Underground sewage treatment plant effluent quality prediction method based on deep learning

Abstract

The invention discloses an underground sewage treatment plant effluent quality prediction method based on deep learning. The prediction method comprises the following steps: collecting inflow and outflow data of an underground sewage treatment plant; preprocessing the water inlet and outlet data based on an isolated forest algorithm to obtain sample data; predicting the effluent quality corresponding to the sample data through a preset prediction model; evaluating the prediction performance of the preset prediction model based on a preset index to determine an optimal prediction model; and predicting the effluent quality of the sewage treatment plant based on the optimal prediction model. According to the invention, the data of the buried sewage treatment plant is intelligently preprocessed, the problems of data abnormity and data deficiency are solved, and the water quality key indexes are subjected to long-time sequence prediction through the deep learning model, so that the precision and reliability of water quality monitoring and prediction of the buried sewage treatment plant are improved.

Description

Technical Field

[0001] The present application relates to the field of intelligent prediction of sewage treatment water quality parameters, and in particular to a method for predicting effluent quality of underground sewage treatment plants based on deep learning. Background Art

[0002] With the acceleration of urbanization and the improvement of industrialization, the amount of sewage discharged in economically active regions such as South China has continued to increase, bringing great pressure to the urban water environment. Traditional sewage treatment methods have been unable to meet the increasingly stringent environmental protection requirements. Therefore, more efficient and intelligent sewage treatment technologies and methods are needed. As a sewage treatment method with small footprint, low noise and small environmental impact, underground sewage treatment plants have been widely used in South China. However, due to the particularity of the operating environment of underground sewage treatment plants, the prediction and control of their effluent water quality face many challenges. Among them, total nitrogen, total phosphorus and ammonia nitrogen are important pollutant indicators in sewage treatment effluent, which are closely related to eutrophication of water bodies, ecological environmental damage and public health risks. However, the removal of these indicators in sewage treatment is affected by many factors, including influent water quality, treatment process, ambient temperature, etc., and it is difficult to predict. According to known operating conditions, the effluent total nitrogen (TNout) of underground sewage treatment plants may often exceed the standard risk, and the removal of effluent total phosphorus (TPout) mainly occurs in the biochemical treatment unit and the deep treatment unit. Chemical phosphorus removal mainly relies on adding phosphorus removal agents in high-efficiency sedimentation tanks, and the consumption of agents has a greater impact on production costs. The prediction of the above effluent water quality parameters can not only improve the refined management capabilities of sewage treatment plants, but also promote the advancement of environmental protection technology and industrial upgrading, and has significant application prospects and social significance.

[0003] Traditional water quality prediction methods mainly rely on physical models, empirical formulas and other means, but these methods have limitations when dealing with complex and changeable water quality data, especially when facing sewage treatment processes with nonlinear changes, the prediction accuracy is limited and the prediction error is large. In recent years, the introduction of machine learning and deep learning methods has provided new ideas for the prediction of water quality variables. However, these methods still face challenges in processing long-term series data and capturing long-distance dependencies. In particular, for data with time series characteristics generated during the operation of underground sewage treatment plants, existing methods still have shortcomings in long-term series prediction.

[0004] The shortcomings of the prior art are mainly reflected in the following aspects:

[0005] 1. Low prediction accuracy: Existing physical models perform poorly when dealing with complex and dynamic water quality data, especially when faced with sewage treatment processes with nonlinear changes, the prediction errors are large.

[0006] 2. Unable to process long time series data: The operating data of underground sewage treatment plants has time dependence and seasonal changes. Existing models find it difficult to capture these long-term dependencies, resulting in inaccurate predictions of future water quality changes.

