Simulation system and method for predicting new pollutant water-sediment distribution behavior

Abstract

The invention discloses a simulation system and method for predicting a new pollutant water-sediment distribution behavior, and relates to the technical field of water environment research, and the system comprises a reactor, a partition plate, a temperature control device, a hydrological index monitoring unit, a water flow circulation unit and a water inlet unit. According to the system, the distribution behavior of new pollutants in a natural water environment can be simulated, then the prediction model can be constructed in combination with the experimental method, and then the distribution behavior of the new pollutants in a target water body is predicted by utilizing the prediction model. The problem that the distribution behavior of the new pollutants between the water body and the sediment is difficult to know in the prior art is solved.

Description

Technical Field

[0001] The present application relates to the technical field of water environment research, and particularly relates to a simulation system and method for predicting the water-sediment partitioning behavior of emerging pollutants. Background Art

[0002] Emerging pollutants are different from conventional pollutants, and their environmental persistence and biological toxicity are significantly greater than those of other common pollutants. Generally speaking, emerging pollutants include persistent organic pollutants, endocrine disruptors, antibiotics, microplastics, etc. The subdivision types of these pollutants are very complicated, and there are a large number of compounds involved. Due to their obvious bioaccumulation and biological toxicity, emerging pollutants have a significant impact on the health of humans and other organisms. As an important storage carrier of emerging pollutants in the environment, the forms of emerging pollutants in water are diverse, mainly including dissolved, suspended and sedimentary forms. Current research has found that different concentrations of emerging pollutants have been detected in the water and sediments of different basins. There is a certain relationship between the forms of emerging pollutants in the water environment and their biological toxicity and persistence. Therefore, it is necessary to clarify the partitioning of emerging pollutants among different forms in the water environment to accurately assess their environmental risks.

[0003] Currently, related research still mainly focuses on the degradation technology and environmental distribution of emerging pollutants, and it is difficult to understand the partitioning behavior of emerging pollutants between water and sediments. Summary of the Invention

[0004] The main purpose of the embodiments of the present application is to propose a simulation system and method for predicting the water-sediment partitioning behavior of emerging pollutants to predict the partitioning behavior of emerging pollutants at the water-sediment interface.

[0005] To achieve the above object, on the one hand, the embodiments of the present application propose a simulation system for predicting the water-sediment partitioning behavior of emerging pollutants, and the simulation system includes: a reactor, a partition plate, a temperature control device, a hydrological index monitoring unit, a water flow circulation unit and a water inlet unit;

[0006] Sediments are provided at the bottom inside the reactor, and circulating water is provided above the sediments; the circulating water is set as a single solution of the target emerging pollutant;

[0007] The partition plate is arranged in the internal space of the reactor, and the partition plate is connected to the bottom of the reactor, and the partition plate is used to separate the internal space of the reactor;

[0008] The temperature control device is used to control the circulating water at a set temperature;

[0009] The hydrological index monitoring unit is used to measure the characteristics of hydrological indexes;

[0010] The water flow circulation unit is used to drive the water flow to circulate to form the circulating water;

[0011] The water inlet unit is used to externally connect a water inlet pump and a storage tank.

[0012] In some embodiments, the hydrological index monitoring unit includes a multi-purpose monitor and an ultrasonic flowmeter;

[0013] The multi-purpose monitor is used to measure temperature, pH, and redox potential.

[0014] In some embodiments, the water flow circulation unit includes at least one impeller and at least one motor;

[0015] Wherein, the impeller is installed between the wall of the reactor and the partition plate;

[0016] The motor is used to drive the impeller to drive the water flow to circulate.

[0017] In some embodiments, the simulation system further includes a water sample sampling port and a sediment sampling spoon;

[0018] Wherein, the water sample sampling port is arranged on the wall of the reactor.

