CN115906626A - Underground water organic pollution mutual-feedback inversion tracing method based on artificial intelligence - Google Patents
Underground water organic pollution mutual-feedback inversion tracing method based on artificial intelligence Download PDFInfo
- Publication number
- CN115906626A CN115906626A CN202211425201.7A CN202211425201A CN115906626A CN 115906626 A CN115906626 A CN 115906626A CN 202211425201 A CN202211425201 A CN 202211425201A CN 115906626 A CN115906626 A CN 115906626A
- Authority
- CN
- China
- Prior art keywords
- model
- inversion
- sample set
- training sample
- optimization
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 83
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 title claims abstract description 37
- 238000013473 artificial intelligence Methods 0.000 title claims abstract description 25
- 238000005457 optimization Methods 0.000 claims abstract description 89
- 238000012549 training Methods 0.000 claims abstract description 57
- 238000004088 simulation Methods 0.000 claims abstract description 27
- 238000006467 substitution reaction Methods 0.000 claims abstract description 27
- 238000004422 calculation algorithm Methods 0.000 claims abstract description 21
- 238000012544 monitoring process Methods 0.000 claims abstract description 21
- 238000012360 testing method Methods 0.000 claims abstract description 11
- 230000008878 coupling Effects 0.000 claims abstract description 6
- 238000010168 coupling process Methods 0.000 claims abstract description 6
- 238000005859 coupling reaction Methods 0.000 claims abstract description 6
- 239000003344 environmental pollutant Substances 0.000 claims description 64
- 231100000719 pollutant Toxicity 0.000 claims description 64
- 238000013508 migration Methods 0.000 claims description 60
- 230000005012 migration Effects 0.000 claims description 59
- 230000008569 process Effects 0.000 claims description 36
- 239000002245 particle Substances 0.000 claims description 12
- 239000008346 aqueous phase Substances 0.000 claims description 11
- 239000000356 contaminant Substances 0.000 claims description 11
- 230000004044 response Effects 0.000 claims description 11
- 239000006185 dispersion Substances 0.000 claims description 10
- 239000003673 groundwater Substances 0.000 claims description 10
- 238000005070 sampling Methods 0.000 claims description 8
- 238000006243 chemical reaction Methods 0.000 claims description 6
- 230000035699 permeability Effects 0.000 claims description 6
- 239000012071 phase Substances 0.000 claims description 5
- 238000007689 inspection Methods 0.000 claims description 4
- 238000012937 correction Methods 0.000 description 25
- 239000013598 vector Substances 0.000 description 20
- 230000006870 function Effects 0.000 description 16
- 238000004364 calculation method Methods 0.000 description 15
- 238000003911 water pollution Methods 0.000 description 14
- 238000011109 contamination Methods 0.000 description 11
- 125000004122 cyclic group Chemical group 0.000 description 8
- 238000009826 distribution Methods 0.000 description 8
- 238000004458 analytical method Methods 0.000 description 5
- 230000000694 effects Effects 0.000 description 5
- 238000010276 construction Methods 0.000 description 4
- 210000002569 neuron Anatomy 0.000 description 4
- 239000002957 persistent organic pollutant Substances 0.000 description 3
- MVPPADPHJFYWMZ-UHFFFAOYSA-N chlorobenzene Chemical compound ClC1=CC=CC=C1 MVPPADPHJFYWMZ-UHFFFAOYSA-N 0.000 description 2
- 150000001875 compounds Chemical class 0.000 description 2
- 238000012790 confirmation Methods 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 230000009189 diving Effects 0.000 description 2
- 230000009977 dual effect Effects 0.000 description 2
- 238000003895 groundwater pollution Methods 0.000 description 2
- 238000011835 investigation Methods 0.000 description 2
- 238000013507 mapping Methods 0.000 description 2
- 239000011159 matrix material Substances 0.000 description 2
- 238000005067 remediation Methods 0.000 description 2
- 230000008439 repair process Effects 0.000 description 2
- 238000012502 risk assessment Methods 0.000 description 2
- 238000012546 transfer Methods 0.000 description 2
- 230000003044 adaptive effect Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000008901 benefit Effects 0.000 description 1
- 238000005314 correlation function Methods 0.000 description 1
- 230000007423 decrease Effects 0.000 description 1
- 230000006866 deterioration Effects 0.000 description 1
- 238000009792 diffusion process Methods 0.000 description 1
- 238000006073 displacement reaction Methods 0.000 description 1
- 238000004090 dissolution Methods 0.000 description 1
- 239000003651 drinking water Substances 0.000 description 1
- 235000020188 drinking water Nutrition 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 230000005284 excitation Effects 0.000 description 1
- 239000012530 fluid Substances 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 238000005259 measurement Methods 0.000 description 1
- 238000010606 normalization Methods 0.000 description 1
- 210000004205 output neuron Anatomy 0.000 description 1
- 230000002085 persistent effect Effects 0.000 description 1
- 230000002028 premature Effects 0.000 description 1
- 238000004445 quantitative analysis Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 230000000452 restraining effect Effects 0.000 description 1
- 230000000717 retained effect Effects 0.000 description 1
- 230000002123 temporal effect Effects 0.000 description 1
- 230000009466 transformation Effects 0.000 description 1
- 230000007723 transport mechanism Effects 0.000 description 1
Images
Classifications
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02A—TECHNOLOGIES FOR ADAPTATION TO CLIMATE CHANGE
- Y02A20/00—Water conservation; Efficient water supply; Efficient water use
- Y02A20/20—Controlling water pollution; Waste water treatment
Landscapes
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses an underground water organic pollution mutual-feed inversion tracing method based on artificial intelligence, which comprises the following steps: acquiring a numerical simulation model of underground water organic pollution multiphase flow based on the monitoring data; acquiring a variable to be identified, and acquiring a training sample set and a test sample set based on the value range of the variable to be identified; based on the training sample set, obtaining an intelligent substitution model of the multi-phase flow numerical model by coupling an artificial intelligence method; obtaining an inversion optimization model based on the intelligent substitution model of the multiphase flow numerical model and the nonlinear programming optimization model; carrying out inversion recognition on the training sample set through a homotopy transformation-group intelligent optimization algorithm to obtain an inversion recognition result, and updating the training sample set based on the inversion recognition result; returning to S4 until a final training sample set is obtained; and re-establishing an intelligent substitution model of the multiphase flow numerical model to obtain an updated inversion identification result.
