness value of the group particle is less than the fitness value corresponding to the global extreme value of the particle, take the position of the particle as the global extreme value of the group particle;

[0082] Step S14: updating the hydrodynamic diffusion coefficient, soil water potential and position of the particle according to the individual extreme value of the particle and the global extreme value;

[0083] Step S15: in the current iteration, adaptively mutate the particle at each position in the particle group;

[0084] Step S16: Repeat steps S12 to S15 until the end condition is triggered; wherein the end condition is that the difference between the position of each particle in the particle group and the target position is less than a preset threshold, or a preset number of iterations are completed.

[0085] In one example of the present application, D(i) represents the hydrodynamic diffusion coefficient of particle i, ψ s (i) represents the soil water potential of particle i, x i represents the position of particle i, F it [i] represents the fitness value, p.best(i) represents the individual extreme value, and g.best(i) represents the global extreme value or the local extreme value. The workflow of the terminal execution convection diffusion optimization method is obtained. Figure 2 is a flow chart of a convection-dispersion optimization method in an example of the present application, such as Figure 2 The process shown in FIG. 1 is to execute the related methods from step S11 to step S16.

[0086] The convection-diffusion optimization method proposed in the embodiment of the present application uses particles to represent soil salt ions, simulates the ion movement of soil salt ions under the action of convection-diffusion, and updates the position of particles by updating the soil water potential and the hydrodynamic diffusion coefficient to simulate the migration of salt ions under the driving force of soil water potential. The global optimal position information is recorded through the mechanical diffusion mechanism, and the individual optimal position information is recorded through molecular diffusion, and the optimal solution is found through iteration. The above-mentioned state soil water potential and dynamic mechanical diffusion coefficient are used as model variable parameters, combined with the dynamic molecular diffusion coefficient search mechanism to update the position of the particles, dynamically simulate the complex environment in which the particles are located in the actual environment, and reflect the mutually coupled relationship between soil moisture and solute transport. Specifically, under the action of soil water potential, salt follows water, and the action of the hydrodynamic diffusion coefficient on the particles is used to simulate the movement process of salt particles in the soil. The mechanical diffusion in hydrodynamic diffusion (the phenomenon of uneven distribution of tracer concentration in soil due to the uneven distribution of flow velocity in microscopic porous media) is abstracted into a global search mechanism, and the molecular diffusion in hydrodynamic diffusion (the material transfer phenomenon formed by the irregular thermal motion of molecules due to the uneven distribution of salt particle concentration in soil) is abstracted into a local search mechanism. Combined with the dynamic mechanical diffusion coefficient and molecular diffusion coefficient search mechanism, the global extreme value and individual extreme value of the particle are searched, which reflect the optimal migration direction and distance of the particle. Therefore, the individual extreme value and global extreme value are used to update the soil water potential and the dynamic mechanical diffusion coefficient to dynamically show the different environmental influences on the particle when it moves to different positions. Furthermore, the individual extreme value, global extreme value, soil water potential and dynamic mechanical diffusion coefficient are used to update the position of the particle at the current moment, which can quickly and accurately complete the task of updating the particle position to the target position.

[0087] Based on the advantages of the above-mentioned convection-dispersion optimization method, that is, the characteristics of fast convergence speed, short time and high precision, the convection-dispersion optimization method can achieve good application prospects in mathematical problems such as combinatorial optimization, constrained optimization, multi-objective optimization, dynamic system optimization, and practical engineering fields: such as power systems, filter design, automatic control, data clustering, pattern recognition and image processing, chemical engineering, machinery, communications, robotics, economics, bioinformatics, medicine, task allocation, TSP and other fields.

[0088] The mechanism adopted by the convection-diffusion optimization method proposed in the embodiment of the present application includes: the first mechanism: the overall migration movement under the action of soil water potential (hydraulic gradient). The larger the hydraulic gradient, the faster the migration of the salt ion particle group. This mechanism reflects the migration driving force of the particle under the action of soil water potential. The second mechanism: the global search mechanism, the movement of the particle between the large gaps of the soil particles under the action of mechanical diffusion. This stage reflects the memory of the particle of the group particle movement trajectory, which represents the tendency of the particle to approach the historical best position of the group. The larger the mechanical diffusion coefficient, the stronger the global exploration ability, and vice versa. The third mechanism: the local search mechanism, the molecular diffusion effect caused by the uneven salt ion concentration of the particle, this stage reflects the memory of the individual particle's own movement trajectory, which represents the tendency of the individual to approach its own historical best position. The larger the molecular diffusion coefficient, the stronger the local exploration ability, and vice versa. The fourth mechanism: the soil water potential weight update mechanism. In this stage, the algorithm adjusts the size of the soil water potential with a dynamic weight mechanism. As the number of iterations gradually increases, the algorithm gradually approaches the optimal value. The fifth mechanism is the particle mutation mechanism. In order to enhance the diversity of the particle group and prevent the algorithm from falling into local search too early and stagnating, the particle’s own mutation is controlled by the mutation probability P.