[0019] To achieve the above object, on the other hand, an experimental method for predicting the water-sediment partitioning behavior of new pollutants is proposed in an embodiment of the present application. The experimental method is applied to the simulation system for predicting the water-sediment partitioning behavior of new pollutants as described in the present application. The experimental method includes the following steps:

[0020] Set the types of water body and sediment in the reactor, and set the physical environment parameters in the reactor;

[0021] Under the condition that the physical environment parameters reach the set conditions, measure the concentrations of new pollutants in the water body and the sediment, and then calculate the partitioning coefficient of the new pollutants;

[0022] Taking the water body characteristics and sediment characteristics of the reactor during the reaction process as characteristic variables, and taking the partitioning coefficient of the new pollutants as the target variable, construct a prediction model for the water-sediment partitioning coefficient;

[0023] According to the partial dependence plot, analyze the influence of different characteristic variables on the partitioning coefficient of the new pollutants;

[0024] Use the prediction model to predict the partitioning behavior of new pollutants in the target water body.

[0025] In some embodiments, when the physical environment parameters reach the set conditions, measuring the concentrations of new pollutants in the water body and the sediment and then calculating the distribution coefficient of the new pollutants includes the following steps:

[0026] When the physical environment parameters reach the set conditions, measuring the concentrations of each new pollutant in the water body and the sediment respectively and then calculating the distribution coefficient of the new pollutant;

[0027] The calculation expression of the distribution coefficient is:

[0028]

[0029] where, K d represents the distribution coefficient, C s represents the concentration of each new pollutant in the sediment, and C w represents the concentration of each new pollutant in the water body.

[0030] In some embodiments, using the water body characteristics and sediment characteristics of the reactor during the reaction process as characteristic variables and the distribution coefficient of the new pollutant as the target variable to construct a prediction model for the water body-sediment distribution coefficient includes the following steps:

[0031] Determining the characteristic variables marked with the target variable as training data;

[0032] Determining the hyperparameters of the deep learning model and the spatial reference range of the hyperparameters;

[0033] Constructing a deep learning model as the prediction model according to the training data and the hyperparameters.

[0034] In some embodiments, determining the hyperparameters of the deep learning model includes the following steps:

[0035] Determining that the hyperparameters of the deep learning model include the number of hidden layers, the number of units in the hidden layer, the learning rate, the optimizer, regularization, and the activation function.

[0036] In some embodiments, constructing a deep learning model as the prediction model according to the training data and the hyperparameters includes the following steps:

[0037] Dividing the training data into five equal parts; one part of the training data is used as the validation set, and the remaining four parts of the training data are used as the training set;

[0038] Using the obtained training set and validation set to construct the deep learning model based on the five-fold cross-validation method;

[0039] Verify the deep learning model using the verification set to obtain the corresponding empirical error;

[0040] Select the hyperparameters corresponding to the minimum empirical error as the target hyperparameters of the deep learning model;

[0041] Calculate the correlation coefficient of the deep learning model set with the target hyperparameters;

[0042] If the correlation coefficient meets the preset conditions, determine the deep learning model set with the target hyperparameters as the prediction model.

[0043] In some embodiments, the analyzing the influence of different feature variables on the distribution coefficient of the new pollutant according to the partial dependence graph includes the following steps:

[0044] Analyze the influence of different feature variables on the distribution coefficient of the new pollutant according to the partial dependence function of the partial dependence graph;

[0045] The expression of the partial dependence function is:

[0046]

[0047] where f S (x S ) represents the partial dependence function, and x S is the variable feature to be plotted in the partial dependence graph; K C is the other variable features in the prediction model except x S .

[0048] The embodiments of the present application at least include the following beneficial effects:

[0049] Through the system of the present application, the distribution behavior of new pollutants in the natural water environment can be simulated. Furthermore, a prediction model can be constructed by combining the experimental method of the present application, and then the distribution behavior of new pollutants in the target water body can be predicted using the prediction model, solving the problem that it is difficult to understand the distribution behavior of new pollutants between water and sediment in the prior art. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the following will briefly introduce the drawings required for the description of the embodiments. Obviously, the following drawings are only some embodiments of the present application. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0051] Figure 1Schematic flow chart of an experimental method for predicting the water-sediment partitioning behavior of emerging pollutants provided by an embodiment of the present application;

[0052] Figure 2 Example structural diagram of a simulation system for predicting the water-sediment partitioning behavior of emerging pollutants provided by an embodiment of the present application;

[0053] Figure 3 Example flow chart of an experimental method for predicting the water-sediment partitioning behavior of emerging pollutants provided by an embodiment of the present application.

