Method and system for reconstructing wastewater biological treatment process based on machine learning

Abstract

The present invention discloses a method and system for reconstructing a wastewater biological treatment process based on machine learning, which includes: collecting basic water quality, ecological risk and process information data of the water supply and discharge of the wastewater biological treatment process and performing data preprocessing; performing feature extraction and multi-unit characterization for the wastewater biological treatment process; building a wastewater basic water quality prediction model and a wastewater ecological risk prediction model for different biological treatment multi-unit combination processes and performing standard achievement detection; simulating and reconstructing short unit combination processes for different wastewater biological treatments, and performing effluent quality standard achievement detection under different feedwater quality scenarios; evaluating the ecological risk of the reconstructed short unit combination process for the biological treatment in which the effluent quality achieves the standard, and determining the optimal wastewater biological treatment process. The present invention can determine a new biological treatment process that meets the basic water quality of the effluent meets the standard and has the lowest biological risk for different feedwater quality scenarios, and solve the selection constraints of traditional biological treatment process methods.

Description

[Technical field]

[0001] The present invention relates to wastewater treatment technology, and more particularly to a method and system for reconstructing a wastewater biological treatment process based on machine learning. [Background technology]

[0002] Sewage treatment plants are an important place for solving the problems of purification and treatment of produced and domestic sewage and standard-achieving discharge, and biological treatment is one of the most widespread wastewater treatment technologies currently used in sewage treatment plants. Among them, the most common biological treatment processes include anaerobic-anoxic-aerobic or anoxic-anaerobic-aerobic A2O process, anoxic-aerobic or anaerobic-aerobic AO process, oxidation ditch OD process, membrane bioreactor MBR process, and circulating activated sludge CAST process.

[0003] At present, there are few studies on the reconstruction of new wastewater biological treatment processes, and most of the studies are only applicable to specified application scenarios and have strict implementation conditions. A Chinese patent with application number 202310107389.9 discloses a wastewater biological treatment process for simultaneous removal of nitrogen, sulfur and phosphorus, which realizes simultaneous removal of three kinds of pollutants, nitrogen, sulfur and phosphorus, at low carbon-nitrogen ratios through alternating reactions between anaerobic and anoxic stages. A Chinese patent with application number 200810104988.0 discloses a method for rapid realization of short-depth denitrification by SBR, which includes the steps of water supply, aeration, adding an equal amount of sewage, aeration, adding an equal amount of sewage again, aeration, adding an external carbon source, aeration, settling, and standing, and realizes short-depth denitrification by SBR for wastewater with low carbon-nitrogen ratios. A Chinese patent with application number 200810104989.5 discloses a method for achieving short-depth denitrification under low temperature conditions, which includes the steps of starting up short-depth denitrification, stabilizing and cooling, and includes the steps of water supply, aeration, denitrification, sedimentation, drainage, sludge discharge and standing, respectively, to achieve stable denitrification under low temperature conditions. The above methods are all optimization and improvement of traditional biological treatment process methods, and are used to enhance the water purification ability of one or several pollutants in a specific application scenario through methods such as adding impurities and changing reaction conditions, but cannot meet the demands of water purification and ecological risk control in different application scenarios, and still have many limitations and cannot meet the current increasing demand for wastewater treatment. Summary of the Invention [Problem to be solved by the invention]

[0004] Objective of the Invention: The objective of the present invention is to provide a method and system for reconstructing a wastewater biological treatment process based on machine learning, which can determine a new biological treatment process that satisfies the basic water quality of the effluent and has the lowest biological risk for different supply water quality scenarios, and resolve the selection constraints of conventional biological treatment process methods. [Means for solving the problem]

[0005] Technical solution: The method for reconstructing a wastewater biological treatment process based on machine learning described in the present invention includes: (1) collecting and pre-processing basic water quality, ecological risk and process information data of water supply and discharge in a wastewater biological treatment process; (2) performing feature extraction and multi-unit characterization of a wastewater biological treatment process; (3) constructing a wastewater basic water quality prediction model and a wastewater ecological risk prediction model for different biological treatment multi-unit combination processes and performing standard achievement detection; (4) simulating and reconstructing the short unit combination process for different wastewater biological treatment, and detecting the achievement of effluent quality standards under different supply water quality scenarios; and (5) evaluating the ecological risk of the reconstructed short unit combination process for biological treatment that achieves the effluent quality standard, and determining the optimal wastewater biological treatment process.