[0089] In the local search process, the individual optimal position is recorded by comparing the fitness values ​​to obtain the individual optimal solution; in the global search process, the global optimal position is recorded by comparing the individual optimal fitness values ​​to obtain the global optimal solution.

[0090] Through the first mechanism and the global search mechanism, each particle moves in the direction of the minimum soil water potential, so that the soil water potential can be used to control the movement direction and distance of the current particle. The global optimal value and individual optimal value are recorded through mechanical diffusion and molecular diffusion, and then the position of the particle at the next moment is updated according to the current global extreme value and individual extreme value, so as to determine the optimal path for the particle migration.

[0091] Initializing the soil water potential, hydrodynamic diffusion coefficient and particle position of each particle in the particle group, and calculating the fitness value of each particle in the particle group includes:

[0092] Initialize the soil water potential according to formula (1);

[0093] ψ0(i)=Rand*(ψ su (i)-ψ sl (i))+ψ sl (i) i = 1…m(1);

[0094] Initialize the hydrodynamic diffusion coefficient according to formula (2);

[0095] D lh0 (i)=

[0096] ωψ0(i)+Rand*D h0 (i)+Rand*D m0 (i) i = 1…m(2);

[0097] Initialize the particle position according to formula (3);

[0098] x(i)=Rand*(U b (i)-L b (i))*Rand+

[0099] L b (i) i = 1…m(3);

[0100] In the above formulas (1) to (3), Rand is a random number in the interval [0, 1]; m is the number of the salt ion particles; ψ0 is the initial value of the soil water potential of the particles; ψ smin is the minimum value of soil water potential at the particle; ψ smax is the maximum value of soil water potential at the particle; D lh0 is the initial value of the particle hydrodynamic diffusion coefficient; ω is the soil water potential weight coefficient; D h0 is the initial value of the particle mechanical dispersion; D m0 is the initial value of the particle molecule diffusion.

[0101] The method for obtaining soil water potential is shown in formula (4):

[0102] ψ s =ψ m +ψ p +ψ g +ψ o (4);

[0103] Among them, ψ s is the total soil water potential; is ψ m The matrix potential is ψ p The pressure potential is ψ g Gravitational potential, ψ o is the seepage potential. The soil water potential can be regarded as the driving force of particle movement in the process of convection-diffusion optimization, ψ s The larger the size, the stronger the global exploration ability, and vice versa.

[0104] The local search mechanisms adopted in the embodiments of the present application include: mechanical diffusion-local search mechanism and molecular diffusion-local search mechanism.

[0105] The mechanical diffusion-local search mechanism uses formula (5) to perform global search and records the global optimal position information according to formula (6).

[0106] D h=λν μ (5);

[0107] D h =λν μ (g.best-x i )(6);

[0108] Among them, D h is the mechanical diffusion coefficient, λ is the mechanical dispersivity, ν is the average pore flow velocity, μ is the mechanical diffusion index. When the average pore flow velocity ν is greater than 1, the global search ability of the mechanical diffusion index μ is stronger, otherwise it is weaker. g.best is the global extreme value; x i is the position of the i-th particle.

[0109] The molecular diffusion-local search mechanism uses formula (7) to perform local search and records the local optimal position information according to formula (8).

[0110] D m =D w e βθ (7);

[0111] D m =D w αe βθ (p.best-x i )(8);

[0112] Among them, D m is the molecular diffusion coefficient, D w is the diffusion coefficient of the solute in free water, θ is the soil moisture content; β is the empirical molecular diffusion index. p.best is the individual extreme value, x i is the position of the i-th particle.