[0054] Reference numerals: 1. Temperature control device; 2. Ultrasonic flowmeter; 3. Temperature, pH, ORP multi-purpose monitor; 4. Motor; 5. Cover plate; 6. Pusher; 6. Water pump; 8. Water storage tank; 9. Inlet; 10. Water sample sampling port; 11. Partition board; 12. Sediment sampling spoon. Detailed implementation manners

[0055] In order to make the objectives, technical solutions and advantages of the present application clearer, the present application will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application. When the following description refers to the accompanying drawings, unless otherwise indicated, the same numerals in different drawings represent the same or similar elements. The implementation manners described in the following exemplary embodiments do not represent all implementation manners consistent with the embodiments of the present application. They are only examples of devices and methods consistent with some aspects of the embodiments of the present application detailed in the appended claims.

[0056] It can be understood that the terms "first", "second", etc. used in the present application may be used herein to describe various concepts, but unless otherwise specified, these concepts are not limited by these terms. These terms are only used to distinguish one concept from another. For example, without departing from the scope of the embodiments of the present application, the first information may also be referred to as the second information, and similarly, the second information may also be referred to as the first information. Depending on the context, the words "if", "when" as used herein may be interpreted as "when...", "when...", or "in response to determining".

[0057] The terms "at least one", "a plurality", "each", "any one", etc. used in the present application, at least one includes one, two or more than two, a plurality includes two or more than two, each refers to each of the corresponding plurality, and any one refers to any one of the plurality.

[0058] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the technical field to which this application belongs. The terms used herein are for the purpose of describing embodiments of this application only and are not intended to limit this application.

[0059] Specifically, the embodiments of this application provide a simulation system for predicting the water-sediment partitioning behavior of new pollutants. The simulation system includes: a reactor, a partition plate, a temperature control device, a hydrological index monitoring unit, a water flow circulation unit, and a water inlet unit;

[0060] Sediments are provided at the bottom inside the reactor, and circulating water is provided above the sediments; the circulating water is set as a single solution of the target new pollutant;

[0061] The partition plate is disposed in the internal space of the reactor. The partition plate is connected to the bottom of the reactor and is used to separate the internal space of the reactor;

[0062] The temperature control device is used to control the circulating water at a set temperature;

[0063] The hydrological index monitoring unit is used to measure hydrological index characteristics;

[0064] The water flow circulation unit is used to push the water flow to circulate to form the circulating water;

[0065] The water inlet unit is used to externally connect a water inlet pump and a water storage tank.

[0066] As an optional implementation manner, the hydrological index monitoring unit includes a multi-purpose monitor and an ultrasonic flowmeter;

[0067] The multi-purpose monitor is used to measure temperature, pH, and redox potential.

[0068] As an optional implementation manner, the water flow circulation unit includes at least one impeller and at least one motor;

[0069] Wherein, the impeller is installed between the inner wall of the reactor and the partition plate;

[0070] The motor is used to drive the impeller to push the water flow to circulate.

[0071] As an optional implementation manner, the simulation system further includes a water sample sampling port and a sediment sampling spoon;

[0072] Wherein, the water sample sampling port is disposed on the inner wall of the reactor.

[0073] Refer to Figure 1, on the other hand, an experimental method for predicting the water-sediment partitioning behavior of emerging pollutants is proposed in an embodiment of the present application. The experimental method is applied to a simulation system for predicting the water-sediment partitioning behavior of emerging pollutants as described in the present application. The experimental method includes the following steps:

[0074] S100: Set the types of water body and sediment in the reactor, and set the physical environment parameters in the reactor;

[0075] S110: Under the condition that the physical environment parameters reach the set conditions, measure the concentrations of emerging pollutants in the water body and the sediment, and then calculate the partitioning coefficient of the emerging pollutants;

[0076] S120: Take the water body characteristics and sediment characteristics during the reaction process of the reactor as characteristic variables, and take the partitioning coefficient of the emerging pollutants as the target variable to construct a prediction model for the water-sediment partitioning coefficient;

[0077] S130: Analyze the influence of different characteristic variables on the partitioning coefficient of the emerging pollutants according to the partial dependence plot;

[0078] S140: Use the prediction model to predict the partitioning behavior of emerging pollutants in the target water body.