Description
Technical Field
The invention belongs to the field of research on underground water environment inverse problems, and particularly relates to an underground water organic pollution mutual-feed inversion tracing method based on artificial intelligence.
Background
Organic pollutants released by accidents, stealing, leakage and the like in the petrochemical production process can enter a groundwater system in a non-aqueous phase fluid form and are accumulated and retained in a water-bearing stratum, so that serious and persistent pollution is caused, and great threats are generated to regional and downstream ecological environments and drinking water safety of residents. The method has the advantages that the groundwater quality condition of the organic pollution site is accurately mastered, the spatial and temporal distribution and evolution rules of pollutants are predicted, the method is a basic premise for formulating a scientific groundwater resource management strategy and constructing an efficient pollution remediation project, and the method has great significance for effectively restraining continuous deterioration of groundwater environment.
The numerical simulation technology is an important tool for quantitative analysis of the time-space characteristics of the groundwater water quality, however, when numerical modeling is carried out on a field, due to the concealment of a pollution source, assignment of source item characteristics has strong subjectivity and uncertainty, and the actual process of pollutant migration is difficult to accurately reproduce. Therefore, it is very important to efficiently and accurately solve the problem of groundwater pollution inversion traceability. For this problem, the inversion calculation using actual observation data combined with data assimilation method is the main solution at present.
Disclosure of Invention
The invention provides an artificial intelligence-based underground water organic pollution mutual-feed inversion tracing method which is used for identifying characteristics of an underground water organic pollution source, and assisting in improving the accuracy of inversion identification of the characteristics of the pollution source by estimating key parameters which are involved in the migration process of organic pollutants in an underground water system and are difficult to directly obtain through the conventional measurement means, so that the accurate simulation and prediction of the migration process of the underground water organic pollution multiphase flow are realized, and important precondition basic conditions are provided for underground water pollution risk assessment, pollution source control removal and responsibility confirmation, and efficient repair engineering construction.
In order to achieve the purpose, the invention provides an underground water organic pollution mutual-feed inversion tracing method based on artificial intelligence, which comprises the following steps:
s1, acquiring monitoring data, and acquiring a numerical simulation model of underground water organic pollution multiphase flow based on the monitoring data;
s2, acquiring a variable to be recognized, and acquiring a training sample set and a test sample set based on the value range of the variable to be recognized;
s3, based on the training sample set, obtaining an intelligent substitution model of the multi-phase flow numerical model by coupling an artificial intelligence method;
s4, constructing a nonlinear programming optimization model, and obtaining an inversion optimization model based on an intelligent substitution model of the multiphase flow numerical model and the nonlinear programming optimization model;
s5, based on the inversion optimization model, performing inversion identification on the training sample set through a homotopy transformation-group intelligent optimization algorithm to obtain an inversion identification result, and updating the training sample set based on the inversion identification result;
s6, returning to S4 until a final training sample set is obtained;
and S7, reestablishing an intelligent substitution model of the multiphase flow numerical model based on the final training sample set to obtain an updated inversion identification result.
Optionally, the variable to be identified includes a pollution source characteristic and a pollutant migration parameter;
the pollution source characteristics include: longitudinal coordinates of the pollution source, transverse coordinates of the pollution source, migration and conversion time length of the pollutants and leakage amount of the pollutants;
the contaminant migration parameters include: porosity, permeability, longitudinal aqueous phase dispersion and transverse aqueous phase dispersion.
Optionally, the method for obtaining the training sample set and the test sample set includes:
collecting a plurality of groups of samples based on the variable to be identified;
inputting a plurality of groups of samples into the underground water organic pollution multiphase flow numerical simulation model to obtain a plurality of model responses;
based on the plurality of model responses, a training sample set and a test sample set are obtained.
Optionally, the method for obtaining the intelligent surrogate model of the multi-phase flow numerical model by coupling the artificial intelligence method based on the training sample set includes:
constructing a Gaussian process model and a kernel limit learning machine model by the artificial intelligence method based on the training sample set;
acquiring the weighting weight and key parameters of the Gaussian process model and the kernel-limit learning machine model, and optimizing the Gaussian process model and the kernel-limit learning machine model by adopting a particle swarm optimization algorithm to obtain the optimal values of the weighting weight and the key parameters;
an intelligent surrogate model of the multiphase flow numerical model is obtained based on the weighted weights and the optimal values of the key parameters.
Optionally, the inversion optimization model includes a pollution source characteristic inversion optimization model and an aquifer parameter inversion optimization model.
Optionally, obtaining the inversion recognition result includes:
rewriting the pollution source characteristic inversion optimization model based on homotopy theory to obtain a plurality of homotopy optimization models;
solving the homotopy optimization models based on a swarm intelligence optimization algorithm to obtain an inversion identification result of the pollution source characteristics;
and acquiring an inversion identification result of the pollutant migration parameters according to a homotopy transformation-group intelligent optimization algorithm based on the inversion identification result of the pollution source characteristics and the aquifer parameter inversion optimization model.
Optionally, updating the training sample set based on the inversion recognition result includes:
and returning the inversion identification result of the pollution source characteristics and the inversion identification result of the pollutant migration parameters to the training sample set, and updating the training sample set.
Optionally, the method for obtaining the final training sample set includes: and obtaining the final training sample set by adopting an improved self-adaptive updating sampling method based on the inversion identification result.
The invention has the following beneficial effects:
(1) According to the invention, an intelligent substitution model of the numerical model of the underground water organic pollution multiphase flow is established, and the calculation efficiency of inversion is improved.