[0113] Another embodiment of the present application updates the soil water potential according to the mechanical diffusion coefficient and the molecular diffusion coefficient. Specifically, according to the individual extreme value of the particle and the global extreme value or the local extreme value, the hydrodynamic diffusion coefficient, soil water potential and position of the particle are updated, including: first updating the hydrodynamic diffusion coefficient according to the mechanical diffusion coefficient and the molecular diffusion coefficient;

[0114] updating the soil water potential according to the updated hydrodynamic diffusion coefficient;

[0115] The position of the particle is updated according to the updated soil water potential.

[0116] Updates the hydrodynamic diffusion coefficient based on the mechanical diffusion coefficient and the molecular diffusion coefficient, including:

[0117] The hydrodynamic dispersion coefficient is updated using equations (9), (10) and (11);

[0118] Dh (i, t) = R1λν μ·log(t / MIT) (g.best t -x i )(9);

[0119]

[0120]

[0121] D(i, t) = D h (i,t)+D m (i, t)(11);

[0122] Among them, D h (i, t) represents the mechanical diffusion coefficient of the i-th particle in the t-th iteration; D m (i, t) represents the molecular diffusion coefficient of the i-th particle in the t-th iteration; D(i, t) represents the hydrodynamic diffusion coefficient of the i-th particle in the t-th iteration. g.best t represents the global or local extreme value in the tth iteration; p.best is the current individual extreme value. R1 and R2 are random numbers between [0, 1], x i represents the i-th particle, MIT is the maximum number of iterations, θ is the soil moisture content; β is the empirical molecular diffusion index, μ is the mechanical dispersion index, λ is the mechanical dispersivity, ν is the average pore flow velocity, and α is the empirical coefficient.

[0123] The soil moisture content in the above formula (11) is calculated according to the following formula (12):

[0124]

[0125] (12) In the formula, θ a is the initial moisture content; θ b is the upper boundary moisture content; D is the saturated diffusivity; K is the saturated hydraulic conductivity; t is the current iteration number; x i is the current position of the i-th particle.

[0126] The present application uses an example to search for the optimal path of an infiltration solution composed of m particles in an n-dimensional target search space, and represents the i-th particle as an n-dimensional vector, denoted as: i =(x i1 , x i2 ,......x in ), i = 1, 2...m. The hydrodynamic diffusion coefficient of the i-th particle is expressed as an m-dimensional vector, denoted by D h =(D i1 , D i2 ,......D im), i = 1, 2...n. The optimal position searched by the i-th particle is called the individual extreme value, denoted by p.best = (p i1 , p i2 ,......p im ), i = 1, 2...n. The global extreme value searched by the particle group is recorded as g.best = (p g1 , p g2 ,......p gm ).

[0127] The soil water potential is updated according to the updated hydrodynamic dispersion coefficient, including:

[0128] Use equation (13) to update soil water potential;

[0129] ψ s (i, t) = ω(Rand*(ψ su -ψ sl )+ψ sl )+D(i, t)(13);

[0130] Among them, ψ su represents the maximum value of soil water potential, ψ sl Represents the minimum value of soil water potential. ω represents the soil water potential weight coefficient of the i-th particle at the current moment.

[0131] Other embodiments of the present application adopt a soil water potential weight update mechanism to update the soil water potential weight coefficient to simulate the dynamic changes of soil water potential in the overall environment, so as to improve the algorithm optimization accuracy and convergence speed. The soil water potential weight coefficient is updated using formula (14).

[0132] ω=(1+log(t / MIT) / log(MIT)(14);

[0133] Among them, MIT is the maximum number of iterations and ω is the soil water potential weight coefficient.

[0134] Furthermore, formula (14) for updating the soil water potential weight coefficient, formula (9) for obtaining the mechanical diffusion coefficient according to the global extreme value, and formula (10) for obtaining the molecular diffusion coefficient according to the individual extreme value are substituted into formula (13) to obtain the specific formula (15) for calculating the soil water potential at the next moment after the current moment.

[0135]

[0136] ψ s (i, t+1) represents the soil water potential of the i-th particle at time t+1.