[0007] 3. Insufficient processing of abnormal data: The monitoring data of underground sewage treatment plants are often affected by factors such as equipment failure and sensor failure, resulting in a large number of abnormal values ​​and missing values ​​in the data. In the existing technology, simple interpolation or median filling methods are usually used, which fail to effectively process abnormal data and affect the overall performance of the prediction model. Summary of the invention

[0008] The purpose of this application is to provide a method for predicting the effluent water quality of underground sewage treatment plants based on deep learning. By intelligently preprocessing the sewage treatment plant data, the problems of data anomalies and missing data are solved, and long-term series prediction of key water quality indicators is performed through a deep learning model to improve the accuracy and reliability of water quality monitoring and prediction in sewage treatment plants.

[0009] To achieve the above objectives, this application provides the following solutions:

[0010] In a first aspect, the present application provides a method for predicting effluent quality of an underground sewage treatment plant based on deep learning, comprising:

[0011] Collect influent and effluent data from underground wastewater treatment plants;

[0012] Preprocessing the water inlet and outlet data based on an isolation forest algorithm;

[0013] Selecting data with characteristics that affect water quality prediction based on the preprocessed data for processing to obtain sample data;

[0014] Predicting the effluent water quality corresponding to the sample data by using a preset prediction model;

[0015] Evaluating the prediction performance of the preset prediction model based on preset indicators to determine the optimal prediction model;

[0016] The effluent water quality of the underground sewage treatment plant is predicted based on the optimal prediction model.

[0017] Optionally, the inlet and outlet data include at least: inlet and outlet flow, sludge concentration, pH value, chemical oxygen demand, biochemical oxygen demand, total nitrogen, and total phosphorus.

[0018] Optionally, the preprocessing includes:

[0019] Identifying abnormal data and missing data in the inlet and outlet data;

[0020] Using an interpolation method of adjacent time series data to repair the abnormal data;

[0021] The missing data are filled by using the moving average method of adjacent time series.

[0022] Optionally, before the step of obtaining sample data, the step further includes:

[0023] performing correlation analysis on the preprocessed data to determine features that affect water quality prediction;

[0024] Obtaining data information corresponding to the feature;

[0025] The data information is normalized to obtain sample data.

[0026] Optionally, the characteristics include at least: water quality monitoring data, environmental data and treatment process parameters.

[0027] Optionally, the preset prediction model includes:

[0028] Prediction model based on random forest;

[0029] Prediction model based on back-propagation neural network;

[0030] Prediction model based on long short-term memory network;

[0031] Transformer-based prediction model;

[0032] Informer-based prediction model.

[0033] Optionally, predicting the effluent water quality corresponding to the sample data by using a preset prediction model includes:

[0034] Acquire a training data set from the sample data to train the preset prediction model;

[0035] The prediction data set obtained from the sample data is predicted based on the trained preset prediction model.

[0036] Optionally, evaluating the prediction performance of the preset prediction model based on preset indicators includes:

[0037] Evaluating the prediction performance of the prediction model based on the mean square error of the prediction results; and / or,

[0038] Evaluating the prediction performance of the prediction model based on the root mean square error of the prediction results; and / or,

[0039] Evaluating the prediction performance of the prediction model based on the mean absolute error of the prediction results; and / or,

[0040] The prediction performance of the prediction model is evaluated based on the determination coefficient of the prediction results.

[0041] Optionally, evaluating the prediction performance of the preset prediction model based on preset indicators further includes:

[0042] Based on the cross-validation method, the preset prediction model is used to predict different prediction data sets;

[0043] The prediction performance of the preset prediction model is evaluated based on the prediction results.

[0044] In a second aspect, the present application provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of any of the above-mentioned methods for predicting effluent water quality of underground sewage treatment plants based on deep learning.