[0079] Further, S110 includes the following steps:

[0080] Under the condition that the physical environment parameters reach the set conditions, measure the concentrations of each emerging pollutant in the water body and the sediment respectively, and then calculate the partitioning coefficient of the emerging pollutants;

[0081] The calculation expression of the partitioning coefficient is:

[0082]

[0083] Among them, K d represents the partitioning coefficient, C s represents the concentration of each emerging pollutant in the sediment, C w represents the concentration of each emerging pollutant in the water body.

[0084] Further, S120 includes the following steps S121 to S123:

[0085] S121: Determine the characteristic variables marked with the target variable as training data;

[0086] S122: Determine the hyperparameters of the deep learning model and the spatial reference range of the hyperparameters;

[0087] S123: Construct a deep learning model based on the training data and the hyperparameters as the prediction model.

[0088] More specifically, the step of determining the hyperparameters of the deep learning model in S122 includes the following steps:

[0089] Determining the hyperparameters of the deep learning model includes the number of hidden layers, the number of units in the hidden layer, the learning rate, the optimizer, regularization, and the activation function.

[0090] More specifically, S123 includes the following steps:

[0091] Divide the training data into five equal parts; one part of the training data is used as the validation set, and the remaining four parts of the training data are used as the training set;

[0092] Construct the deep learning model based on the training set and the validation set obtained after division using the five-fold cross-validation method;

[0093] Validate the deep learning model using the validation set to obtain the corresponding empirical error;

[0094] Select the hyperparameters corresponding to the minimum empirical error as the target hyperparameters of the deep learning model;

[0095] Calculate the correlation coefficient of the deep learning model set with the target hyperparameters;

[0096] If the correlation coefficient meets the preset conditions, determine the deep learning model set with the target hyperparameters as the prediction model.

[0097] Further, S130 includes the following steps:

[0098] Analyze the influence of different feature variables on the partition coefficient of the new pollutant according to the partial dependence function of the partial dependence graph;

[0099] The expression of the partial dependence function is:

[0100]

[0101] where f S (x S ) represents the partial dependence function, x S is the variable feature to be plotted in the partial dependence graph; X C is the other variable features in the prediction model except x S .

[0102] Next, the solution of the embodiment of the present application will be introduced and described in detail with specific application examples.

[0103] The distribution of emerging pollutants between the dissolved and sedimentary states is comprehensively affected by environmental factors and the properties of pollutants, and there may also be interactions between different factors. In this embodiment, a combination of deep learning and simulation experiments is used to analyze the influence of various factors on the interfacial distribution behavior of emerging pollutants and to have a certain predictive effect on the distribution of emerging pollutants in actual water bodies.

[0104] Specifically, this embodiment provides a simulation system for predicting the water-sediment distribution behavior of emerging pollutants based on deep learning. Referring to Figure 2 , the system may include: a reactor, a partition, a temperature control device, a hydrological index monitoring unit, a water flow circulation unit, a water inlet unit, a water sample sampling port, and a sediment sampling spoon.

[0105] This embodiment also provides an experimental method for predicting the water-sediment distribution behavior of emerging pollutants based on deep learning. Referring to Figure 3 , the experimental method includes the following steps:

[0106] Set the types of water bodies and sediments in the device, and set parameters such as temperature and flow rate. Measure the concentrations of emerging pollutants in water samples and sediments under stable conditions and calculate their distribution coefficients;

[0107] Use the measurable water body characteristics such as the temperature, flow rate, pH, oxidation-reduction potential (ORP) of the reactor and the measurable sediment characteristics such as particle size and total organic carbon (TOC) during the experiment as feature variables;

[0108] Build a prediction model for the water-sediment distribution coefficient based on a deep learning algorithm;

[0109] Combine the Partial Dependence Plot to analyze the influence of different environmental variables on the water-sediment distribution coefficient of emerging pollutants and predict the distribution behavior of emerging pollutants in the natural environment. Among them, the Partial Dependence Plot (abbreviated as PDP) is a visualization tool for showing the influence of one or two features on the prediction results of a machine learning model. The PDP reveals the relationship between features and prediction results by marginalizing other features, only changing the values of one or two features, and observing the changes in the model prediction results.