[0006] Preferably, in step (1), the basic water quality of the water supply and wastewater from the wastewater biological treatment process includes chemical oxygen demand, ammonia nitrogen, total nitrogen, total phosphorus, suspended solids, biochemical oxygen demand, animal and vegetable oils, petroleum, nitrate nitrogen, total organic carbon, dissolved oxygen, temperature, pH value and fecal coliform count, the water supply and wastewater ecological risk data includes trace pollutant content and biological toxicity, and the process information of the wastewater biological treatment process includes information data of the wastewater biological treatment process using an activated sludge method.

[0007] Preferably, step (2) comprises: The method includes: performing feature extraction on the biopond links of the wastewater biological treatment process using the activated sludge method, and dividing the biopond links into three feature units, i.e., aerobic ponds, anoxic ponds and anaerobic ponds, based on dissolved oxygen; and performing multi-unit characterization on the wastewater biological treatment process using the activated sludge method according to the extracted feature units.

[0008] Preferably, the multi-unit characteristic method includes a one-hot algorithm.

[0009] Preferably, performing multi-unit characterization for a biological treatment process using a one-hot algorithm comprises: First, we express the aerobic pond unit as [1,0,0], the anoxic pond unit as [0,1,0], the anaerobic pond unit as [0,0,1], and the blank unit as [0,0,0]. The method further includes performing feature splicing for the biological pond unit, performing n-unit characterization collectively for the wastewater biological treatment process using the activated sludge method, and supplementing with blank units when the number of units of the represented process is insufficient.

[0010] Preferably, the step (3) of constructing a effluent basic water quality prediction model of the different biological treatment multi-unit combined processes includes: The method includes: determining an input feedwater quality feature set and presetting a effluent quality target; selecting different machine learning algorithms to respectively establish effluent basic water quality prediction models for different biological treatment multi-unit combination processes based on the input feedwater quality feature set; and conducting cross-validation experiments to evaluate the performance of the prediction models to obtain optimal effluent basic water quality prediction models; and screening effluent ecological risk prediction models that achieve the pre-set effluent quality target. The effluent quality target includes but is not limited to effluent total phosphorus, effluent total nitrogen, ammonia nitrogen, nitrate nitrogen, biochemical oxygen demand, total organic carbon, etc. One or more different effluent basic water quality prediction models can be constructed and standard achievement detection screening can be performed.

[0011] Preferably, the step (3) of constructing the wastewater ecological risk prediction model of the different biological treatment multi-unit combined processes includes: The method includes: determining an input water supply quality feature set and presetting an ecological risk target; selecting different machine learning algorithms to respectively establish wastewater ecological risk prediction models for different biological treatment multi-unit combination processes based on the input water supply quality feature set; and conducting cross-validation experiments to evaluate the performance of the prediction models to obtain an optimal wastewater ecological risk prediction model; and screening the wastewater ecological risk prediction model that achieves the pre-set ecological risk target, the ecological risk target including but not limited to trace pollutant content, biological toxicity, etc.

[0012] Preferably, when the above-mentioned cross-validation experiment is performed to evaluate the performance of the prediction model, the following three evaluation indexes are used, and the calculation methods of each index are as follows:

number

[0013] Preferably, step (4) comprises: (4.1) simulating and reconstructing all possible short unit combination processes for wastewater biological treatment, and performing n-unit characterization on the reconstructed short unit combinations at once; and (4.2) predicting the effluent quality of each reconstructed biological treatment short unit combined process under different feedwater quality scenarios, and then screening the reconstructed biological treatment short unit combined processes whose effluent quality achieves the standard based on whether the predicted results achieve the preset effluent quality target.

