A method, device and computer readable medium for selecting a dosing regimen

Abstract

The application discloses a method, device and computer readable medium for selecting a medicament dosing scheme, which constructs an autoregressive machine learning model based on operation state information of a preset pollutant treatment process under different reaction conditions; on this basis, reaction conditions corresponding to different to-be-tested medicament dosing schemes are taken as model inputs, the model is used to predict corresponding pollutant residual concentration kinetic curves, and the medicament dosing scheme is optimized according to a preset selection rule. Through the machine learning method, the application can accurately find the mapping relationship between the reaction conditions of the pollutant treatment process and the reaction kinetic curves from data, and the accurate and intelligent optimization of the medicament dosing scheme of the pollutant treatment process is realized by using the prediction model in combination with the selection rule.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of wastewater treatment, and particularly relates to a method and device for selecting a reagent dosing scheme and a computer readable medium. BACKGROUND

[0002] Fenton oxidation technology, as a kind of efficient water treatment technology, has the characteristics of strong oxidation capacity, simple equipment and easy operation, and is often used as the final process for upgrading wastewater discharge of enterprises and treatment plants, and is widely used in actual engineering. In the Fenton reaction process of Fenton oxidation process, the reagent dosing scheme must be accurately controlled to avoid problems such as degradation of wastewater / s sewage treatment effect, cost increase and the like caused by unreasonable reagent dosing.

[0003] The traditional technology generally realizes the selection of the reagent dosing scheme through engineering experience or orthogonal experiment, but it is difficult to accurately obtain a high-quality scheme, and it is accordingly difficult to realize efficient wastewater / s sewage treatment, which affects the wastewater / s sewage treatment effect and may lead to cost increase. SUMMARY

[0004] Therefore, the application provides a method and device for selecting a reagent dosing scheme and a computer readable medium, which are used for optimizing the reagent dosing scheme based on an autoregressive machine learning model to accurately select a high-quality reagent dosing scheme and provide support for efficient wastewater / s sewage treatment.

[0005] The specific scheme is as follows:

[0006] A method for selecting a reagent dosing scheme, comprising:

[0007] obtaining reaction conditions corresponding to pollutant treatment of a to-be-treated liquid by a to-be-tested reagent dosing scheme under a preset pollutant treatment process;

[0008] inputting the reaction conditions into a prediction model, and predicting a pollutant residual concentration kinetic curve corresponding to pollutant treatment under the reaction conditions of the to-be-tested reagent dosing scheme by the prediction model; the prediction model is an autoregressive machine learning model constructed based on process operation state information of the preset pollutant treatment process under different reaction conditions in advance;

[0009] obtaining the pollutant residual concentration kinetic curves respectively predicted and output by the prediction model for different to-be-tested reagent dosing schemes;

[0010] selecting a target to-be-tested reagent dosing scheme as an optimized scheme for pollutant treatment of the to-be-treated liquid according to the obtained pollutant residual concentration kinetic curves and a preset selection rule.

[0011] Optionally, the construction process of the prediction model comprises:

[0012] determining, through experiments, corresponding pollutant residual concentration kinetic curves of the preset pollutant treatment process under different reaction conditions;

[0013] constructing a data set containing sample data; each piece of sample data includes reaction conditions and corresponding pollutant residual concentration kinetic curves of the preset pollutant treatment process under the corresponding reaction conditions;

[0014] based on a machine learning algorithm and an autoregressive algorithm, constructing an autoregressive machine learning model with the reaction conditions in the sample data contained in the data set as input and the pollutant residual concentration kinetic curves as output, to obtain the prediction model.

[0015] Optionally, the preset pollutant treatment process is a Fenton oxidation process, and the pollutant is an organic pollutant.

[0016] The liquid to be treated is contaminated water; and the pollutant residual concentration kinetic curve is composed of residual concentrations of organic pollutants at different sampling time points in the contaminated water.

[0017] The reaction conditions include the concentration, type and corresponding reagent dosing scheme of the organic pollutant in the contaminated water.

[0018] Optionally, the reagent dosing scheme includes Fe 2+ concentration and H2O2 concentration.

[0019] Optionally, constructing the autoregressive machine learning model with the reaction conditions in the sample data contained in the data set as input includes:

[0020] using a topological molecular fingerprint algorithm to encode the types of organic pollutants in the reaction conditions of the sample data of the data set into vectors, and using the encoded vectors as one of the inputs to construct the autoregressive machine learning model.

[0021] Optionally, before using the prediction model, the method further includes:

[0022] performing performance evaluation and / or model interpretation on the constructed prediction model to verify the availability of the prediction model.