[0110] Furthermore, the reactor is a cuboid, including sediments deposited on the substrate and circulating water above the sediments;

[0111] The partition is located in the middle of the reactor and is connected to the bottom of the reactor for separating the two long sides of the reactor;

[0112] The temperature control device can set the circulating water in the reactor at a certain temperature condition;

[0113] The hydrological index monitoring unit includes a multi - purpose monitor for temperature, pH, and ORP, and an ultrasonic flowmeter, which is used to measure the characteristics of hydrological indexes in real - time;

[0114] The water flow circulation unit includes a pusher and a corresponding motor. The pusher is installed between the wall of the short side of the reactor and the partition board. Four pushers are arranged in a single reactor to promote water circulation, and the motors of the pushers are installed on the cover plate of the reactor;

[0115] The water inlet unit includes a water inlet installed on the side of the reactor, an external water inlet pump, and a storage tank; The circulating water above the sediment initially comes from the storage tank.

[0116] Further, a water level line is set on the reactor to mark the water volume in the reactor; The size of the reactor is 100cm×40cm×60cm, and the size can be adjusted appropriately.

[0117] Further, the circulating water above the sediment is generally set as a single solution of a specified new pollutant.

[0118] Further, the distribution coefficient can be used to reflect the water - sediment interface distribution behavior of new pollutants. The calculation method of the distribution coefficient is as follows:

[0119]

[0120] In the formula: K d is the water - sediment distribution coefficient of the pollutant, L / kg; C s is the concentration of a single pollutant in the sediment, ng / g dw; C w is the concentration of a single pollutant in the water, μg / L.

[0121] Further, the step of constructing a prediction model of the water - sediment distribution coefficient based on the deep - learning algorithm specifically includes:

[0122] Dividing the experimental data into a training set and a test set at a ratio of 8:2;

[0123] Presetting the reference range of the hyperparameter space of the deep - learning model;

[0124] Constructing a deep - learning model based on the five - fold cross - validation method;

[0125] Selecting the distribution coefficient prediction model with the best prediction performance through an optimization algorithm.

[0126] Exemplarily, in the step of presetting the hyperparameter space reference range of the deep learning model, the hyperparameters of the deep learning model include the number of hidden layers, the number of hidden layer units, the learning rate, the optimizer, the regularization, and the activation function.

[0127] Furthermore, the step of building a deep learning model based on the five-fold cross-validation method specifically includes:

[0128] The training set data is divided into 5 independent equal parts, 1 of which is used as a validation set and 4 as training sets;

[0129] The water-sediment partition coefficient in the training set is used as the target variable, and a partition coefficient prediction model is constructed based on the deep learning algorithm;

[0130] Use the validation set to validate the prediction model and obtain the corresponding empirical error;

[0131] The partition coefficient prediction model with the smallest empirical error was selected and its correlation coefficient was calculated to ensure that the correlation coefficient met expectations.

[0132] Specifically, the partial dependency graph shows the marginal effect of a feature on the prediction results of the machine learning model, that is, by marginalizing other features, to show the impact of a feature on the target variable. For regression problems, the partial dependency graph needs to be based on the partial dependency function, as follows:

[0133]

[0134] Where: x S The features to be plotted for the partial dependence graph; X C In addition to x in the deep learning model S Other features besides .

[0135] A partial dependence diagram is drawn for each characteristic variable on the distribution coefficient, that is, the influence of each characteristic variable on the distribution coefficient can be analyzed.