[0014] The wastewater biological treatment process reconstruction system based on machine learning according to the present invention comprises: A data collection and pre-processing module for collecting and pre-processing basic water quality, ecological risk and process information data of the water supply and discharge in the wastewater biological treatment process; a feature extraction and representation module for performing feature extraction and multi-unit characterization for a wastewater biological treatment process; A prediction model building module for building a wastewater basic water quality prediction model and a wastewater ecological risk prediction model for different biological treatment multi-unit combination processes; A wastewater biological treatment process reconstruction and standard achievement detection module for simulating and reconstructing all possible short unit combination processes for wastewater biological treatment, and performing effluent quality standard achievement detection under different feedwater quality scenarios, and screening the reconstructed short unit combination processes for biological treatment that achieve the effluent quality standard; and a wastewater biological treatment process determination module for evaluating the ecological risk of a reconstructed short unit combination process for biological treatment that achieves a standard for wastewater quality based on the wastewater ecological risk prediction model, and determining an optimal wastewater biological treatment process based on the ecological risk control effect. Effect of the Invention

[0015] Beneficial Effects: Compared with the prior art, the present invention has the following significant advantages: 1. A multi-unit characterization is performed on the biological treatment process, and a reasonable and effective biological treatment process reconstruction method is further proposed, which can determine a new biological treatment process that meets the basic quality of the effluent and has the lowest biological risk for different feedwater quality scenarios, and resolve the selection constraints of the traditional biological treatment process method; 2. A model is established using knowledge and process data of the wastewater treatment process, and the effluent quality and biological risk index of the biological treatment process of the sewage treatment plant can be accurately predicted for different feedwater quality scenarios. [Brief description of the drawings]

[0016] [Figure 1] 1 is a method flow chart of the present invention. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0017] The technical solution of the present invention will be further described below in conjunction with the drawings.

[0018] As shown in FIG. 1 , the method for reconstructing a wastewater biological treatment process based on machine learning according to the present invention includes the following steps:

[0019] (1) Collect basic water quality, ecological risk and process information data of wastewater biological treatment process of sewage treatment plant and perform data pre-processing.

[0020] Basic water quality, ecological risk and biological treatment process information data of sewage treatment plant water supply and wastewater are collected respectively. Basic water quality characteristics of water supply and wastewater include chemical oxygen demand, ammonia nitrogen, total nitrogen, total phosphorus, suspended solids, biochemical oxygen demand, animal and vegetable oils, petroleum, nitrate nitrogen, total organic carbon, dissolved oxygen, temperature, pH value and fecal coliform count. Water supply and wastewater ecological risk data include trace pollutant content and biological toxicity. Information data of wastewater biological treatment process using activated sludge method used in sewage treatment plant is obtained.

[0021] Data pre-processing is carried out on the basic water quality and ecological risk data of the sewage treatment plant. The data pre-processing methods include abnormal data processing, default data processing, outlier data processing, repeatability and consistency detection, data normalization and standardization, etc.

[0022] (2) Feature extraction and multi-unit characterization are performed for wastewater biological treatment processes.

[0023] Feature extraction is performed on the biological pond links of the wastewater biological treatment process using the activated sludge method in a sewage treatment plant, specifically, the biological pond links are divided into three feature units, aerobic ponds, anoxic ponds, and anaerobic ponds, based on dissolved oxygen, and multi-unit feature combination is performed on the wastewater biological treatment process using the activated sludge method using a method not limited to the one-hot algorithm.

[0024] The method for performing multi-unit characterization for a biological treatment process using the one-hot algorithm is as follows: First, we represent the aerobic pond unit as [1,0,0], the anoxic pond unit as [0,1,0], the anaerobic pond unit as [0,0,1], and the blank unit as [0,0,0].

[0025] Then, n-unit characterization is performed collectively for the wastewater biological treatment process using the activated sludge method, and if the number of unitized biological pond links in the process being represented is smaller than n, blank units are used to supplement the unitized biological pond links, and the dimension of the feature represented by n units is 3 x n. For example, the "tri-unit" of the Anaerobic-Anoxic-Oxic process in the wastewater biological treatment process is represented as [0,0,1,0,1,0,1,0,0], the "double unit" of the Anaerobic-Oxic process as [0,0,1,1,0,0], and the "tri-unit" as [0,0,1,1,0,0,0,0,0].

[0026] (3) The wastewater basic water quality prediction model and wastewater ecological risk prediction model of the n-unit combination process for different biological treatments were established, and the standard achievement detection was carried out.

[0027] Specifically, the method includes the following steps:

[0028] (3.1) A basic water quality prediction model for wastewater from n-unit combination processes for different biological treatment was established, which specifically includes the following:

[0029] In step (1), a set of input feedwater quality characteristics is determined from the basic water quality characteristics of the biological treatment process that have been data preprocessed, and effluent water quality targets are preset. The effluent water quality targets include ammonia nitrogen, total nitrogen, total phosphorus, nitrate nitrogen, total organic carbon, etc. One or more effluent basic water quality prediction models can be established based on different effluent water qualities.