[0023] Optionally, the performance evaluation and / or model interpretation on the constructed prediction model includes:

[0024] performing performance evaluation on the prediction model from at least one of accuracy, robustness, ability of the model to train based on a kinetic curve containing less than a preset number of pairs of time points and residual concentrations of organic pollutants, and ability of the model to train based on a data set less than a preset size;

[0025] A post-modeling explanation algorithm is used to perform model explanation on the prediction model.

[0026] Optionally, the target measured medicament dosing scheme is selected as the preferred scheme for the pollutant treatment of the to-be-treated liquid according to the obtained pollutant residual concentration kinetic curve and the preset selection rule.

[0027] The pollutant removal effect of each measured medicament dosing scheme that meets the effect condition is selected according to the obtained pollutant residual concentration kinetic curve.

[0028] The target measured medicament dosing scheme that meets the cost condition is selected from the selected each measured medicament dosing scheme as the preferred scheme for the pollutant treatment of the to-be-treated liquid.

[0029] A medicament dosing scheme selection device comprises:

[0030] A first acquisition module is configured to acquire reaction conditions corresponding to the pollutant treatment of a to-be-treated liquid by a measured medicament dosing scheme under a preset pollutant treatment process.

[0031] A prediction module is configured to input the reaction conditions into a prediction model, and predict a pollutant residual concentration kinetic curve corresponding to the pollutant treatment under the reaction conditions of the measured medicament dosing scheme by the prediction model.

[0032] A second acquisition module is configured to acquire pollutant residual concentration kinetic curves respectively predicted and output by the prediction model for different measured medicament dosing schemes.

[0033] A selection module is configured to select a target measured medicament dosing scheme as a preferred scheme for the pollutant treatment of the to-be-treated liquid according to the obtained pollutant residual concentration kinetic curve and a preset selection rule.

[0034] A computer readable medium having stored thereon a computer program, the computer program comprising program code for executing the method of any one of the preceding claims.

[0035] To sum up, the method, device and computer readable medium for selecting a reagent dosing scheme provided by the application, based on the process operation state information corresponding to different reaction conditions under the preset pollutant treatment process, an autoregressive machine learning model is constructed as a prediction model; on this basis, the reaction conditions corresponding to the pollutant treatment of the to-be-treated liquid by the to-be-tested reagent dosing scheme under the preset pollutant treatment process are obtained, and the prediction model is input, and the model predicts the pollutant residual concentration kinetic curve corresponding to the to-be-tested reagent dosing scheme under the required reaction conditions, and then, based on the prediction model, the pollutant residual concentration kinetic curves predicted and output for different to-be-tested reagent dosing schemes are further predicted and output, and the preset selection rule is used to optimize the reagent dosing scheme.

[0036] The application can accurately find the mapping relationship between the reaction conditions and the reaction kinetic curve of the pollutant treatment process from the data by the machine learning method, and the prediction model constructed by the machine learning is combined with the selection rule to realize the accurate and intelligent optimization of the reagent dosing scheme of the pollutant treatment process, so as to guarantee the treatment effect of the to-be-treated liquid such as wastewater / sewage. BRIEF DESCRIPTION OF DRAWINGS

[0037] The above and other features, advantages and aspects of the embodiments of the application will become more apparent by referring to the following detailed description in conjunction with the accompanying drawings. Throughout the drawings, the same or similar reference numerals indicate the same or similar elements. It should be understood that the drawings are schematic, and the original and elements are not necessarily drawn according to the scale.

[0038] Figure 1 is a flowchart of constructing a prediction model provided by the application;

[0039] Figure 2 is a model architecture diagram of the prediction model provided by the application;

[0040] Figure 3 is a model performance evaluation result diagram provided by the application;

[0041] Figure 4 is a model explanation result diagram provided by the application;

[0042] Figure 5 is a flowchart of the method for selecting a reagent dosing scheme provided by the application;

[0043] Figure 6 is a component structure diagram of the reagent dosing scheme selection device provided by the application. DETAILED DESCRIPTION

[0044] Embodiments of the present application will be described in more detail with reference to the drawings. While certain embodiments of the present application are shown in the drawings, it is understood that the present application can be embodied in various forms and should not be construed as being limited to the embodiments set forth herein, but rather should be interpreted in the context of what is claimed. It is understood that the drawings and embodiments of the present application are for illustrative purposes only and are not intended to limit the scope of the present application.

[0045] The term "comprising" and variations thereof as used herein are used inclusively, i.e., "comprising, but not limited to". The term "based on" means "based, at least in part, on". The term "one embodiment" means "at least one embodiment"; the term "another embodiment" means "at least one additional embodiment"; the term "some embodiments" means "at least some embodiments". Related terms are defined in the description below.

[0046] It should be noted that the terms "first", "second", and the like in the present application are only used to distinguish different devices, modules or units, and are not intended to limit the functions of these devices, modules or units or the sequence or interdependence of these functions.

[0047] It should be noted that the terms "one", "multiple" in the present application are illustrative and not restrictive, and those skilled in the art should understand that unless otherwise explicitly stated in the context, it should be understood as "one or more".