[0137] According to formula (15), there are three factors that affect the soil water potential renewal. The first part (1+log(t / MIT) / log(MIT)*(Rand*(ψ su -ψ sl )+ψ sl ) is the soil water potential part, which reflects the driving force of the particle movement. Its size affects the speed of the particle movement. The larger its value, the faster the algorithm converges, and the corresponding accuracy is also poor. On the contrary, the slower the convergence, the better the accuracy. The second part Rand*λν μ·log(t / MIT) (g.best t -x i ) is the mechanical diffusion part, which reflects the memory of the movement path of the salt ion group particles, representing the tendency of the particles to approach the historical best position; the third part This is the molecular diffusion part, which reflects the random Brownian motion of the particle toward the neighborhood space. This stage reflects the tendency of the particle to approach its individual historical best position.

[0138] The hydrodynamic diffusion coefficient, soil water potential and position of the particle are updated according to the individual extreme value of the particle and the global extreme value or the local extreme value, including:

[0139] The updated soil water potential at the current moment is added to the current position of the mass point to obtain the updated position. The specific method for calculating the updated position is shown in formula (16):

[0140] x(i, t+1)=x(i, t)+ψ s (i, t)(16);

[0141] Among them, x(i, t+1) represents the position of the i-th particle at time t+1, and x(i, t) represents the position of the i-th particle at time t.

[0142] Another embodiment of the present application proposes a method for adaptively mutating a particle at each position in a particle group. Adaptive mutation of a particle may refer to simulating the change in the position of a particle due to external forces in an actual environment. In the current iteration, adaptive mutation of a particle at each position in the particle group includes:

[0143] Formula (17) is used to perform adaptive variation calculation on the particles.

[0144]

[0145] Where n is the number of particles, b l (i) represents the lower boundary of the position of the i-th particle, b u (i) is the upper boundary of the position of the i-th particle, P is the mutation probability, and x(i, j) is the position of the i-th particle at the j-th iteration.

[0146] In one example of the present application, the Goldstein-Price function and the Branin RCOS function are used to calculate the fitness value.

[0147] In order to verify the superiority of the algorithm described in the present invention, five algorithms, including particle swarm optimization (PSO), vulture algorithm (BES), grey wolf algorithm (GWO), moth-to-fire algorithm (MFO) and convection-dispersion algorithm (CDEOA) described in the present invention, are used to optimize and solve the following three international general test functions. The other four algorithm parameters are built-in optimal parameters. The parameters of the CDEOA algorithm described in the present invention are: weight coefficient = 0.05, mutation probability = 0.976, initial water content = 0.15, saturated water content = 0.05, saturated diffusivity = 0.0005, saturated hydraulic conductivity = 0.0005, mechanical dispersion = 0.5, average pore flow velocity = 0.95, dispersion index = 1.5, solute diffusion coefficient = 1.5, diffusion coefficient index = 2, minimum water potential gradient = -2.5, maximum water potential gradient = 2.5. The population number of the five algorithms is 50, and the number of iterations is 100. The application examples are as follows:

[0148] Application Example 1: Bohachevskyl# function;

[0149]

[0150] The Bohachevskyl# function has a global minimum value of 0 at (x1, x2) = (0, 0). The fitness of the five optimization algorithms varies with the number of iterations as follows: Figure 3 As shown. Figure 3 It can be seen that for the Bohachevskyl# function, the convergence speeds of the five algorithms are from fast to slow as follows: CDEOA>GWO>PSO>BES>MFO, among which the CDEOA algorithm of the present invention has the fastest convergence speed and MFO converges the slowest.

[0151] Application Example 2: Bohachevsky2# function;

[0152]

[0153] The Bohachevsky2# function has a global minimum value of 0 at (x1, x2) = (0, 0). The fitness of the five optimization algorithms varies with the number of iterations as follows: Figure 4 As shown. Figure 4 It can be seen that for the Bohachevsky2# function, the convergence speeds of the five algorithms are from fast to slow as follows: CDEOA>GWO>BES>PSO>MFO, among which the CDEOA algorithm of the present invention has the fastest convergence speed and MFO converges the slowest.

[0154] Application Example 3: Bohachevsky3# function;

[0155]

[0156] The Bohachevsky3# function has a global minimum value of 0 at (x1, x2) = (0, 0). The fitness of the five optimization algorithms varies with the number of iterations as follows: Figure 5 As shown. Figure 5 It can be seen that for the Bohachevsky3# function, the convergence speeds of the five algorithms are from fast to slow as follows: CDEOA>GWO>PSO>BES>MFO, among which the CDEOA algorithm described in the present invention has the fastest convergence speed and MFO converges the slowest.