[0030] Select different machine learning algorithms, respectively establish effluent basic water quality prediction models for different biological treatment n-unit characteristic combination processes based on the input feedwater quality feature sets, and perform cross-validation experiments to evaluate the performance of the prediction models to obtain the optimal effluent basic water quality prediction model. Based on whether the model prediction results achieve the preset effluent quality target, then screen the effluent basic water quality prediction models whose effluent quality meets the standard.

[0031] (3.2) A wastewater ecological risk prediction model was established for different n-unit combination processes for biological treatment, specifically including:

[0032] In step (1), a set of input water quality characteristics is determined based on the basic water quality and ecological risk characteristics of the biological treatment process that have been data preprocessed, and an ecological risk target is preset.

[0033] Select different machine learning algorithms, respectively establish ecological risk prediction models for different biological treatment n-unit feature combination processes based on the input feedwater quality feature sets, and perform cross-validation experiments to evaluate the performance of the prediction models to obtain the optimal ecological risk prediction model. Then, screen the wastewater ecological risk prediction models that achieve the ecological risk criteria based on whether the model prediction results achieve the preset ecological risk targets.

[0034] When conducting cross-validation experiments to evaluate the predictive performance of the established wastewater basic water quality prediction model and ecological risk prediction model, the following three evaluation indexes are used, and the calculation methods of each index are as follows:

number

number

[0035] (4) Simulate and reconstruct a short unit combination process for wastewater biological treatment using different activated sludge processes, and detect the achievement of basic water quality standards under different supply water quality scenarios, including:

[0036] (4.1) All possible short unit combination processes for wastewater biological treatment using activated sludge are simulated and reconstructed, and n-unit characterization is performed collectively for the short unit combinations, which are used as input variables for the biological process information of the prediction model. The short unit combinations include double unit combinations, tri-unit combinations, tetra-unit combinations, etc.

[0037] (4.2) Under different supply water quality scenarios, use the optimal effluent basic water quality prediction model of the n-unit characteristic combination process for biological treatment trained in step (3.1) to respectively predict the effluent quality of the short unit combination process for biological treatment reconstructed in step (4.1), and judge whether the prediction result achieves the predetermined effluent quality target, the input of the effluent basic water quality prediction model is the input characteristics determined in step (3.1) and the short unit combination process for wastewater biological treatment reconstructed by simulation in step (4.1), the output of the model is the prediction result of effluent quality, and then screen the reconstructed short unit combination process for biological treatment whose prediction result meets the effluent quality target as the reconstructed short unit combination process for biological treatment whose effluent quality achieves the standard.

[0038] (5) Based on the wastewater ecological risk prediction model, evaluate the ecological risk of the reconstructed short unit combination process for biological treatment that achieves the wastewater quality standard, and determine the wastewater biological treatment process based on the ecological risk control effect.

[0039] Using the wastewater ecological risk prediction model of the n-unit feature combination process for biological treatment trained in step (3.2), predict the ecological risk of the short unit combination process for reconstructed biological treatment that achieves the standard in step (4.2), respectively; the input of the ecological risk prediction model is the input features determined in step (3.2) and the short unit combination process for reconstructed biological treatment that achieves the standard in effluent quality screened in step (4.2); the output of the model is the predicted result of ecological risk; then, rescreen the short unit combination process for reconstructed biological treatment whose predicted result meets the wastewater ecological risk target, and determine the optimal wastewater biological treatment process based on the ecological risk control effect.

[0040] In order to further illustrate the method of the present invention, the wastewater biological treatment process data of sewage treatment plants with different wastewater biological treatment processes across the country are collected. The specific wastewater biological treatment process reconstruction steps are as follows:

[0041] (1) Collect and preprocess basic water quality, ecological risks, and specific activated sludge wastewater biological treatment process information data from 180 sewage treatment plants in many provinces and cities, including Shanghai, Jiangsu, Guangdong, Henan, and Yunnan.