[0048] In the Fenton reaction process, the dosage scheme must be accurately controlled to avoid unreasonable dosage and cost increase and effect decline. However, the applicant found that due to the influence of multiple factors such as influent water quality, water quantity, pH value and Fenton reagent dosage on Fenton reaction, the traditional way of optimizing the dosage scheme through engineering experience or orthogonal experiment is difficult to accurately obtain a high-quality dosage scheme, which ultimately affects the wastewater / sewage treatment effect and may lead to cost increase.

[0049] To solve the above technical problems and realize accurate optimization of the dosage scheme, the present application discloses a method, device and computer readable medium for selecting a dosage scheme.

[0050] The method disclosed in the present application is pre-constructed with a prediction model, which is an autoregressive machine learning model constructed based on the process running condition information corresponding to different reaction conditions of a preset pollutant treatment process.

[0051] Machine learning method can discover and mine the potential value contained in data by studying how computers simulate or implement human learning behavior. The present application discovers the mapping relationship between the reaction conditions of the preset pollutant treatment process and the reaction kinetics curve (pollutant residual concentration kinetics curve) from the data by the machine learning method, so as to train and construct a prediction model that can be used to predict the pollutant residual concentration kinetics curve under the given reaction conditions based on the preset pollutant treatment process, thereby providing an evaluation basis for the dosing scheme corresponding to the given reaction conditions.

[0052] Referring to Figure 1 , the construction process of the prediction model includes:

[0053] Step 101, determining the pollutant residual concentration kinetics curve corresponding to different reaction conditions of the preset pollutant treatment process through experiments.

[0054] Different reaction conditions, in particular, refer to different reaction conditions required for treating the liquid by using multiple different dosing schemes of the preset pollutant treatment process.

[0055] Optionally, the preset pollutant treatment process is Fenton oxidation process, and the liquid to be treated can be contaminated water such as wastewater, sewage, etc., and the pollutant in the liquid to be treated is organic pollutant. In actual application, the type of the preset pollutant treatment process, the liquid to be treated, and the pollutant can be determined according to the demand.

[0056] The reaction conditions include the concentration, type of the organic pollutant in the contaminated water, and the corresponding dosing scheme.

[0057] For Fenton oxidation process, the dosing scheme in the model training and use stage includes Fe 2+ concentration and H2O2 concentration.

[0058] The present application embodiment sets multiple different types and concentrations of organic pollutants, and forms different dosing schemes based on Fe 2+ and H2O2 of different concentrations, and constructs multiple groups of reaction conditions containing the concentration, type of the organic pollutant, and the corresponding dosing scheme.

[0059] When constructing the above reaction conditions, preferably:

[0060] The type of the organic pollutant is selected from phenol, hydroquinone, p-cresol, p-nitrophenol, p-chlorophenol, p-hydroxybenzyl alcohol, p-hydroxybenzaldehyde, p-hydroxyacetophenone, p-hydroxybenzoic acid, p-hydroxybenzoic acid methyl ester, p-hydroxyphenylacetamide, and p-hydroxyanisole as alternatives;

[0061] The concentration of the organic pollutant is selected from 10 mg / L, 50 mg / L, and 100 mg / L as alternatives;

[0062] Fe 2+ Concentrations are alternatively 1, 2.5, 5, 7.5, 10, 15, 20 mg / L;

[0063] H2O2 concentrations are alternatively 5, 7.5, 10, 12.5, 15, 20, 30, 40, 50, 70 mg / L.

[0064] For each set of reaction conditions, a corresponding pollutant residual concentration kinetic curve is determined by experiment for the Fenton oxidation process under the reaction conditions for treating wastewater.

[0065] An alternative example of determining pollutant residual concentration kinetic curves by experiment is provided below.

[0066] In this example, the Fenton reaction is carried out in a 25 mL beaker containing 20 mL of a solution of organic pollutant (e.g., contaminated water) at a set concentration, the temperature of the solution is controlled at 20°C using a circulating water cooler, and the pH of the solution is adjusted to 3.0 ± 0.2 using 0.1 M H2SO4 (i.e., 0.1 moles of sulfuric acid per liter of solution). Then, according to the set reaction conditions, a certain volume of H2O2 and FeSO4 stock solutions are added to the solution to initiate the reaction, 0.5 mL of the reaction solution is taken at a given number of different time points, the reaction is terminated using 0.5 mL of ascorbic acid solution (5 g / L), and the concentration of the organic pollutant is analyzed using ultra-high performance liquid chromatography with a C18 column.

[0067] For example, the sampling time points are 2, 5, 9, 15, 22, and 30 minutes after the reaction starts. Then, using the concentration test results at each time point, a pollutant residual concentration kinetic curve under the corresponding reaction conditions is constructed, which is composed of the residual concentrations of the organic pollutant at different sampling time points in the solution.