[0157] In summary, through three nonlinear optimization function optimization tests, it is shown that the CDEOA optimization algorithm proposed in the present invention has good convergence speed and accuracy.

[0158] As for the device embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment.

[0159] Each embodiment in this specification is described in a progressive or illustrative manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the embodiments can be referenced to each other.

[0160] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, devices, or computer program products. Therefore, the present application may be provided in the form of a complete hardware embodiment, a complete software embodiment, or a combination of software and hardware embodiments. Furthermore, the present application may be provided in the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0161] The embodiments of the present application are described by flowcharts and / or block diagrams of the methods, devices, and computer program products of the embodiments of the present application. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing terminal device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing terminal device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1A device that provides the functions specified in a block or multiple blocks.

[0162] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing terminal device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce a manufactured product including an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0163] These computer program instructions can also be loaded onto a computer or other programmable data processing terminal device so that a series of operating steps are executed on the computer or other programmable terminal device to produce a computer-implemented process, thereby providing instructions for executing on the computer or other programmable terminal device to implement the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0164] Although the preferred embodiments of the present application have been described, those skilled in the art may make additional changes and modifications to these embodiments once they are aware of the basic creative concepts. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications that fall within the scope of the embodiments of the present application.

[0165] Finally, it should be noted that, in this article, relational terms such as first and second, etc. are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process or method including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process or method. In the absence of further restrictions, the elements defined by the sentence "includes a..." do not exclude the existence of other identical elements in the process or method including the elements.

[0166] The above is a detailed introduction to a convection-dispersion optimization method provided by the present application. The description of the above embodiments is only used to help understand the method of the present application and its core idea. At the same time, for a person skilled in the art, according to the idea of ​​the present application, there may be changes in the specific implementation method and application scope. In summary, the content of this specification should not be understood as a limitation on the present application.

Claims

1. A convection-dispersion optimization method, characterized in that , the method comprises: Initializing a particle group and the hydrodynamic diffusion coefficient, soil water potential and position of each particle in the particle group, and calculating the fitness value of each particle in the particle group; For an individual particle, the fitness value of the particle is compared with the fitness value corresponding to the individual extreme value. When the fitness value of the particle is less than the fitness value corresponding to the individual extreme value of the particle, the position of the particle is taken as the individual extreme value of the current particle. For a group of particles, the fitness value of the group of particles is compared with the fitness value corresponding to the global extreme value, and when the fitness value of the group of particles is less than the fitness value corresponding to the global extreme value, the position of the particle is used as the current global extreme value; According to the individual extreme value of the particle and the global extreme value, the hydrodynamic diffusion coefficient, soil water potential and position of the particle are updated; In the current iteration, adaptively mutate the mass point at each position in the mass point group, and then return to the fitness value comparison step; Before triggering the end condition, return to the step of comparing the fitness value of the particle with the individual extreme value of the particle, wherein the end condition is that the error between the position of each particle in the particle group and the target position is less than a preset threshold, or a preset number of iterations are completed; Wherein, the updating of the hydrodynamic diffusion coefficient, soil water potential and position of the particle according to the individual extreme value of the particle and the global extreme value includes: Update the hydrodynamic diffusion coefficient based on the mechanical diffusion coefficient and the molecular diffusion coefficient; updating the soil water potential according to the updated hydrodynamic diffusion coefficient; The position of the particle is updated according to the updated soil water potential.

2. The method according to claim 1, characterized in that , initializing the soil water potential, hydrodynamic diffusion coefficient and particle position of each particle in the particle group, and calculating the fitness value of each particle in the particle group includes: Initialize the soil water potential according to formula (1); (1); Initialize the hydrodynamic diffusion coefficient according to formula (2); (2) Initialize the particle position according to formula (3); (3); The above formulas (1) to (3), is a random number in the interval [0,1]; is the number of salt ion particles; is the initial value of soil water potential of the particle; is the minimum value of soil water potential of the particle; is the maximum value of soil water potential at the particle; The maximum value of soil water potential Indicates the minimum value of soil water potential; is the initial value of the hydrodynamic diffusion coefficient of the particle; is the soil water potential weight coefficient; is the initial value of the mechanical dispersion of the particle; is the initial value of the particle molecule diffusion.