[0045] According to the specific embodiments provided in this application, this application discloses the following technical effects:

[0046] The present application provides a method for predicting the effluent water quality of underground sewage treatment plants based on deep learning. It solves the problems of data anomalies and missing data by intelligently preprocessing the sewage treatment plant data, and performs long-term series prediction of key water quality indicators through a deep learning model, so as to improve the accuracy and reliability of water quality monitoring and prediction of underground sewage treatment plants. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0048] Figure 1 A schematic diagram of a process flow of a method for predicting effluent quality of an underground sewage treatment plant based on deep learning provided in an embodiment of the present application;

[0049] Figure 2 A schematic diagram of a deep learning-based effluent quality prediction process framework for an underground sewage treatment plant provided in an embodiment of the present application;

[0050] Figure 3 Schematic diagram of the isolation forest algorithm provided in the embodiment of the present application;

[0051] Figures 4A-4B A schematic diagram showing the comparison of data before and after isolation forest processing provided in an embodiment of the present application;

[0052] Figure 5A feature correlation heat map provided for an embodiment of the present application;

[0053] Figure 6 A schematic diagram of the structure of a prediction model based on a long short-term memory network LSTM provided in an embodiment of the present application;

[0054] Figure 7 A schematic diagram of the structure of the Informer-based prediction model provided in the embodiment of the present application;

[0055] Figure 8 The SHAP value of the effluent total nitrogen TN provided in the embodiment of the present application;

[0056] Fig. 9 The SHAP value of the effluent total phosphorus TP provided in the embodiment of the present application;

[0057] Fig.10 A schematic diagram showing the comparison between the actual value and the predicted value of the effluent total nitrogen TN provided in the embodiment of the present application;

[0058] Fig.11 A schematic diagram showing the comparison between the actual value and the predicted value of the effluent total phosphorus TP provided in the embodiment of the present application;

[0059] Fig.12 A schematic diagram of the structure of a computer device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0060] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0061] In order to make the above-mentioned objects, features and advantages of the present application more obvious and easy to understand, the present application is further described in detail below with reference to the accompanying drawings and specific implementation methods.

[0062] refer to Figure 1 and Figure 2 As shown, the present application provides a method for predicting effluent quality of underground sewage treatment plants based on deep learning, including:

[0063] Collect influent and effluent data from wastewater treatment plants;

[0064] In an embodiment of the present application, when predicting the effluent quality of a sewage treatment plant, it is first necessary to collect various parameters of the inlet and outlet of the underground sewage treatment plant, including: inlet and outlet flow, sludge concentration, pH value, COD, total nitrogen, total phosphorus, chemical oxygen demand (COD), biochemical oxygen demand (BOD) and other key data information. In this embodiment, in order to ensure that the latest inlet and outlet data information can be obtained in a timely manner, the data collection frequency can be set to once a day, and the various data indicators of the inlet and outlet can be recorded in detail. In another embodiment scenario, in order to be able to grasp the data indicators of the inlet and outlet at any time, the various data indicators of the inlet and outlet can also be collected in real time and recorded in detail.

[0065] Preprocessing the inlet and outlet data based on an isolation forest algorithm to obtain sample data;

[0066] In the embodiment of the present application, after collecting the inlet and outlet data of the sewage treatment plant, the collected inlet and outlet data are further preprocessed using the isolation forest algorithm. Specifically, the isolation forest algorithm is used to process the abnormal values ​​and vacancy values ​​in the data. Figure 3 As shown in the figure, the isolation forest algorithm is an unsupervised abnormal data detection method that can detect abnormal data in the data by constructing random trees and automatically mark the detected abnormal data, including extreme values ​​caused by sensor failure or other reasons. For these abnormal data, further data repair or elimination is performed through interpolation based on adjacent time series data. For missing data, the isolation forest algorithm uses the moving average method of adjacent time series to fill in the data missing area according to the time series characteristics of the missing data and environmental related factors, thereby ensuring data continuity and consistency, which is complete and reliable. Reference Figure 4A , 4B As shown, the comparison of data before and after processing by the isolation forest algorithm.

[0067] In the embodiment of the present application, based on the processed data, data correlation analysis is performed to select the most influential feature for predicting water quality, and the corresponding data information obtained based on the feature is normalized to obtain sample data. The specific steps are as follows:

[0068] performing correlation analysis on the preprocessed data to determine features that affect water quality prediction;

[0069] Obtaining data information corresponding to the feature;

[0070] The data information is normalized to obtain sample data.