(2) According to the invention, a mutual-feed cyclic correction inversion system for respectively and independently carrying out optimization identification on the characteristics of the pollution source and the migration parameters of the pollutants is constructed, so that the identification results of the pollution source and the migration parameters of the pollutants are mutually promoted and gradually improved, and the calculation accuracy of inversion is improved.
(3) The method solves the problems of identification of characteristics of the underground water organic pollution source and correction of key parameters in the multiphase flow migration model, thereby realizing accurate simulation and prediction of the space-time distribution of the underground water organic pollution and providing important precondition basic conditions for underground water pollution risk assessment, pollution source control removal and responsibility confirmation and high-efficiency repair engineering construction; meanwhile, the invention can provide an effective means for similar multi-type variable identification problems, and avoids the effect of 'different participation and same effect' to the maximum extent.
Drawings
The accompanying drawings, which are incorporated in and constitute a part of this application, illustrate embodiments of the application and, together with the description, serve to explain the application and are not intended to limit the application. In the drawings:
FIG. 1 is a flow chart of an embodiment of the present invention of a groundwater contamination source characteristic and contaminant migration parameter cross-feed inversion system;
FIG. 2 is a diagram of the relative positions of the generalized site and water quality monitoring wells according to the second embodiment of the present invention;
FIG. 3 is a fitting scatter diagram of the inspection output of the intelligent surrogate model and the output of the numerical simulation model, which are established in the second step 4 and the second step 5 of the embodiment of the present invention;
FIG. 4 is a convergence curve of different variable identification values in the iterative loop solution process of feedback correction according to the second embodiment of the present invention;
FIG. 5 is a comparison graph of an actual pollution plume and an identified pollution plume in accordance with an embodiment of the present invention, wherein (a) -the actual pollution plume and (b) -the identified pollution plume after being modified by a cross-feed loop.
Detailed Description
It should be noted that, in the present application, the embodiments and features of the embodiments may be combined with each other without conflict. The present application will be described in detail below with reference to the accompanying drawings in conjunction with embodiments.
It should be noted that the steps illustrated in the flowcharts of the figures may be performed in a computer system such as a set of computer-executable instructions and that, although a logical order is illustrated in the flowcharts, in some cases, the steps illustrated or described may be performed in an order different than presented herein.
Example one
As shown in fig. 1, the present embodiment provides an underground water organic pollution mutual-feed inversion tracing method based on artificial intelligence, including:
step 1: the hydrogeological conditions of the polluted site are mastered by means of field on-site investigation, dynamic monitoring and the like, and groundwater water quality dynamic monitoring data (pollutant concentration data) are obtained.
Step 2: carrying out generalization treatment on the hydrogeological conditions of the site, and primarily establishing a multiphase flow numerical simulation model of the organic pollution of the underground water so as to describe the transport mechanism of the organic pollutants in the underground water.
And step 3: and determining the pollution source characteristics to be identified and the pollutant migration parameters in the pollution tracing problem. And (3) randomly sampling a plurality of groups of samples according to the value range of the variable to be identified in the model, and substituting the samples into the multiphase flow numerical model established in the step (2) one by one to obtain model responses corresponding to each group of samples, thereby forming a training sample set and an inspection sample set which are formed by 'model input-model response' sample pairs.
And 4, step 4: and (3) establishing an intelligent substitution model of the multiphase flow numerical model by coupling different artificial intelligence methods according to the input-output training sample set obtained in the step (3) so as to greatly improve the calculation efficiency of the cross-feed inversion iteration.
And 5: and (4) respectively establishing two independent nonlinear programming optimization models for inversion identification of pollution sources and pollutant migration parameters, and embedding the intelligent substitution model established in the step (4) into the optimization model as one of constraint conditions. When identifying the pollution source characteristics, the quantity expressing the pollution source index is regarded as a variable and the quantity expressing the pollutant migration parameter is regarded as a constant in both the substitution model and the optimization model. When identifying the pollutant migration parameter, the quantity expressing the pollutant migration parameter is regarded as a variable and the quantity expressing the pollutant source index is regarded as a constant in both the surrogate model and the optimization model.
Step 6: and (4) rewriting the pollution source characteristic inversion optimization model established in the step (5) based on homotopy theory to obtain a series of homotopy optimization models.
And 7: and (4) sequentially solving the homotopy optimization model series established in the step (6) by adopting a swarm intelligence optimization algorithm, wherein the solution of the last optimization model is the inversion identification result of the underground water pollution source.
And 8: and (4) feeding back and utilizing the identification result obtained in the step (7) into identification calculation of the pollutant migration parameters, and solving the aquifer parameter inversion optimization model established in the step (5) by applying a homotopy transformation-group intelligent optimization algorithm to obtain an inversion identification result of the pollutant migration parameters.
And step 9: and (4) connecting the two optimized identification processes of the step 6, the step 7 and the step 8 to form a feedback correction iteration loop, and feeding back the identification result obtained by the front iteration into the identification calculation of the next iteration by using the iteration loop. Along with the progress of the feedback correction iteration process, the identification results of the pollution source and the pollutant migration parameters are mutually promoted and are continuously and gradually improved.
Step 10: in the feedback correction iteration process of step 9, after each recognition is completed, the training sample set of step 3 is updated according to the recognition result, and the intelligent surrogate model is reestablished by using the updated training sample, so that the local precision of the intelligent surrogate model in the optimal solution neighborhood is improved.
And 2, constructing a multiphase flow migration model of the underground water organic pollution by using a UTCHEM program.
The pollution source characteristics to be identified in step 3 include: longitudinal coordinates of the pollution source, transverse coordinates of the pollution source, migration and conversion time of the pollutants and leakage amount of the pollutants; the contaminant migration parameters to be identified include: porosity, permeability, longitudinal aqueous phase dispersion, transverse aqueous phase dispersion. The random sampling is realized by a Latin hypercube sampling method.