[0042] The basic water quality characteristics of water supply and wastewater include ammonia nitrogen, total nitrogen, total phosphorus, nitrate nitrogen, total organic carbon, temperature, and pH value, and the water supply and wastewater ecological risk data includes biological toxicity.

[0043] Data pre-processing is carried out on the basic water quality and ecological risk data of the sewage treatment plant. The data pre-processing methods include abnormal data processing, default data processing, outlier data processing, repeatability and consistency detection, data normalization and standardization, etc.

[0044] (2) Feature extraction and short multi-unit characterization are performed for wastewater biological treatment processes.

[0045] Each link in the wastewater biological treatment process using the activated sludge method at a sewage treatment plant is unitized, and the biological pond unit is divided into three feature units, namely aerobic ponds, anoxic ponds and anaerobic ponds, based on dissolved oxygen, feature extraction is performed, and a short multi-unit feature combination is performed for the activated sludge biological treatment process using a method such as but not limited to the one-hot algorithm.

[0046] The method for performing multi-unit characterization for a biological treatment process using the one-hot algorithm is as follows: First, we represent the aerobic pond unit as [1,0,0], the anoxic pond unit as [0,1,0], the anaerobic pond unit as [0,0,1], and the blank unit as [0,0,0].

[0047] Then, a "tetraunit" characterization is performed collectively for all biological treatment processes using the activated sludge method, and blank units are used to supplement the number of units in the biological pond of the process being represented if the number of units is less than 4; the dimension of the feature represented by the "tetraunit" is 12. The "tetraunit" of the anaerobic-anoxic-aerobic process in the wastewater biological treatment process is represented as [0,0,1,0,1,0,1,0,0,0,0,0], and the "tetraunit" of the anaerobic-aerobic process is represented as [0,0,1,1,0,0,0,0,0,0,0,0,0].

[0048] (3) Establish a basic water quality prediction model and an ecological risk prediction model for wastewater from biological treatment process, and detect the achievement of the standards.

[0049] Specifically, the method includes the following steps:

[0050] (3.1) Establish a basic water quality prediction model for biological treatment wastewater and detect whether it meets the standards. Specifically, the following items are included:

[0051] Determine the input feedwater quality characteristic set and preset the effluent quality target. The feedwater quality characteristic set includes total phosphorus, total nitrogen, ammonia nitrogen, nitrate nitrogen, total organic carbon, temperature and pH value, and the effluent quality target is set to a total effluent nitrogen removal rate of 60%.

[0052] Select support vector machine and random forest algorithm to respectively establish wastewater basic water quality prediction models for different biological treatment short tetra unit characteristic combination processes based on the input feedwater quality feature set, and carry out cross-validation experiments to evaluate the performance of the prediction models to obtain the optimal wastewater basic water quality prediction model. Based on whether the model prediction result reaches the preset total wastewater nitrogen removal rate of 60%, then screen the wastewater basic water quality prediction models whose wastewater quality meets the standard.

[0053] (3.2) Establish wastewater ecological risk prediction model for biological treatment process and detect standard achievement, specifically including:

[0054] Determine the input feedwater quality characteristic set and preset the ecological risk target. The feedwater quality characteristic set includes total phosphorus, total nitrogen, ammonia nitrogen, nitrate nitrogen, total organic carbon, temperature, and pH value, and the ecological risk target is set to a biotoxicity reduction rate of 60%.

[0055] Select support vector machine and random forest algorithm to respectively establish wastewater ecological risk prediction models for different short tetra-unit characteristic combination processes for biological treatment based on the input water supply quality feature set, and carry out cross-validation experiments to evaluate the performance of the prediction models to obtain the optimal wastewater ecological risk prediction model. Based on whether the model prediction result reaches the preset biological toxicity reduction rate of 60%, then screen the wastewater ecological risk prediction models that achieve the wastewater ecological risk standard.

[0056] When the predictive performance of the established wastewater basic water quality prediction model and ecological risk prediction model was evaluated through cross-validation experiments, the MAE, MSE and R 2 The results are shown in Table 1 below. As can be seen from Table 1, the support vector machine method has a smaller evaluation index value for the wastewater basic water quality prediction model, and the random forest method has a smaller evaluation index value for the wastewater ecological risk prediction model. Therefore, the support vector machine algorithm is selected to build the optimal wastewater basic water quality prediction model, and the random forest algorithm is selected to build the optimal ecological risk prediction model.