[0071] In an embodiment of the present application, after obtaining the treated influent and effluent data, the relationship between the variables in the treated influent and effluent data is further analyzed, the feature data with cross correlations are eliminated, and the feature data with the most influence on the predicted water quality is selected. In this embodiment, the selected feature data include: historical water quality monitoring related data, historical environmental data, and treatment process parameters. In an exemplary scenario, after obtaining the pre-treated influent and effluent data, the Pearson correlation coefficient is used to analyze the relationship between the variables in the influent and effluent data, and the features with poor correlation are eliminated. Finally, the influent and effluent flow, sludge concentration, pH value, COD, BOD, total nitrogen, total phosphorus, etc. are selected as the main features of the prediction model input.

[0072] In the embodiment of the present application, after the characteristic data for water quality prediction is selected, in order to ensure the stability and efficiency of the prediction model training process, the selected characteristic data is further processed to eliminate the dimensional differences of different variables. In an exemplary embodiment, the Min-Max normalization method is used to normalize all the selected characteristic data, and the sample values ​​of all characteristic data are mapped and scaled to the range of [0,1], so as to eliminate the influence of the inconsistent dimensions of different indicators on the model. Figure 5 As shown. The calculation expression of normalization processing is as follows:

[0073]

[0074] Among them, x is the sample value; minValue is the minimum value in the sample; maxValue is the maximum value in the sample.

[0075] Predicting the effluent water quality corresponding to the sample data by using a preset prediction model;

[0076] In the embodiments of the present application, multiple preset regression test models, such as a prediction model based on random forest (ET), a prediction model based on back propagation neural network (BPNN), a prediction model based on long short-term memory network (LSTM), a prediction model based on Transformer, and a prediction model based on Informer, are used to perform training and prediction based on the acquired sample data. In order to ensure that the prediction performance of the prediction model is optimal, each prediction model is hyperparameter optimized. In the embodiments of the present application, reference Figure 6 As shown in Figure 2, a prediction model based on the long short-term memory network (LSTM) is used to handle long-term dependencies. Figure 7As shown, the Informer-based prediction model is used to solve the sparse attention mechanism problem in long time series prediction, thereby improving the efficiency and accuracy of the model. In one embodiment of the present application, for the training of the prediction model, the adam optimizer is used to set the appropriate learning rate, batch size and number of training times according to the characteristics of the sample data set.

[0077] In a scenario of an embodiment, by sampling the inlet and outlet data of an underground sewage treatment plant, a total of 2191 sets of complete sample data were obtained. When the above five prediction models were trained and predicted using the original data, the training data set and the test data set were divided in a ratio of 8:2, and 1752 sets of training data and 439 sets of test data were obtained. The Adam optimizer was used in the training process, and the learning rate was set to 0.0005. The number of training times was 300 to ensure the optimal performance of the model.

[0078] In this embodiment, after the prediction model is trained, the model is further explained using SHAP analysis. The SHAP value of each feature is calculated to identify the feature that has the greatest impact on the effluent water quality.

[0079] Among them, the prediction of effluent total nitrogen TNout, chemical oxygen demand COD and suspended solids SS are the most important characteristics. The increase of chemical oxygen demand COD will increase the predicted value of total nitrogen TN, while suspended solids SS may have a negative impact. Figure 8 SHAP values ​​of effluent total nitrogen (TN) are shown.

[0080] Among them, the chemical oxygen demand (COD) is also the most influential feature for the prediction of effluent total phosphorus (TPout), which is more critical to the removal effect of total phosphorus (TP). Fig. 9 SHAP values ​​of effluent total phosphorus TP are shown.