[0068] A plurality of sets of reaction conditions respectively correspond to a plurality of sets of organic pollutant residual concentration kinetic curves obtained by experiment.

[0069] Step 102, constructing a data set containing sample data; each piece of sample data includes reaction conditions and a pollutant residual concentration kinetic curve corresponding to the pre-set pollutant treatment process under the corresponding reaction conditions.

[0070] After a plurality of sets of reaction conditions respectively correspond to a plurality of sets of organic pollutant residual concentration kinetic curves are obtained by experiment, a data set is constructed based on the experimental results, and the data set includes a plurality of pieces of sample data, each piece of sample data includes a set of reaction conditions and a corresponding pollutant residual concentration kinetic curve.

[0071] Optionally, the data set is divided into a training set and a test set in a certain proportion, which are used for model training and evaluation, respectively. For example, the data set includes 300 groups of reaction conditions and their corresponding organic pollutant kinetic curves, and is divided into a training set and a test set in a ratio of 8:2.

[0072] Step 103, based on the machine learning algorithm and the autoregressive algorithm, an autoregressive machine learning model is constructed with the reaction conditions in the sample data in the data set as input and the pollutant residual concentration kinetic curve as output, to obtain the prediction model.

[0073] In this embodiment, the autoregressive machine learning model is trained with the reaction conditions in the sample data in the training set as input and the pollutant residual concentration kinetic curve in the sample data as output, and the prediction model is obtained after training.

[0074] Preferably, the machine learning algorithm is XGBoost algorithm.

[0075] For organic pollutant species, a topological molecular fingerprint algorithm is used to encode it into a vector.

[0076] Preferably, the topological molecular fingerprint is Morgan fingerprint.

[0077] The vector obtained by encoding the organic pollutant species is a numerical vector, which can be converted from the chemical formula of the organic pollutant molecule by using Morgan molecular fingerprint algorithm.

[0078] The numerical vector can be in the form of a mathematical matrix.

[0079] Correspondingly, in the model training, the numerical vector obtained by converting the chemical formula of the organic pollutant molecule by using Morgan molecular fingerprint algorithm, and the pollutant, Fe 2+ and H2O2 concentration as input, and the pollutant residual concentration kinetic curve in the sample data as output, and combining the autoregressive algorithm and the XGBoost algorithm, a corresponding prediction model is trained, and the model architecture can be referred to as shown in Figure 2 .

[0080] Through the above model training process, the accurate simulation of the whole process concentration change of the organic pollutant in the Fenton oxidation process is realized. Based on the trained prediction model, the accurate optimization of the Fenton oxidation process reagent dosing scheme is carried out, which overcomes the previous selection mode relying on artificial experience, effectively improves the wastewater treatment efficiency and quality of the Fenton oxidation process, ensures that the effluent meets the standard, and greatly reduces the wastewater treatment cost.

[0081] Optionally, after obtaining the prediction model by training the autoregressive machine learning, the performance of the prediction model is further evaluated and / or the model is explained to verify the availability of the model.

[0082] wherein the performance of the prediction model is evaluated from at least one of accuracy, robustness, ability of the model to be trained based on a kinetic curve containing less than a preset number of pairs of time points and residual concentration of organic pollutants, and ability of the model to be trained based on a data set less than a preset size.

[0083] That is, the evaluation content can include, but is not limited to, any one or more of model accuracy, robustness, ability of the model to be trained based on a kinetic curve containing a small number of time-concentration points, and ability to be trained based on a small size data set.

[0084] Preferably, the evaluation index includes mean absolute error (MAE), root mean square error (RMSE), and determination coefficient (r2) of the model.

[0085] The prediction model is evaluated by using the constructed test set, and the evaluation results are shown in Table 1. Figure 3 wherein (a) represents the accuracy evaluation result, (b) represents the robustness evaluation result, (c) represents the evaluation result of “ability to be trained based on a kinetic curve containing a small number of time-concentration points”, and (d) represents the evaluation result of “ability to be trained based on a small size data set”.

[0086] According to the evaluation results, the prediction model has small error and high accuracy on the test set, with MAE, RMSE, and R2 being 4.015%, 5.326%, and 0.935, respectively. To evaluate the robustness of the model, Gaussian errors are artificially introduced into the training data. After introducing random errors with a standard deviation of 2%, the model still maintains high accuracy, with MAE, RMSE, and R2 being 4.511%, 6.012%, and 0.904, respectively, indicating that the model has good robustness. In addition, the number of time-pollutant concentration data points contained in the kinetic curve in the training set is reduced, and when the number of data points is 3, the model still maintains good accuracy, with MAE, RMSE, and R2 being 5.346%, 7.135%, and 0.897, respectively, indicating that the model can be applied to kinetic curves containing a small number of data points. Furthermore, the size of the training set is continuously reduced, and when the number of input-output data pairs contained in the training set reaches 40, the model still maintains good accuracy, with MAE, RMSE, and R2 being 8.357%, 10.549%, and 0.735, respectively, indicating that the model can maintain good performance on a small data set.