[0057] [Table 1]

[0058] (4) Simulate and reconstruct a short unit combination process for wastewater biological treatment using different activated sludge methods, and perform wastewater quality standard achievement detection based on the optimal wastewater basic water quality prediction model constructed in step (3). The specific steps are as follows: (4.1) Simulate and reconstruct all possible double-unit, tri-unit, and tetra-unit combination processes for wastewater biological treatment using activated sludge, and perform a one-hot one-hot algorithm to characterize the short unit combination processes collectively as "tetra-units"; (4.2) Under different feedwater quality scenarios, for example, total phosphorus is 2 mg / L, total nitrogen is 15 mg / L, ammonia nitrogen is 13.5 mg / L, nitrate nitrogen is 1 mg / L, total organic carbon is 30 mg / L, temperature is 25°C, and pH value is 6.5, the optimal effluent basic water quality prediction model of the "Tetra Unit" characteristic combination process trained in step (3.1) is used to respectively predict the effluent quality of different reconstructed wastewater biological treatment "Tetra Unit" characteristic combination processes, the prediction results are shown in Table 2 below, and it is judged whether the prediction results reach the preset effluent quality target (total nitrogen removal rate 60%), i.e., whether the effluent total nitrogen is less than 6 mg / L, and then screen the short unit combination process for reconstructed biological treatment whose effluent quality achieves the standard.

[0059] (5) Predict the ecological risk of each of the short unit combination processes for reconstructed biological treatment that achieve the wastewater quality standard, and determine the optimal wastewater biological treatment process from the reconstructed biological treatment processes that achieve the standard based on the ecological risk control effect. The specific steps are as follows: For the reconstructed biological treatment short unit combination process in which the wastewater quality meets the standard in step (4.2), the wastewater ecological risk model of the "tetra-unit" characteristic combination process trained in step (3.2) is used to predict the ecological risk, and the ecological risk is compared with the preset ecological risk target (biological toxicity reduction rate 60%), as shown in Table 2. The ecological risk control effects of different reconstructed biological treatment short unit combination processes are evaluated, and then the optimal wastewater biological treatment short unit combination process is selected.

[0060] [Table 2]

[0061] As can be seen from Table 2, under the current water quality scenario, the predicted results of the basic water quality and ecological risk of the effluent of the above-mentioned reconstructed short unit combination process for wastewater biological treatment both meet the preset targets, and based on the ecological risk control effect, the anaerobic-aerobic-anoxic-aerobic combination process, which is the short unit combination process for wastewater biological treatment with the highest biological toxicity reduction rate, is selected, and its biological toxicity reduction rate reaches 67.3%.

Claims

1. A method for reconstructing a wastewater biological treatment process based on machine learning, comprising the steps of: (1) collecting basic water quality, ecological risk and process information data of water supply and discharge in a wastewater biological treatment process and performing data preprocessing; (2) performing feature extraction and multi-unit characterization for a wastewater biological treatment process; (3) constructing a wastewater basic water quality prediction model and a wastewater ecological risk prediction model for different biological treatment multi-unit combination processes and carrying out standard achievement detection; (4) simulating and reconstructing the short unit combination process for different wastewater biological treatment, and performing effluent quality standard achievement detection under different feedwater quality scenarios; and (5) evaluating the ecological risk of a short unit combination process for reconstructing biological treatment that achieves a standard for wastewater quality, and determining an optimal wastewater biological treatment process.

2. The method for reconstructing a wastewater biological treatment process based on machine learning as described in claim 1, characterized in that in step (1), the basic water quality of the water supply and wastewater of the wastewater biological treatment process includes chemical oxygen demand, ammonia nitrogen, total nitrogen, total phosphorus, suspended solids, biochemical oxygen demand, animal and vegetable oils, petroleum, nitrate nitrogen, total organic carbon, dissolved oxygen, temperature, pH value, and fecal coliform count, the water supply and wastewater ecological risk data includes trace pollutant content and biological toxicity, etc., and the process information of the wastewater biological treatment process includes information data of a wastewater biological treatment process using an activated sludge method.