The artificial intelligence substitution modeling method in the step 4 comprises the following steps: gaussian process method, kernel extreme learning machine method. The Gaussian process modeling method comprises the following steps:
the regression equation of Gaussian Process (GP) can be expressed as:
wherein f (x) = [ f ] 1 (x)，f 2 (x)，…，f k (x)] T Is a known basis function, β = (β) 1 ，β 2 ，…，β k ) T And obtaining the regression parameters corresponding to the basis functions by estimating the training samples. Z (x) is a local deviation term, the mean is 0, and the variance is
Its covariance can be expressed as:
r (u, v) is a correlation function of n-dimensional vectors u and v:
in the formula, alpha i For the parameter to be determined, u i And v i The ith element of u and v.
For a given sample P = [ P ] 1 ，p 2 ，…，p m ] T And corresponding output response Q = [ Q ] 1 ，q 2 ，…，q m ] T The predicted output response y (x) for any input vector x can be written as:
where r is the correlation vector between x and sample P:
r(x)＝[R(x，p 1 )，R(x，p 2 )，…，R(x，p m )] T (5)
f is the response column vector for sample P:
r is a correlation matrix among sample points of the sample P:
the Kernel Extreme Learning Machine (KELM) modeling method is as follows:
for training sample (X) j ，t j ) J =1, …, N, output y of Extreme Learning Machine (ELM) j Can be expressed as:
q(X j )＝[p(ω 1 X j +b 1 )，p(ω 2 X j +b 2 )，…，p(ω L X j +b L )] T (10)
wherein p (-) is an excitation function, and the input node passes through a weight vector omega i Connected to the ith hidden neuron, i =1, …, L, b i Threshold for the ith hidden neuron, p (ω) i X j +b i ) As an output function of the ith hidden neuron, beta i The weight vector for the ith implicit neuron connected to the output neuron.
Formula (9) may be represented as follows:
Qβ＝Y (11)
wherein β = [ β ] 1 ，…，β L ] T Y＝[y 1 ，…，y N ] T Q is the hidden layer output matrix of ELM:
if an extreme learning machine model with L hidden nodes is able to learn N training samples unbiased, then the following formula exists:
in the formula, t j Representing the target value.
Formula (13) may be abbreviated as:
Qβ＝T (14)
least square solution (14)
β＝Q + T (15)
In the formula, Q + Moore-Penrose generalized inverse of Q. Q + Calculated from the following equation
Q + ＝Q T (QQ T ) -1 (16)
For training sample (X) j ，t j ) J =1, …, N, KELM's original optimization problem is expressed as
In the formula, C represents a regularization parameter capable of balancing training errors and algorithm complexity, ξ j Representative error。
The optimization problem can be converted into a lagrange dual form solution:
in the formula, theta i Representing the lagrange operator. The dual problem can be solved by applying the Karush-Kuhn-Tucker (KKT) condition:
a least squares solution for the weight vector β can be calculated from equation (19):
constructing a wavelet implicit mapping function K (x) according to a kernel function theory j ，x j ) Instead of a random mapping function q (x) j )：
K ELM ＝QQ T (21)
K ELM(i，j) ＝q(x i ) T ·q(x j )＝K(x i ，x j ) (22)
The trained KELM output function is expressed as follows:
wherein T = [ T ] 1 ，…，t N ] T 。
The expression of the kernel function is:
in the formula alpha w ，γ 1，w ，γ 2，w Is an adjustable parameter.
Training a Gaussian process model and a nuclear extreme learning machine model by using training samples to enable the Gaussian process model and the nuclear extreme learning machine model to approximate to the input-output relation of a multiphase flow numerical model, establishing an optimization model as follows, and determining the weighting weight and the key parameter value of the Gaussian process model and the nuclear extreme learning machine model. The decision variables of the optimization model are the weighted weight values of the intelligent substitution model and the variable parameter values of each model, and the optimal solution of the optimization model enables the root mean square error between the substitution model output and the simulation model output obtained by applying the test sample to be minimum.
In the formula, y i (x k ，p i ) Using parameter vector p for ith substitution model i And the kth test sample input x k To input a corresponding output response, i =1,2,.., N, k =1,2,. For M, y actual (x k ) Inputting x for the k test sample k And outputting a response by the corresponding simulation model, wherein M is the number of the test samples. And solving the optimization model by using a particle swarm optimization algorithm, obtaining the optimal values of the weighting weight and the variable parameters of the substitution model, and constructing a Gaussian Process (GP) -Kernel Extreme Learning Machine (KELM) intelligent substitution model (GP-KELM).
The two nonlinear optimization models in the step 5 take the minimum square sum of absolute deviations of the simulation calculation concentration value and the actually measured concentration value of the monitoring well as an objective function; respectively taking the pollution source characteristics to be identified and the pollutant migration parameters as decision variables, wherein the decision variables comprise: longitudinal coordinates of the pollution source, transverse coordinates of the pollution source, migration and conversion time of the pollutants, leakage amount of the pollutants, porosity, permeability, longitudinal aqueous phase dispersity and transverse aqueous phase dispersity; and taking an intelligent substitution model as equality constraint of a pollutant migration and transformation rule, and taking the value ranges of pollution source characteristics and pollutant migration parameters as inequality constraint conditions. The expression is as follows:
pollution source identification
Migration parameter identification>
C k For the actual monitored value of contaminant concentration in the kth monitored well, k =1,2, ·, n,
calculated for the corresponding simulation. />
The pollutant migration parameter value vector is obtained by calculating an intelligent substitution model instead of a multiphase flow numerical model, wherein s is a pollution source related variable value vector, and p is a pollutant migration parameter value vector. The constraints respectively represent the longitudinal coordinates (X) of the pollution source i ) Transverse coordinate (Y) of the source of pollution i ) Duration of pollutant transfer (T) i ) Amount of leakage of contaminants (V) i ) Porosity (theta), permeability (K), longitudinal oil phase dispersion (D) oil，l ) (ii) degree of transverse oil phase dispersion (D) oil，t ) The upper and lower bounds of the variation range. In the contamination Source identification model, the contamination Source signature is considered as ^ er>
And the contaminant migration parameter is set to a constant; in the contaminant migration parameter identification model, the migration parameter is considered as ≥ er>
And the contamination source signature is set to a constant.