[0087] Furthermore, this application utilizes a post-modeling interpretation algorithm to interpret the prediction model. Specifically, the ALE (accumulated local effect) algorithm can be used, but is not limited to, to estimate the correlation between reaction conditions and reaction rate.

[0088] The results of estimating the correlation between reaction conditions and reaction rate using the ALE algorithm are as follows: Figure 4 As shown, (a) represents Fe 2+ The correlation between concentration and reaction rate, (b) shows the correlation between H2O2 concentration and reaction rate, and (c) shows the correlation between Fe concentration and reaction rate. 2+ The correlation between the H2O2 concentration ratio and the reaction rate is shown in (d), which represents the correlation between the organic pollutant concentration and the reaction rate. According to... Figure 4 It can be seen that the reaction rate increases with Fe 2+ The reaction accelerates with increasing H2O2 concentration because the production of more reactive substances speeds up the degradation of organic pollutants. Fe 2+ An excessively high Fe concentration to H2O2 ratio will lead to a decreased degradation rate, due to the excess Fe. 2+ This leads to an increase in the proportion of ineffective H2O2 decomposition. Furthermore, the reaction rate also accelerates with increasing organic pollutant concentration. All of these analyses are consistent with our understanding of the Fenton reaction, indicating that the predictive model has good reliability.

[0089] This embodiment confirms the performance and reliability of the prediction model through model evaluation and interpretation, and verifies the usability of the prediction model.

[0090] Based on the trained and validated prediction model, see [link / reference]. Figure 5 The method for selecting the drug dosing regimen disclosed in this application includes:

[0091] Step 501: Obtain the reaction conditions corresponding to the treatment of the liquid under the pollutant treatment process by using the test reagent dosing scheme under the preset pollutant treatment process.

[0092] As mentioned above, the preset pollutant treatment process can be the Fenton oxidation process, the liquid to be treated can be polluted water, such as wastewater or sewage, and the pollutants in the liquid to be treated are organic pollutants.

[0093] For the dosing scheme of the reagent to be tested, the corresponding reaction conditions can be obtained, for example, including the concentration and type of organic pollutants in the water, and the Fe content provided by the dosing scheme of the reagent to be tested. 2+ Concentration, H2O2 concentration.

[0094] Step 502, input the reaction condition into the prediction model, and predict the pollutant residual concentration kinetic curve corresponding to the pollutant treatment under the reaction condition of the to-be-tested agent dosing scheme by the prediction model; the prediction model is an autoregressive machine learning model constructed in advance based on the process operation state information of the preset pollutant treatment process under different reaction conditions.

[0095] Then, the obtained reaction condition is input into the prediction model, wherein the topological molecular fingerprint algorithm is used to encode the organic pollutant species into a vector.

[0096] The prediction model receives the input reaction condition and outputs the pollutant residual concentration kinetic curve matched therewith. For a plurality of to-be-tested agent dosing schemes, the prediction model can finally obtain the pollutant residual concentration kinetic curve as the output corresponding to the reaction condition of each to-be-tested agent dosing scheme.

[0097] Step 503, obtain the pollutant residual concentration kinetic curves respectively predicted and output by the prediction model for different to-be-tested agent dosing schemes.

[0098] Step 504, select a target to-be-tested agent dosing scheme as a preferred scheme for treating the to-be-treated liquid according to the obtained pollutant residual concentration kinetic curves and a preset selection rule.

[0099] The preset selection rule can be, but is not limited to, any one of the following:

[0100] Rule one: a rule formed by a set pollutant removal effect condition;

[0101] Rule two: a rule formed by a set pollutant removal effect condition and a cost condition.

[0102] In actual application, rule two can be preferred.

[0103] For example, the selection rule is set to be that the organic pollutant removal rate reaches at least 80% at 30 minutes and the agent cost is the lowest.

[0104] After obtaining the pollutant residual concentration kinetic curves corresponding to a plurality of to-be-tested agent dosing schemes based on the prediction model, the obtained pollutant residual concentration kinetic curves are further used as the basis to select the agent dosing scheme from the to-be-tested agent dosing schemes according to the above preset selection rule.

[0105] With the preset selection rule as the rule two, the predicted pollutant residual concentration kinetic curve can be used as a basis. First, each of the to-be-tested reagent dosing schemes that meet the effect condition is selected. Then, a scheme that meets the cost condition is selected from the schemes that meet the effect condition, as the preferred scheme for treating the liquid to be treated.