3. Step (2) is The method for reconstructing a wastewater biological treatment process based on machine learning according to claim 1, further comprising: performing feature extraction on the biological pond links of the activated sludge wastewater biological treatment process, dividing the biological pond links into three feature units, i.e., aerobic ponds, anoxic ponds and anaerobic ponds, based on dissolved oxygen; and performing multi-unit characterization on the activated sludge wastewater biological treatment process according to the extracted feature units.

4. The method for reconstructing a wastewater biological treatment process based on machine learning according to claim 3, wherein the multi-unit characteristic method includes a one-hot algorithm.

5. Performing multi-unit characterization of biological treatment processes using the one-hot algorithm: First, the aerobic pond unit is represented as [1,0,0], the anoxic pond unit as [0,1,0], the anaerobic pond unit as [0,0,1], and the blank unit as [0,0,0]. The method for reconstructing a wastewater biological treatment process based on machine learning as described in claim 4, further comprising: performing feature splicing for the biological pond unit, performing n-unit characterization collectively for the wastewater biological treatment process using the activated sludge method, and supplementing with blank units if the number of units of the represented process is insufficient.

6. In step (3), constructing a effluent basic water quality prediction model of the above-mentioned different biological treatment multi-unit combined processes includes: The method for reconstructing a wastewater biological treatment process based on machine learning as described in claim 5, characterized in that it includes: determining an input water supply quality feature set and presetting a wastewater quality target; selecting different machine learning algorithms, and respectively establishing wastewater basic water quality prediction models for different biological treatment multi-unit combination processes based on the input water supply quality feature set, and conducting cross-validation experiments to evaluate the performance of the prediction models to obtain an optimal wastewater basic water quality prediction model; and screening a wastewater ecological risk prediction model that achieves the pre-set wastewater quality target.

7. In step (3), constructing a wastewater ecological risk prediction model of the different biological treatment multi-unit combined processes includes: The method for reconstructing a wastewater biological treatment process based on machine learning as described in claim 6, characterized in that it includes: determining an input water supply quality feature set and pre-setting an ecological risk target; selecting different machine learning algorithms, and respectively establishing wastewater ecological risk prediction models for different biological treatment multi-unit combination processes based on the input water supply quality feature set, and conducting cross-validation experiments to evaluate the performance of the predictive models to obtain the optimal wastewater ecological risk prediction model; and screening the wastewater ecological risk prediction model that achieves the pre-set ecological risk target.

8. When the cross-validation experiment is performed to evaluate the performance of the prediction model, the following three evaluation indexes are used, and the calculation methods for each index are as follows: [0045] MAE is the mean absolute error between the predicted value and the true value, MSE is the mean squared error between the predicted value and the true value, R ２ The method for reconstructing a wastewater biological treatment process based on machine learning according to claim 6 or 7, characterized in that: represents the overall performance of the predictive model.

9. Step (4) is (4.1) simulating and reconstructing all possible short unit combination processes for wastewater biological treatment, and performing n-unit characterization on the reconstructed short unit combinations at once; The method for reconstructing a wastewater biological treatment process based on machine learning according to claim 7, further comprising the step (4.2) of predicting the effluent quality of each of the reconstructed short unit combination processes for biological treatment under different feedwater quality scenarios, and then screening the reconstructed short unit combination processes for biological treatment whose effluent quality achieves the standard based on whether the prediction results achieve the preset effluent quality target.

10. A wastewater biological treatment process reconstruction system based on machine learning, A data collection and pre-processing module for collecting and pre-processing basic water quality, ecological risk and process information data of water supply and discharge in the wastewater biological treatment process; a feature extraction and representation module for performing feature extraction and multi-unit characterization for a wastewater biological treatment process; A prediction model building module for building a wastewater basic water quality prediction model and a wastewater ecological risk prediction model for different biological treatment multi-unit combination processes; A wastewater biological treatment process reconstruction and standard achievement detection module for simulating and reconstructing all possible short unit combination processes for wastewater biological treatment, and performing effluent quality standard achievement detection under different feedwater quality scenarios, and screening the reconstructed short unit combination processes for biological treatment that achieve the effluent quality standard; A wastewater biological treatment process reconstruction system based on machine learning, comprising: a wastewater biological treatment process determination module for evaluating the ecological risk of a short unit combination process for reconstructed biological treatment that achieves a standard for wastewater quality based on a wastewater ecological risk prediction model, and determining an optimal wastewater biological treatment process based on the ecological risk control effect.

Images