[0136] Predicting the distribution behavior of new pollutants in the natural environment includes the following steps:

[0137] Collect the water and sediment characteristics of a natural area as characteristic variables of the model;

[0138] The characteristic variables are input into the constructed optimal prediction model to obtain the predicted value of the local new pollutant distribution coefficient.

[0139] Beneficial effects:

[0140] The system provided in this embodiment simulates the water flow in the watershed and river channel systems by setting baffles and a plug-flow device inside the reactor, changing the water flow direction and the way of plug flow in the reactor, and through the coordinated control of equipment such as a temperature control device, making the experimental environment in the reactor more stable and effectively avoiding the influence of external factors on the operation process of the reactor;

[0141] The experimental method provided in this embodiment is based on deep learning means, and a large amount of experimental data obtained by the system is used to build a machine learning model, which can effectively analyze the influence of various environmental factors on the distribution behavior of new pollutants between water and sediment.

[0142] Next, this embodiment will illustrate more specific implementation manners.

[0143] The reactor is set with dimensions of 100 cm × 40 cm × 60 cm. The baffle is located in the middle of the reactor, with a height and length of 60 cm × 60 cm and a thickness limited within 3 cm. The water inlet is set at 20 cm from the bottom on the side of the reactor and is connected to a water pump outside, and the water pump is connected to a water storage tank. The sampling port is set at 15 - 20 cm from the bottom on the side of the reactor, and the lower end of the sampling port is slightly higher than the top of the sediment. Manual valves are set at both the water inlet and the sampling port. The pusher uses a small paddle-type plug-flow device, which is set between the side of the baffle and the side of the reactor, with two sets arranged in parallel on each side. The wires on the upper part are encapsulated with a sealed insulating wire tube and connected to the motor on the reactor cover. The monitoring probes or electrodes of temperature control equipment, ultrasonic flowmeter, monitor, etc. are placed in the stable water flow between the front of the reactor and the baffle, without contacting the sediment. The handle of the surface sediment sampling spoon should ensure that sediment samples of 0 - 5 cm at the top layer of the sediment can be collected.

[0144] Prepare circulating water and sediment. Select a new pollutant A, configure a solution of a certain concentration of A with a standard sample of A, and place it in the water storage tank for standby; wash sand, powder, clay, etc. and mix them in a certain proportion to prepare sediment, and add it to the reactor. The thickness of the sediment layer is set to 15 cm.

[0145] Store an aqueous solution of A with a certain concentration in the water storage tank, inject the solution into the reactor through the water pump and the water inlet, and close the water inlet valve when the stable water level in the reactor reaches the top of the water level line. The temperature control equipment controls the temperature at 20 °C, and the pusher is not started temporarily. At this time, it is in a "static water" state.

[0146] After maintaining the "static water" state for 24 hours, a small amount of water sample is taken through the sampling port to measure the concentration of the new pollutant A in the water sample. Three sediment samples are taken from at least three positions at 0-5 cm of the sediment with a sampling spoon to measure the particle size ratio, TOC, and the concentration of the new pollutant A in the sediment respectively, and the three groups of data are averaged. The ratio of the concentration of A in the water sample to the concentration of A in the sediment is the distribution coefficient of A measured in this experiment under this condition.

[0147] Optionally, different temperatures are set through the temperature control device to measure the change of the water-sediment distribution coefficient of the new pollutant under different water temperature conditions;

[0148] Optionally, control the frequency of the pusher motor, feedback the flow rate through the ultrasonic flowmeter, and measure the change of the water-sediment distribution coefficient of the new pollutant under different water flow rates;

[0149] Optionally, adjust the sampling depth of the sampling spoon in the sediment layer to observe the concentration distribution of the new pollutant at different depths of the sediment;

[0150] Optionally, configure different types of new pollutant influents to observe the differences in the distribution coefficients of different pollutants;

[0151] Optionally, adjust the proportion of different particle sizes in the sediment to observe the influence of the sediment particle size distribution on the concentration distribution of the new pollutant.

[0152] During the experiment process, the reactor should be placed in a dark environment to avoid different degrees of photolysis of the new pollutant caused by light. After each adjustment of the reactor conditions, the reactor should be operated for 24 hours before sampling.