[0106] Preferably, the reagent dosing cost is calculated by the following formula:

[0107]

[0108] wherein, Cost refers to the reagent dosing cost, C F , C H , C P respectively refer to the concentrations of Fe 2+ , H2O2 and organic pollutants, P F and P H respectively refer to the prices of unit mass of Fe 2+ and H2O2, for example, 400 yuan / ton and 600 yuan / ton.

[0109] In actual application, all possible schemes can be traversed to select the global optimal result based on the constructed prediction model and the set selection rule.

[0110] The following provides an application example of the method of the present application:

[0111] Suppose the type of organic pollutants is phenol, and the pollutant concentration is 10 mg / L. All the Fenton reagent dosing schemes composed of the selected Fe 2+ and H2O2 concentrations are traversed. The selected Fe 2+ concentration is 1, 2.5, 5, 7.5, and 10 mg / L, and the selected H2O2 concentration is 5, 7.5, 10, and 12.5 mg / L, forming 5*4, i.e., 20 to-be-tested reagent dosing schemes. Based on the prediction model and the selection rule, first, the Fenton reagent dosing scheme that can make the pollutant removal rate reach 80% in 30 minutes is obtained. Then, from the obtained Fenton reagent dosing schemes, the one with the lowest reagent cost is selected as the preferred scheme. For example, for 10 mg / L of phenol pollutants, the preferred scheme is 2.5 mg / L of Fe 2+ and 10 mg / L of H2O2.

[0112] To sum up, the method for selecting a reagent addition scheme provided in the application, based on the process operation state information of the preset pollutant treatment process under different reaction conditions, constructs an autoregressive machine learning model as a prediction model; on this basis, the reaction conditions required for treating the liquid to be treated by using the reagent addition scheme to be tested under the preset pollutant treatment process are obtained, and the prediction model is input, and the model predicts the pollutant residual concentration kinetic curve corresponding to the reagent addition scheme to be tested under the required reaction conditions, and then, based on the pollutant residual concentration kinetic curves respectively predicted and output by the prediction model for different reagent addition schemes to be tested and the preset selection rule, the reagent addition scheme is optimized.

[0113] The application can accurately find the mapping relationship between the reaction conditions and the reaction kinetic curve of the pollutant treatment process from the data by the machine learning method, and realizes accurate and intelligent optimization of the reagent addition scheme of the pollutant treatment process by using the prediction model constructed by machine learning combined with the selection rule, so as to guarantee the treatment effect of the liquid to be treated such as wastewater / sewage.

[0114] Corresponding to the above method, the application also provides a device for selecting a reagent addition scheme, the composition structure of the device is as shown in Figure 6 The device comprises:

[0115] A first acquisition module 601 is configured to acquire reaction conditions corresponding to pollutant treatment of a liquid to be treated by using a reagent addition scheme to be tested under a preset pollutant treatment process.

[0116] A prediction module 602 is configured to input the reaction conditions into a prediction model, and predict a pollutant residual concentration kinetic curve corresponding to pollutant treatment under the reaction conditions of the reagent addition scheme to be tested by the prediction model; the prediction model is an autoregressive machine learning model constructed based on process operation state information of the preset pollutant treatment process under different reaction conditions.

[0117] A second acquisition module 603 is configured to acquire pollutant residual concentration kinetic curves respectively predicted and output by the prediction model for different reagent addition schemes to be tested.

[0118] A selection module 604 is configured to select a target reagent addition scheme to be tested as an optimized scheme for pollutant treatment of the liquid to be treated according to the acquired pollutant residual concentration kinetic curves and a preset selection rule.

[0119] In an embodiment, the above device further comprises a model construction module configured to:

[0120] The pollutant residual concentration kinetic curves corresponding to the preset pollutant treatment process under different reaction conditions are determined by experiments.

[0121] constructing a dataset containing sample data; each piece of sample data comprises reaction conditions and a corresponding pollutant residual concentration kinetic curve of the preset pollutant treatment process under the reaction conditions;

[0122] based on a machine learning algorithm and an autoregressive algorithm, constructing an autoregressive machine learning model with the reaction conditions in the sample data contained in the dataset as input and the pollutant residual concentration kinetic curve as output, to obtain the prediction model.

[0123] In an embodiment, the preset pollutant treatment process is a Fenton oxidation process, and the pollutant is an organic pollutant.

[0124] The liquid to be treated is contaminated water; and the pollutant residual concentration kinetic curve is composed of residual concentrations of organic pollutants at different sampling time points in the contaminated water.

[0125] The reaction conditions include the concentration, type and corresponding reagent dosing scheme of the organic pollutant in the contaminated water.

[0126] In an embodiment, the reagent dosing scheme includes Fe 2+ concentration and H2O2 concentration.