[0081] Evaluating the prediction performance of the preset prediction model based on preset indicators to determine the optimal prediction model;

[0082] In the embodiment of the present application, after the preset prediction model is used to predict the test data set to obtain the prediction results, the prediction results are comprehensively evaluated to determine the prediction performance of each prediction model. In this embodiment, the mean square error (MSE), root mean square error (RMSE), mean absolute error (MAE) and determination coefficient (R 2 ) and other indicators to evaluate the prediction performance of the prediction model. In this embodiment, in order to further improve the accuracy of the model, the present invention adopts a cross-validation method to evaluate the performance of the model under different data set divisions to ensure the stability and generalization ability of the model. At the same time, through hyperparameter tuning, such as adjusting the learning rate, the number of hidden layer neurons, etc., the model performance is further optimized.

[0083]

[0084]

[0085]

[0086]

[0087] Among the above evaluation indicators, MSE, MAE and RMSE all reflect the error between the predicted value and the true value. The smaller the error, the better the model prediction effect. R2 represents the degree of fit of the model to the data, r represents the linear relationship between the true value and the predicted value, and the closer the value is to 1, the better the model fitting effect.

[0088] In an exemplary embodiment, after each prediction module makes a prediction based on the above 439 groups of test data, its prediction result is evaluated according to the above indicators. Table 1 shows the model prediction comparison of effluent total nitrogen (TN).

[0089] Model MSE MAE RMSE <![CDATA[R 2 ]]> ET 4.3872 1.4211 2.1374 0.562 BPNN 0.48788 0.5482 0.7133 0.4773 LSTM 0.0084 0.0707 0.0918 0.6458 Transformer 0.87 0.694 0.937 0.4475 Informer 0.672 0.62 0.8201 0.683

[0090] Table 1

[0091] Table 2 shows the model prediction comparison of effluent total phosphorus (TP).

[0092] Model MSE MAE RMSE <![CDATA[R 2 ]]> ET 0.0016 0.0309 0.0384 0.2483 BPNN 0.50889 0.7291 0.50889 0.4773 LSTM 0.025 0.1258 0.1582 0.7153 Transformer 0.5071 0.551 0.712 0.6605 Informer 0.531 0.5562 0.728 0.793

[0093] Table 2

[0094] According to the analysis of the prediction data in Tables 1 and 2 above, the prediction model based on the long short-term memory network (LSTM) shows the best prediction performance in various indicators, including the mean square error (MSE) of the prediction results, the average error (MAE) of the prediction results, and the determination coefficient (R 2 ) are better than other models, indicating its superiority in water quality prediction tasks. In the effluent total nitrogen (TN) task, the mean square error (MSE) of the prediction results based on the long short-term memory network (LSTM) prediction model is 0.0084, and the determination coefficient (R 2 ) is 0.6458. In the effluent total phosphorus (TP) task, the mean square error (MSE) of the prediction results based on the long short-term memory network (LSTM) prediction model is 0.025, and the determination coefficient (R 2 ) is 0.7153, both showing very high fit and low error.

[0095] The effluent quality of the sewage treatment plant is predicted based on the optimal prediction model.

[0096] In an embodiment of the present application, the optimal prediction model after training and evaluation is deployed to the online monitoring system of the sewage treatment plant for real-time prediction of effluent water quality. The optimal prediction model can output the effluent water quality in the future in real time based on the monitored inlet parameters, so that the operator can adjust the sewage treatment process parameters in time to ensure that the effluent water quality meets the standards. In this embodiment, after the above-mentioned training and evaluation of the preset prediction model, the prediction model based on the long short-term memory network LSTM is the optimal prediction model, and the optimized prediction model based on the long short-term memory network LSTM is deployed to the online monitoring system of the underground sewage treatment plant to realize the prediction of future effluent water quality.

[0097] In this embodiment, the prediction results are further compared with the actual measured values, and the error is within an acceptable range, indicating that the model has high practicality. Fig.10 and Fig.11 shown.