3. The method according to claim 2, characterized in that , including: calculating soil water potential, i.e., the particle motion driving force mechanism, according to formula (1), calculating the hydrodynamic diffusion coefficient, i.e., the global search mechanism and the local search mechanism, according to formulas (2) and (3); During the search process, the algorithm records the global optimal position and the individual optimal position, and obtains the global minimum fitness value and the individual minimum fitness value; The individual optimal position searched for each individual particle is used as the individual particle extreme value, and the global optimal position searched for the group particle is used as the group particle global extreme value; Driving force mechanism of particle motion: Under the action of soil water potential, the particle group searches in the search space according to formula (4): (4) In the above formula is the total soil water potential; Matrix potential Pressure potential Gravity potential is the osmotic potential; in order to reduce the algorithm parameters, Instead of overall soil water potential The faster the particle moves in the search space, the slower the particle moves in the search space; Global search mechanism, the particle performs global search according to formula (5) and records the global optimal position information according to formula (6); (5); (6); in Mechanical diffusion coefficient Mechanical diffusion is the average pore flow velocity Mechanical diffusion index is the global extreme value; For the i Particle position; Molecular diffusion-local search mechanism, the particle performs local search according to formula (7) and records the local optimal position information according to formula (8); (7); (8); in is the molecular diffusion coefficient is the diffusion coefficient of the solute in free water is the soil moisture content; is the empirical coefficient is the empirical molecular diffusion index, Individual extreme value For the i Particle position.

4. The method according to claim 1, characterized in that , update the hydrodynamic diffusion coefficient based on the mechanical diffusion coefficient and the molecular diffusion coefficient, including: Update the hydrodynamic dispersion coefficient according to equations (9), (10) and (11); (9); (10); (11); in Indicates In the iteration The mechanical diffusion coefficient of each particle; Indicates In the iteration The molecular diffusion coefficient of each particle; Indicates Second in The hydrodynamic diffusion coefficient of a particle is is the diffusion coefficient of the solute in free water; Indicates The global extremum in the iteration; is the current individual extreme value; A random number between [0,1] Indicates Mass is the maximum number of iterations For the Particle Soil moisture content in iterations; is the empirical molecular diffusion index Mechanical diffusion index Mechanical diffusion is the average pore flow velocity is the empirical coefficient; The soil moisture content in the above formula (11) is calculated according to the following formula (12): (12); (12) In the formula, is the initial moisture content; is the upper boundary moisture content; is the saturation diffusivity; is the saturated hydraulic conductivity; is the current iteration number; For the current The position of a particle.

5. The method according to claim 4, characterized in that , based on the updated hydrodynamic diffusion coefficient, updating the soil water potential, including: Use equation (13) to update soil water potential; (13); in, The maximum value of soil water potential represents the minimum value of soil water potential, Indicates the current moment The soil water potential weight coefficient of each particle.

6. The method according to claim 5, characterized in that , the method further comprises: The soil water potential weight updating mechanism is adopted to update the soil water potential weight coefficient; the soil water potential weight updating mechanism is shown in formula (14); (14); is the maximum number of iterations is the soil water potential weight coefficient.

7. The method according to claim 6, characterized in that , the method further comprises: The mechanical diffusion coefficient that records the global extreme value, the molecular diffusion coefficient that records the individual extreme value, the updated soil water potential, and the soil water potential weight update mechanism are integrated to obtain the calculation formula (15) of the soil water potential of the group of particles at any number of iterations; (15); Indicates The mass point The soil and water potential at each moment.

8. The method according to claim 7, characterized in that , according to the updated soil water potential, updating the position of the particle, including: Update the position of the mass point using equation (16); (16); in Indicates The mass point The position at the iteration number Indicates The mass point The position at the iteration number.

9. The method according to claim 1, characterized in that ,In the current iteration, adaptive mutation is performed on the particle at each position in the particle group, including: Formula (17) is used to perform adaptive variation calculation on the mass points; (17); in is the number of particles Indicates The lower boundary of the particle position For the The upper boundary of the particle position, P is the mutation probability For the Iteration No. The ceil function means rounding the value in the brackets toward positive infinity. The meaning of formula (17) is when Rand The random number generated is greater than the mutation probability P , then the particle position To mutate, that is, to expand the diversity of the population; otherwise Rand The random number generated is less than the mutation probability P , particle position No mutation is performed.