[0127] In an embodiment, when the model construction module constructs an autoregressive machine learning model with the reaction conditions in the sample data contained in the dataset as input, it is specifically configured to: use a topological molecular fingerprint algorithm to encode the types of organic pollutants in the reaction conditions of the sample data of the dataset into vectors, and use the encoded vectors as one of the inputs to construct the autoregressive machine learning model.

[0128] In an embodiment, the above device further comprises a model verification module configured to: before using the prediction model, perform performance evaluation and / or model interpretation on the constructed prediction model to verify the availability of the prediction model.

[0129] In an embodiment, when the model verification module performs performance evaluation and / or model interpretation on the constructed prediction model, it is specifically configured to:

[0130] performing performance evaluation on the prediction model from at least one of the following: accuracy, robustness, the ability of the model to train a kinetic curve based on a pair of data containing less than a preset number of time points and residual concentrations of organic pollutants, and the ability of the model to train based on a dataset less than a preset size;

[0131] performing model interpretation on the prediction model using a post-modeling interpretation algorithm.

[0132] In an embodiment, the selection module 604 is specifically configured to:

[0133] According to the obtained pollutant residual concentration kinetic curve, each to-be-tested medicament dosing scheme satisfying the pollutant removal effect condition is selected;

[0134] A target to-be-tested medicament dosing scheme satisfying the cost condition is selected from the selected each to-be-tested medicament dosing scheme as a preferred scheme for pollutant treatment of the to-be-treated liquid.

[0135] For the medicament dosing scheme selection device provided in the embodiments of the present application, since it corresponds to the medicament dosing scheme selection method provided in the method embodiments above, the description is relatively simple, and the relevant similarities can be seen from the description of the method embodiments above, which will not be described in detail here.

[0136] The present application also provides a computer readable medium having a computer program stored thereon, the computer program comprising program codes for executing the medicament dosing scheme selection method provided in any of the method embodiments above.

[0137] In the context of the present application, the computer readable medium (machine readable medium) can be a tangible medium, which can contain or store programs for use by or in conjunction with an instruction execution system, apparatus or device. The machine readable medium can be a machine readable signal medium or a machine readable storage medium. The machine readable medium can include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus or device, or any suitable combination of the above. More specific examples of machine readable storage media can include one or more wires, portable computer disks, hard disks, random access memory (RAM), read only memory (ROM), erasable programmable read only memory (EPROM or flash memory), optical fiber, compact disk read only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination of the above.

[0138] It should be noted that the computer-readable medium described above can be a computer-readable signal medium or a computer-readable storage medium or a combination thereof. The computer-readable storage medium can be, for example, but not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any suitable combination of the above. More specific examples of the computer-readable storage medium can include, but are not limited to, an electrical connection having one or more wires, a portable computer diskette, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or Flash memory), an optical fiber, a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In the present application, the computer-readable storage medium can be any tangible medium that contains or stores a program used by or in connection with an instruction execution system, apparatus, or device. In the present application, the computer-readable signal medium can include a computer-readable storage medium in a baseband or propagated as a carrier wave in a propagated signal, where the computer-readable program code is contained in the baseband or propagated as the carrier wave in the propagated signal. Such a propagated signal can take a variety of forms, including, but not limited to, electro-magnetic, optical, or any suitable combination thereof. Computer-readable signal media can also be any computer-readable medium that is not a storage medium and that can communicate or propagate program code, such as computer-readable program code embodied in a computer-readable medium for communication or propagated as a carrier wave. Program code embodied on a computer-readable medium can be transmitted using any appropriate medium, including, but not limited to, wireless, wire line, optical fiber cable, RF, etc., or any suitable combination of the above.

[0139] The computer-readable medium described above can be contained in an electronic device or can exist separately from the electronic device.

[0140] It should be noted that although the subject matter has been described in language specific to structural features and / or methodological acts, the subject matter defined in the appended claims is not necessarily limited to the specific features or acts described above. Rather, the specific features and acts described above are disclosed as example forms of implementing the claims.

[0141] Although specific implementation details have been included in the above discussion, these should not be construed as limiting the scope of the application. Certain features described in the context of separate embodiments can also be implemented in combination with each other. Conversely, various features described in the context of a single embodiment can also be implemented separately or in any appropriate subcombination. For example, the features described in the context of the first embodiment can be implemented in the context of the second embodiment, and vice versa.

[0142] The above description is only the preferred embodiment of the present application and the explanation of the technical principles. It should be understood by those skilled in the art that the application scope of the present application is not limited to the technical solutions with the specific combination of the above technical features, and should also cover other technical solutions formed by combining the above technical features or their equivalent features without departing from the application concept. For example, the technical solutions formed by replacing the above features with the technical features with similar functions applied in the present application (but not limited to) with each other.