And preferentially identifying the characteristics of the pollution source, estimating and assigning the migration parameters of the pollutants to be identified according to the field investigation result and experience, and using the estimation and assignment as the known constant in the process of independently identifying the characteristics of the pollution source. The pollution source inversion homotopy optimization model series construction in the step (7) needs to introduce homotopy parameters lambda and construct homotopy functions H (s, lambda) by means of homotopy algorithm thought, so that when lambda =0, the solution of the equation set H (s, lambda) =0 is the assumed underground water pollution source characteristic, and when lambda =1, the solution of the equation set H (s, lambda) =0 is the true underground water pollution source characteristic to be solved. Starting from any assumed pollution source information, sequentially solving optimization problems corresponding to a series of homotopic equations through path tracking, gradually approaching and finally obtaining real pollution source information to be solved.
The homotopy function takes a linear homotopy as follows:
H(s，λ)＝λF(s)+(1-λ)G(s) (27)
wherein F(s) = F(s) -C obs ，G(s)＝f(s)-C 0 . In the formula, s is a value vector of a pollution source characteristic related variable; λ is homotopy parameter, and the value range [0,1 ]](ii) a f () represents a multiphase flow simulation model or a surrogate model thereof; c obs Actually monitoring concentration vectors for monitoring points; c 0 And the concentration vector is calculated by substituting the characteristics of the arbitrarily assumed underground water pollution source into the monitoring points obtained by the simulation model.
The homotopy function H continuously depends on homotopy parameter lambda, and the value range of the lambda is within the range of 0,1]Make a division of 0 ＝0＜λ 1 ＜…＜λ N =1, a series of equations is obtained:
H(s，λ i )＝f(s)-(λ i ·C obs +(1-λ i )·C 0 )＝0，i＝1，2，...，N (28)
if λ i+1 -λ i Sufficiently small, the solution(s) of the adjacent equations i+1 And(s) i Are very close.
And (4) rewriting the optimization model in the step (5) according to the homotopy algorithm idea to obtain an optimization model series corresponding to the homotopy equation set series of the underground water pollution source inversion identification problem, as shown in the formula (29).
And the calculated concentration of the pollutants in the k-th monitoring well obtained by substituting the characteristics of the arbitrarily assumed underground water pollution source into the simulation model is shown.
The group intelligent optimization algorithm adopted in the step (8) is a particle swarm optimization algorithm.
The basic steps of the swarm intelligence optimization inversion can be summarized as:
(1) And setting an iteration counter t =0, and randomly initializing the position and the speed of the particle swarm. Suppose the velocity and position of the ith particle are X i ＝(x i，1 ，x i，2 ，...，x i，d ) And V i ＝(v i，1 ，v i，2 ，...，v i，d ) And d is the number of kernel parameters to be optimized, each position corresponds to a group of kernel parameters, and the positions are the feature vectors of the particles.
(2)
(2) The fitness of each particle is calculated by inputting the objective function in the formula (29) to the position vector, and each position vector is stored as P i ＝(p i，1 ，p i，2 ，...，p i，d ) Saving the optimal position vector as P g ＝(p g，1 ，p g，2 ，...，p g，d )。
(3) Calculating adaptive inertial weights to balance global search capabilities and local search capabilities:
in the formula w max And w min As maximum and minimum values of the weights, f i For the current i-th particle, the value of the objective function, f avg And f min The average value and the minimum value of all current objective function values.
(4) Updating the speed and position of the particle swarm:
in the formula, c1 and c2 are learning factors, and r1 and r2 are random numbers uniformly distributed between 0,1.
(5) The fitness of the particle population is recalculated and compared to previous values, and pbest, gbest is then updated, while t = t + 1. When t reaches the maximum iteration times, making gbest as the optimal kernel parameter; otherwise, returning to the step (3).
And (3) solving the homotopy optimization model series constructed in the step (6) in sequence by using a Particle Swarm Optimization (PSO), wherein in the solving process, because the path tracking process is gradually evolved, the solutions of two adjacent optimization problems are also very close to each other:
in the formula (I), the compound is shown in the specification,
for the initial population of the ith homotopy optimization model, s i，opt The optimal solution of the ith homotopy optimization model obtained by applying the PSO can be used as a basis for generating the initial population of the next optimization model:
in the formula, v s，low ，v s，up And representing the upper limit and the lower limit of the adjustment displacement vector corresponding to the pollution source characteristics in the PSO algorithm.
The solution of the ith optimization problem is used as a generation basis for solving the initial value of the (i + 1) th optimization problem, so that the optimization process is ensured to be gradually evolved, and the premature convergence problem caused by the fact that the initial value is far from the optimal solution is avoided. And the solution of the last optimization model is the inversion identification result of the underground water pollution source.
In step 9, solving the pollutant migration parameter inversion optimization model constructed in step 5 by using a homotopy transformation-group intelligent optimization algorithm, setting the pollution source characteristics as the optimal value optimally solved in step 7 as a known constant in the optimization solving process, and performing independent identification on the pollutant migration parameters:
in the formula (I), the compound is shown in the specification,
updating is carried out according to the characteristic value of the identified pollution source.
And starting a new round of pollution source characteristic inversion identification according to the obtained pollutant migration parameter identification value.
And 9, the iterative circulation system for the mutual feed correction of the pollution source characteristics and the pollutant migration parameters connects the pollution source characteristic inversion identification and the pollutant migration parameter inversion identification in the steps 6, 7 and 8, and repeatedly and alternately carries out until the optimal solution of the optimization model meets the convergence condition. With the progress of the feedback correction iteration process, the identification result obtained by the front iteration is fed back and utilized in the identification calculation of the next iteration, so that the identification results of the pollution source and the migration parameter are mutually promoted and are continuously and gradually improved.