[0153] Record the results of multiple experiments. Take the water temperature, flow rate, pH, ORP, sediment particle size, and TOC of each experiment as the characteristic variables of the deep learning model, and take the calculation result of the distribution coefficient as the target variable to build a deep learning model.

[0154] Divide the above experimental data into a training set and a test set according to 8:2, and the division ratio can be appropriately adjusted according to the experimental requirements.

[0155] In the process of hyperparameter optimization, the method of Hyperopt is adopted to find the optimal model parameters.

[0156] Optionally, this example uses the hyperopt module in Python as the parameter tuning tool for random search. The random search space settings for the main parameters in the deep learning model are shown in Table 1:

[0157] Table 1

[0158]

[0159]

[0160] The method of five - fold cross - validation is adopted to calculate the loss returned by each random search, which is used as the objective function in the hyperparameter tuning process of hyperopt. The smaller the loss, the better the model fitting effect.

[0161] Combined with the optimized deep - learning model, a partial dependence plot is drawn to analyze the influence of a single feature variable on the target variable.

[0162] According to the variable requirements in the experiment, monitoring data such as the water quality of the river channel water body in the actual basin is collected to verify the prediction effect of the deep - learning model obtained from the experiment.

[0163] The embodiments described in the embodiments of this application are for more clearly illustrating the technical solutions of the embodiments of this application, and do not constitute a limitation on the technical solutions provided by the embodiments of this application. Those skilled in the art know that with the evolution of technology and the emergence of new application scenarios, the technical solutions provided by the embodiments of this application are equally applicable to similar technical problems.

[0164] Those skilled in the art can understand that the technical solutions shown in the figures do not constitute a limitation on the embodiments of this application, and may include more or fewer steps than those shown, or combine some steps, or different steps.

[0165] It should be understood that in this application, "at least one (item)" means one or more, and "multiple" means two or more. "And / or" is used to describe the association relationship of associated objects, indicating that three relationships can exist. For example, "A and / or B" can mean: only A exists, only B exists, and both A and B exist at the same time. Among them, A and B can be singular or plural. The character " / " generally represents an "or" relationship between the associated objects before and after. "At least one (one) of the following" or its similar expression refers to any combination of these items, including any combination of single item (one) or plural items (ones). For example, at least one (one) of a, b, or c can mean: a, b, c, "a and b", "a and c", "b and c", or "a and b and c", where a, b, c can be single or multiple.

[0166] The preferred embodiments of the embodiments of this application have been described above with reference to the accompanying drawings, and thus do not limit the scope of rights of the embodiments of this application. Any modifications, equivalent replacements, and improvements made by those skilled in the art without departing from the scope and essence of the embodiments of this application shall be within the scope of rights of the embodiments of this application.

Claims

1. A simulation system for predicting the water-sediment distribution behavior of new pollutants, characterized in that: The simulation system includes: a reactor, a baffle, a temperature control device, a hydrological index monitoring unit, a water flow circulation unit and a water inlet unit; A sediment is arranged at the bottom of the reactor, and circulating water is arranged on the top of the sediment; the circulating water is arranged as a single solution of the target new pollutant; The partition is arranged in the internal space of the reactor, the partition is connected to the bottom of the reactor, and the partition is used to separate the internal space of the reactor; The temperature control device is used to control the circulating water to be at a set temperature; The hydrological index monitoring unit is used to measure the characteristics of the hydrological index; The water circulation unit is used to promote water circulation to form the circulating water; The water inlet unit is used to connect to an external water inlet pump and a water storage tank.

2. The simulation system for predicting the water-sediment distribution behavior of new pollutants according to claim 1, characterized in that: The hydrological indicator monitoring unit includes a multi-purpose monitoring instrument and an ultrasonic flow meter; The multi-purpose monitor is used to measure temperature, pH and redox potential.

3. The simulation system for predicting the water-sediment distribution behavior of new pollutants according to claim 1, characterized in that: The water circulation unit comprises at least one flow promoter and at least one motor; Wherein, the flow promoter is installed between the wall of the reactor and the partition; The motor is used to drive the flow promoter to promote water circulation.