Claims

1. A method for selecting a drug dosing scheme, characterized in that, include: Obtain the reaction conditions corresponding to the treatment of the liquid under the pollutant treatment process using the reagent dosing scheme under the preset pollutant treatment process; the preset pollutant treatment process is the Fenton oxidation process, and the pollutant is an organic pollutant; The reaction conditions are input into a prediction model, which predicts the pollutant residual concentration kinetic curve corresponding to the pollutant treatment under the reaction conditions of the test reagent dosing scheme. The prediction model is an autoregressive machine learning model constructed in advance based on the process operation information of the preset pollutant treatment process under different reaction conditions. The liquid to be treated is polluted water. The pollutant residual concentration kinetic curve consists of the residual concentration of organic pollutants in the polluted water at different sampling time points. The reaction conditions include the concentration and type of organic pollutants in the polluted water and the corresponding reagent dosing scheme. Using the reaction conditions in the sample data contained in the dataset as input, an autoregressive machine learning model is constructed, including: using a topological molecular fingerprint algorithm to encode the types of organic pollutants in the reaction conditions contained in the sample data of the dataset into vectors, and using the encoded vectors as one of the inputs to construct the autoregressive machine learning model. Obtain the pollutant residual concentration kinetic curves predicted and output by the prediction model for different test reagent dosing schemes; Based on the obtained pollutant residual concentration kinetic curve and preset selection rules, the target reagent dosing scheme is selected as the preferred scheme for treating the pollutants in the liquid to be treated.

2. The method according to claim 1, characterized in that, The process of constructing the prediction model includes: The kinetic curves of the residual pollutant concentration under different reaction conditions were determined through experiments. Construct a dataset containing sample data; each sample data includes reaction conditions and the kinetic curve of the residual concentration of pollutants under the corresponding reaction conditions of the preset pollutant treatment process. Based on machine learning and autoregressive algorithms, an autoregressive machine learning model is constructed using the reaction conditions in the sample data contained in the dataset as input and the pollutant residual concentration kinetic curve as output, to obtain the prediction model.

3. The method according to claim 1, characterized in that, The reagent dosing regimen includes Fe 2+ Concentration and H2O2 concentration.

4. The method according to claim 1, characterized in that, Before using the prediction model, the following are also included: The constructed prediction model is subjected to performance evaluation and / or model interpretation to verify its usability.

5. The method according to claim 4, characterized in that, The performance evaluation and / or model interpretation of the constructed prediction model includes: The predictive model is evaluated based on at least one of the following: accuracy, robustness, ability to train the model on kinetic curves containing fewer than a preset number of time points and residual concentrations of organic pollutants, and ability to train the model on a dataset of a smaller than a preset size. The prediction model is interpreted using a post-modeling interpretation algorithm.

6. The method according to claim 1, characterized in that, The step of selecting a target reagent dosing scheme as the preferred scheme for treating the liquid to be treated, based on the obtained pollutant residual concentration kinetic curve and preset selection rules, includes: Based on the obtained pollutant residual concentration kinetic curves, select the dosing schemes of each test reagent that meet the effect conditions for pollutant removal. Choose the target reagent dosing scheme that meets the cost requirements from the selected reagent dosing schemes, and use it as the preferred scheme for treating the contaminants in the liquid to be treated.

7. A device for selecting a drug dosing scheme, characterized in that, include: The first acquisition module is used to acquire the reaction conditions corresponding to the treatment of the liquid under the pollutant treatment process by using the test reagent dosing scheme under the preset pollutant treatment process; the preset pollutant treatment process is the Fenton oxidation process, and the pollutant is an organic pollutant. The prediction module is used to input the reaction conditions into a prediction model, which then predicts the pollutant residual concentration kinetic curve corresponding to the pollutant treatment under the reaction conditions of the test reagent dosing scheme. The prediction model is an autoregressive machine learning model pre-constructed based on the process operation information of the preset pollutant treatment process under different reaction conditions. The liquid to be treated is polluted water. The pollutant residual concentration kinetic curve consists of the residual concentration of organic pollutants in the polluted water at different sampling time points. The reaction conditions include the concentration and type of organic pollutants in the polluted water and the corresponding reagent dosing scheme. The model building module, when constructing an autoregressive machine learning model using the reaction conditions in the sample data contained in the dataset as input, specifically uses the topological molecular fingerprint algorithm to encode the types of organic pollutants in the reaction conditions contained in the sample data of the dataset into vectors, and uses the encoded vectors as one of the inputs to construct an autoregressive machine learning model. The second acquisition module is used to acquire the pollutant residual concentration kinetic curves predicted and output by the prediction model for different dosing schemes of the test reagents; The selection module is used to select the target reagent dosing scheme as the preferred scheme for treating the liquid to be treated, based on the obtained pollutant residual concentration kinetic curve and preset selection rules.

8. A computer-readable medium, characterized in that, It stores a computer program thereon, the computer program containing program code for performing the method as described in any one of claims 1-6.

Images