The training sample updating in the step 10 adopts an improved self-adaptive updating sampling method: on the basis of solving the optimization model, adding the current optimal solution into the original training sample, simultaneously removing two samples with the largest distance from the current optimal solution, and adding the average value of the current optimal solution and the samples with the largest distance from the current optimal solution into the training sample as another newly added sample. The distance between the sample point and the current optimal solution is expressed in terms of euclidean distance.
Two n-dimensional vectors a (x) 11 ，x 12 ，…，x 1n ) And b (x) 21 ，x 22 ，…，x 2n ) The Euclidean distance between them is represented by the following formula:
and (4) reestablishing the substitution model and the optimization model by using the updated training sample, and solving the optimization model to obtain a new optimal solution. This process is repeated until the optimal solution of the optimization model satisfies the convergence condition.
In the underground water pollution source and pollutant migration parameter mutual feedback circulating system, by the self-adaptive updating sampling mode, the inversion identification result of each iteration underground water pollution source and parameter is used as feedback information to improve the training sample of the substitution model, so that the training sample effectively covers the neighborhood of the optimal solution; in addition, the improved self-adaptive updating sampling method not only simply adds the optimal solution as a new sample into the training sample, but also adds effective samples in a certain mode according to the optimal solution and simultaneously eliminates ineffective samples, so that the structure of the training sample is more reasonable, and the training samples are prevented from being too concentrated or too dispersed. By establishing the self-adaptive sample updating process, the training sample, the substitution model and the optimization model are gradually improved, and the precision of the identification result of the pollution source and the pollutant migration parameter is improved.
Example two
As shown in fig. 1-5, this embodiment is a hypothetical organically-polluted diving aquifer that can be generalized to a homogeneous, isotropic three-dimensional multiphase flow model. There is no natural boundary near the contaminated site, and the boundary is defined at a position where the migration effect of the contaminants is negligible. Wherein, the northeast boundary and the southwest boundary are generalized into a type of boundary; the southeast boundary and the northwest boundary are composed of flow surfaces and generalized to zero-flux boundaries; the lower part of the calculation simulation area is a water-resisting layer and can be generalized to a zero-flux boundary, the upper part of the calculation simulation area is a diving surface and a water exchange boundary, and the thickness of the aquifer is slowly changed along the direction of underground water flow, so the calculation simulation area is generalized to an equal-thickness aquifer. The physicochemical parameters of water and organic contaminants chlorobenzene are detailed in table 1.
TABLE 1
Organic pollution in the field is caused by point-like pollution sources, the pollutants enter an aquifer in a short time, and most of the following time is a natural dissolution and diffusion stage of the pollutants in the aquifer. Therefore, the variables to be identified in the simulation model (the variables to be identified in the pollution source inversion identification problem) can be finally determined as: 1. lateral, longitudinal coordinates of the contamination source (contamination source location). The source of contamination is by default located at the top of the aquifer, the vertical coordinates need not be identified. 2. The pollutant migration conversion is long (simulation period). 3. And (4) leakage amount of pollutants. (the migration and conversion time of the pollutants and the leakage amount of the pollutants are the pollution source release history in the pollution source inversion identification problem). 4. Contaminant migration parameters (including porosity, permeability, longitudinal aqueous phase dispersion, transverse aqueous phase dispersion). And 5 water level and water quality monitoring wells in the field are used for acquiring pollutant concentration monitoring data as known information for inversion solution. The site generalization and monitoring well location are shown in figure 2.
Actual values of the contamination source information and the contamination migration parameters given in the hypothetical example are given in table 2.
TABLE 2
Based on the information, the UTCHEM is used for model construction, a model of the groundwater DNAPS polluted multiphase flow related to the hypothetical example is built, the built simulation model is used for forward forecasting calculation, the space-time distribution condition of the pollutant concentration in the seepage field is obtained, namely the pollutant concentration at the monitoring well is used as actual monitoring data, and the pollutant concentration at the bottom of the aquifer at the five monitoring wells at the last moment is calculated and output (actual monitoring data) by the hypothetical example forward model shown in the table 3.
TABLE 3
By using the method, the characteristics of each pollution source and the migration parameters of the pollutants are subjected to inversion calculation according to the concentration monitoring data in the table 3;
in this embodiment:
in the step 3, the prior intervals of all parameters are shown in a parameter prior distribution characteristic in a table 4;
TABLE 4
The number of training samples in the step 3 is 100, and 20 test samples are adopted;
and 7, when the homotopy transformation-group intelligent optimization algorithm is applied to solve the optimization model in the step 7, taking 0.1 interval of homotopy parameters, constructing an equation series containing 10 homotopy equations, and rewriting the optimization model according to the homotopy equation series to obtain corresponding 10 optimization models.
When solving the homotopy optimization model series in the step 8, the maximum evolutionary algebra solved by the first nine optimization models is 60, and the maximum evolutionary algebra solved by the last optimization model is 120.
The end conditions of the feedback correction iteration loop set in the step 10 are as follows: and the Euclidean distance of the optimal solution obtained by the two adjacent optimization solutions after normalization is less than 0.02. The identification results of the cyclic correction process are shown in table 5, table 6, table 7, and table 5, the cyclic correction results of the contamination source characteristics and the contamination transfer parameters (1), table 6, and table 7. And the result after the twelfth correction is the final identification result of the underground water pollution source and the pollutant migration parameter.