4. The simulation system for predicting the water-sediment distribution behavior of new pollutants according to any one of claims 1 to 3, characterized in that: The simulation system also includes a water sampling port and a sediment sampling spoon; Wherein, the water sample sampling port is arranged on the wall of the reactor.

5. An experimental method for predicting the water-sediment partitioning behavior of new pollutants, characterized in that: The experimental method is applied to the simulation system for predicting the water-sediment distribution behavior of new pollutants as claimed in any one of claims 1 to 4, and the experimental method comprises the following steps: Setting the types of water and sediment in the reactor, and setting the physical environment parameters in the reactor; When the physical environmental parameters reach the set conditions, the concentration of the new pollutant in the water body and the sediment is measured and the distribution coefficient of the new pollutant is calculated; Using the water body characteristics and sediment characteristics of the reactor during the reaction process as characteristic variables and the distribution coefficient of the new pollutant as a target variable, a prediction model for the water body-sediment distribution coefficient is constructed; Analyzing the influence of different characteristic variables on the partition coefficient of the new pollutant according to a partial dependence diagram; The prediction model is used to predict the distribution behavior of new pollutants in the target water body.

6. The experimental method for predicting the water-sediment distribution behavior of new pollutants according to claim 5, characterized in that: The method of measuring the concentration of the new pollutant in the water body and the sediment and then calculating the distribution coefficient of the new pollutant under the set conditions of the physical environmental parameters comprises the following steps: When the physical environmental parameters reach the set conditions, the concentration of a single new pollutant in the water body and the sediment is measured respectively, and then the distribution coefficient of the new pollutant is calculated; The calculation expression of the distribution coefficient is: Among them, K d represents the partition coefficient, C s represents the concentration of a single new pollutant in the sediment, C w Represents the concentration of a single new pollutant in the water body.

7. The experimental method for predicting the water-sediment distribution behavior of new pollutants according to claim 5, characterized in that: The method of constructing a prediction model of water-sediment partition coefficient by taking the water body characteristics and sediment characteristics of the reactor during the reaction process as characteristic variables and the partition coefficient of the new pollutant as the target variable comprises the following steps: Determining the feature variable labeled with the target variable as training data; Determining hyperparameters of a deep learning model and spatial reference ranges of the hyperparameters; A deep learning model is constructed according to the training data and the hyperparameters as the prediction model.

8. The experimental method for predicting the water-sediment distribution behavior of new pollutants according to claim 7, characterized in that: Determining the hyperparameters of the deep learning model comprises the following steps: The hyperparameters of the deep learning model are determined to include the number of hidden layers, the number of units in the hidden layers, the learning rate, the optimizer, the regularization, and the activation function.

9. The experimental method for predicting the water-sediment distribution behavior of new pollutants according to claim 7, characterized in that: The step of constructing a deep learning model as the prediction model according to the training data and the hyperparameters comprises the following steps: Divide the training data into five equal parts; one part of the training data is used as a validation set, and the remaining four parts of the training data are used as training sets; Using the training set and the validation set obtained after the division to construct the deep learning model based on a five-fold cross validation method; Using the validation set to validate the deep learning model, and obtaining a corresponding empirical error; Selecting the hyperparameter corresponding to the minimum empirical error as the target hyperparameter of the deep learning model; Calculating a correlation coefficient of the deep learning model set with the target hyperparameters; If the correlation coefficient meets the preset conditions, the deep learning model set with the target hyperparameters is determined as the prediction model.

10. The experimental method for predicting the water-sediment distribution behavior of new pollutants according to any one of claims 5 to 9, characterized in that: The method of analyzing the influence of different characteristic variables on the distribution coefficient of the new pollutant according to the partial dependence diagram comprises the following steps: Analyzing the influence of different characteristic variables on the distribution coefficient of the new pollutant according to the partial dependence function of the partial dependence graph; The expression of the partial dependence function is: Among them, f S (x S ) represents the partial dependence function, x S The variable feature that needs to be drawn for the partial dependence diagram; X C For the prediction model except x S Other variable characteristics other than those mentioned above.

Images