TABLE 5
TABLE 6
TABLE 7
The results of the precision analysis in the cross-feed cyclic correction process are shown in table 8, cyclic correction result precision analysis (1), table 9, cyclic correction result precision analysis (2), and table 10, cyclic correction result precision analysis (3). In the loop correction process, the average relative error of the recognition result slightly increases after the first correction, but gradually decreases as the correction process progresses. Fig. 5 can more intuitively show the process of gradually approaching the identification values of the pollution source characteristics and the pollutant migration parameters to the actual values in the correction cycle. After the cycle was terminated, the average relative error of the final recognition was only 1.02%, which is significantly lower than the average relative error of the initial recognition by 4.21%. In addition, the maximum relative error in the initial recognition result was 8.51%, and after the cycle correction, the maximum relative error was only 2.25%. The result shows that the identification precision can be obviously improved by establishing a mutual feed inversion system of the pollution source characteristics and the pollutant migration parameters in the underground water pollution inversion tracing process to iteratively correct the identification result.
TABLE 8
TABLE 9
The recognition results obtained by the cross-feed inversion system were compared with the results obtained by directly applying the homotopy transform-group intelligent optimization algorithm, as shown in the feedback correction iterative loop performance analysis of table 11. Along with the progress of the feedback correction iteration loop, the pollution source identification and the pollutant migration parameter identification are mutually promoted step by step, the average relative error of the pollution source identification result is reduced to 1.13% from 2.46%, and the average relative error of the pollutant migration parameter identification result is reduced to 0.93% from 2.07%. And the maximum relative error of the identification result of the cross-feed inversion system is 2.25%, which is obviously lower than 4.26% of the identification result at the same time. These show that the cross-feed inversion system applying independent identification of pollution source characteristics and pollutant migration parameters can effectively inhibit the effect of 'different parameters and same effects' and improve the identification accuracy.
TABLE 11
And establishing a multiphase flow simulation model according to the corrected optimal solution, namely the corrected recognition result of the pollution source and the pollutant migration parameter, calculating the distribution condition of the pollutants in the aquifer at the last moment, and comparing the distribution condition with the actual distribution of the pollutants at the last moment of the hypothetical example, as shown in figure 5. After the cross-feed cyclic correction, the range of each concentration contour line of the identified pollution plume is very close to that of the actual pollution plume, and the shapes of the two pollution plumes are basically consistent. And establishing a simulation model by applying the corrected recognition result, and calculating the current or forecasting the distribution condition of the pollutants in the future aquifer more accurately, thereby providing a reliable basis for designing a groundwater pollution remediation scheme and evaluating risks.
The polluted aquifer multiphase flow simulation model in the example is operated on a computer with a CPU of Intel core i 5.0 GHz and an internal memory of 8GB, and the operation time is about 600 seconds on average; and the artificial intelligence substitution model only needs 1.2 seconds for running once, and if the intelligent model is used for substituting the simulation model, the calculation efficiency of the optimization solving process can be improved by about 500 times.
Therefore, the method can realize the inversion identification of the characteristics of the underground water pollution source and the parameter correction of the multiphase flow numerical model with small calculation load, and greatly improve the accuracy of the identification result.
The above description is only for the preferred embodiment of the present application, but the scope of the present application is not limited thereto, and any changes or substitutions that can be easily conceived by those skilled in the art within the technical scope of the present application should be covered within the scope of the present application. Therefore, the protection scope of the present application shall be subject to the protection scope of the claims.
Claims (8)
1. An underground water organic pollution mutual-feed inversion tracing method based on artificial intelligence is characterized by comprising the following steps:
s1, acquiring monitoring data, and acquiring a numerical simulation model of underground water organic pollution multiphase flow based on the monitoring data;
s2, acquiring a variable to be recognized, and acquiring a training sample set and a test sample set based on the value range of the variable to be recognized;
s3, based on the training sample set, obtaining an intelligent substitution model of the multi-phase flow numerical model by coupling an artificial intelligence method;
s4, constructing a nonlinear programming optimization model, and obtaining an inversion optimization model based on an intelligent substitution model of the multiphase flow numerical model and the nonlinear programming optimization model;
s5, based on the inversion optimization model, performing inversion identification on the training sample set through a homotopy transformation-group intelligent optimization algorithm to obtain an inversion identification result, and updating the training sample set based on the inversion identification result;
s6, returning to S4 until a final training sample set is obtained;
and S7, reestablishing an intelligent substitution model of the multiphase flow numerical model based on the final training sample set to obtain an updated inversion identification result.
2. The underground water organic pollution mutual-feed inversion tracing method based on artificial intelligence as claimed in claim 1,
the variables to be identified comprise pollution source characteristics and pollutant migration parameters;
the pollution source characteristics include: longitudinal coordinates of the pollution source, transverse coordinates of the pollution source, migration and conversion time length of the pollutants and leakage amount of the pollutants;
the contaminant migration parameters include: porosity, permeability, longitudinal aqueous phase dispersion and transverse aqueous phase dispersion.
3. The underground water organic pollution mutual feedback inversion tracing method based on artificial intelligence as claimed in claim 2, wherein the method for obtaining the training sample set and the inspection sample set comprises:
collecting a plurality of groups of samples based on the variable to be identified;
inputting a plurality of groups of samples into the underground water organic pollution multiphase flow numerical simulation model to obtain a plurality of model responses;
based on the plurality of model responses, a training sample set and a test sample set are obtained.
4. The underground water organic pollution mutual-feed inversion tracing method based on artificial intelligence as claimed in claim 3, wherein the method for obtaining the intelligent substitution model of the multiphase flow numerical model based on the training sample set by coupling the artificial intelligence method comprises:
constructing a Gaussian process model and a kernel limit learning machine model by the artificial intelligence method based on the training sample set;
acquiring the weighting weight and key parameters of the Gaussian process model and the kernel-limit learning machine model, and optimizing the Gaussian process model and the kernel-limit learning machine model by adopting a particle swarm optimization algorithm to obtain the optimal values of the weighting weight and the key parameters;
and obtaining an intelligent substitute model of the multiphase flow numerical model based on the weighting weight and the optimal value of the key parameter.
5. The underground water organic pollution mutual-feed inversion tracing method based on artificial intelligence as claimed in claim 2, wherein the inversion optimization model comprises a pollution source characteristic inversion optimization model and an aquifer parameter inversion optimization model.
6. The underground water organic pollution mutual feedback inversion tracing method based on artificial intelligence of claim 5, wherein obtaining the inversion identification result comprises:
rewriting the pollution source characteristic inversion optimization model based on homotopy theory to obtain a plurality of homotopy optimization models;
solving the homotopy optimization models based on a swarm intelligence optimization algorithm to obtain an inversion identification result of the pollution source characteristics;
and acquiring an inversion identification result of the pollutant migration parameters according to a homotopy transformation-group intelligent optimization algorithm based on the inversion identification result of the pollution source characteristics and the aquifer parameter inversion optimization model.
7. The artificial intelligence-based groundwater organic pollution mutual feedback inversion tracing method of claim 6, wherein updating the training sample set based on the inversion recognition result comprises:
and returning the inversion identification result of the pollution source characteristics and the inversion identification result of the pollutant migration parameters to the training sample set, and updating the training sample set.
8. The artificial intelligence based groundwater organic pollution mutual feedback inversion tracing method according to claim 7, wherein the method for obtaining the final training sample set comprises: and obtaining the final training sample set by adopting an improved self-adaptive updating sampling method based on the inversion identification result.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211425201.7A CN115906626A (en) | 2022-11-14 | 2022-11-14 | Underground water organic pollution mutual-feedback inversion tracing method based on artificial intelligence |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211425201.7A CN115906626A (en) | 2022-11-14 | 2022-11-14 | Underground water organic pollution mutual-feedback inversion tracing method based on artificial intelligence |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN115906626A true CN115906626A (en) | 2023-04-04 |
Family
ID=86485133
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211425201.7A Pending CN115906626A (en) | 2022-11-14 | 2022-11-14 | Underground water organic pollution mutual-feedback inversion tracing method based on artificial intelligence |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115906626A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116151488A (en) * | 2023-04-19 | 2023-05-23 | 中科三清科技有限公司 | Pollution data analysis method, system and equipment |
| CN116739386A (en) * | 2023-08-10 | 2023-09-12 | 哈尔滨工业大学(深圳)(哈尔滨工业大学深圳科技创新研究院) | Multi-index fusion pollution tracing method, equipment and readable storage medium |
-
2022
- 2022-11-14 CN CN202211425201.7A patent/CN115906626A/en active Pending
Non-Patent Citations (2)
| Title |
|---|
| 侯泽宇: "地下水 DNAPLs 污染源反演识别及不确定性分析研究", 中国博士学位论文全文数据库 工程科技Ⅰ辑》, pages 027 - 73 * |
| 潘红光，等: "采用 FFT－PSO 策略的多约束稀疏阵天线方向图综合", 《电讯技术》, pages 1135 - 117 * |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116151488A (en) * | 2023-04-19 | 2023-05-23 | 中科三清科技有限公司 | Pollution data analysis method, system and equipment |
| CN116739386A (en) * | 2023-08-10 | 2023-09-12 | 哈尔滨工业大学(深圳)(哈尔滨工业大学深圳科技创新研究院) | Multi-index fusion pollution tracing method, equipment and readable storage medium |
| CN116739386B (en) * | 2023-08-10 | 2024-03-08 | 哈尔滨工业大学(深圳)(哈尔滨工业大学深圳科技创新研究院) | Multi-index fusion pollution tracing method, equipment and readable storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN115906626A (en) | Underground water organic pollution mutual-feedback inversion tracing method based on artificial intelligence | |
| CN104318325B (en) | Many basin real-time intelligent water quality prediction methods and system | |
| CN106529818B (en) | Water quality assessment Forecasting Methodology based on Fuzzy Wavelet Network | |
| Kişi | Evolutionary fuzzy models for river suspended sediment concentration estimation | |
| CN114707395A (en) | Groundwater organic pollution source inversion method based on heuristic homotopy algorithm | |
| CN107274030A (en) | Runoff Forecast method and system based on hydrology variable year border and monthly variation characteristic | |
| CN114548591B (en) | Sequential data prediction method and system based on mixed deep learning model and Stacking | |
| KR101474874B1 (en) | computing system for well placement optimization developed by SA/ANN and well placement optimization method using Thereof | |
| CN114492164A (en) | Organic pollutant migration numerical model substitution method based on multi-core extreme learning machine | |
| CN114399073A (en) | Ocean surface temperature field prediction method based on deep learning | |
| CN111311577A (en) | Intelligent water seepage detection method based on generation of confrontation network and reinforcement learning | |
| Xiaoyi et al. | Research on hybrid mechanism modeling of algal bloom formation in urban lakes and reservoirs | |
| CN110119534B (en) | Method for determining monitoring range of underground water pollution monitoring well | |
| CN115310348A (en) | Stacking-based grouting amount integrated agent prediction model and prediction method | |
| Han et al. | Predict water quality using an improved deep learning method based on spatiotemporal feature correlated: a case study of the Tanghe Reservoir in China | |
| Jiang et al. | Deep-reinforcement-learning-based water diversion strategy | |
| Gupta et al. | Evaluation of water quality based on a machine learning algorithm and water quality index for mid gangetic region (south Bihar plain), India | |
| CN116502539B (en) | VOCs gas concentration prediction method and system | |
| Damian | A critical review on artificial intelligence models in hydrological forecasting how reliable are artificial intelligence models | |
| CN117034808A (en) | Natural gas pipe network pressure estimation method based on graph attention network | |
| CN115510977A (en) | Geological statistical pattern recognition method based on Bayes ensemble learning machine | |
| CN116029331A (en) | Big data crowd sensing collaborative prediction method for safety state of coal mine working face | |
| CN115293024A (en) | Site soil pollutant concentration three-dimensional spatial interpolation method integrating multi-source cooperative variables | |
| Todorovski et al. | Integrating domain knowledge in equation discovery | |
| Alagha et al. | Artificial intelligence based modelling of hydrological processes |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20230404 |