[0098] In an exemplary embodiment, a computer device is provided. The computer device may be a server or a terminal. The internal structure diagram thereof may be as follows: Fig.12 As shown. The computer device includes a processor, a memory, an input / output interface (Input / Output, referred to as I / O) and a communication interface. The processor, the memory and the input / output interface are connected through a system bus, and the communication interface is connected to the system bus through the input / output interface. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store video tag processing data. The input / output interface of the computer device is used to exchange information between the processor and an external device. The communication interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, a video tag processing method is implemented.

[0099] Those skilled in the art will understand that Fig.12 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0100] In an exemplary embodiment, a computer device is further provided, including a memory and a processor, wherein a computer program is stored in the memory, and the processor implements the steps in the above-mentioned method embodiments when executing the computer program.

[0101] In an exemplary embodiment, a computer-readable storage medium is provided, storing a computer program, and when the computer program is executed by a processor, the steps in the above method embodiments are implemented.

[0102] In an exemplary embodiment, a computer program product is provided, including a computer program, and when the computer program is executed by a processor, the steps in the above method embodiments are implemented.

[0103] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with relevant regulations.

[0104] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to the memory, database or other medium used in the embodiments provided in the present application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetoresistive random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM may be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM).

[0105] The database involved in each embodiment provided in this application may include at least one of a relational database and a non-relational database. The non-relational database may include a distributed database based on blockchain, etc., but is not limited thereto. The processor involved in each embodiment provided in this application may be a general-purpose processor, a central processing unit, a graphics processor, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, etc., but is not limited thereto.

[0106] In summary, the water quality prediction method and storage medium for a sewage treatment plant provided by this application have the following advantages compared with traditional technologies:

[0107] 1. Higher prediction accuracy. Compared with traditional physical models and empirical formulas, this application uses deep learning methods to better process complex water quality data and significantly improve the accuracy of predictions. By using a comprehensive evaluation of multiple regression prediction models, it ensures that the most suitable model is selected for different scenarios, greatly improving the reliability of water quality predictions.

[0108] 2. Prediction capability of long time series: By introducing prediction models based on long short-term memory networks and informers, this application can effectively capture the long-term dependencies of water quality data, which is particularly suitable for scenarios such as underground sewage treatment plants that require long-term monitoring and control. This allows the present invention to maintain a high prediction accuracy when facing complex water quality fluctuations.

[0109] 3. Data processing automation: Through the automation of data cleaning, normalization and feature selection, the need for human intervention is reduced, and the efficiency and reliability of the prediction process are improved. Outlier processing and feature selection in the data preprocessing process further ensure data quality, thereby improving the stability of the prediction model.

[0110] 4. Strong interpretability: After model training, this application uses SHAP values ​​to perform an interpretative analysis of the model to identify which features have the greatest impact on effluent quality, thereby providing managers with actionable suggestions to help optimize the sewage treatment process.

[0111] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0112] This article uses specific examples to illustrate the principles and implementation methods of this application. The description of the above embodiments is only used to help understand the method and core ideas of this application. At the same time, for those skilled in the art, according to the ideas of this application, there will be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as limiting this application.

Claims

1. A method for predicting effluent quality of underground sewage treatment plants based on deep learning, characterized in that: The prediction method comprises: Collect influent and effluent data from underground wastewater treatment plants; Preprocessing the inlet and outlet data based on an isolation forest algorithm to obtain sample data; Predicting the effluent water quality corresponding to the sample data by using a preset prediction model; Evaluating the prediction performance of the preset prediction model based on preset indicators to determine the optimal prediction model; The effluent water quality of the underground sewage treatment plant is predicted based on the optimal prediction model.

2. The method for predicting effluent quality of underground sewage treatment plants based on deep learning according to claim 1 is characterized in that: The inlet and outlet data include at least: inlet and outlet flow, sludge concentration, pH value, chemical oxygen demand, biochemical oxygen demand, total nitrogen, and total phosphorus.

3. The method for predicting effluent quality of underground sewage treatment plants based on deep learning according to claim 1 is characterized in that: The pre-processing comprises: Identifying abnormal data and missing data in the inlet and outlet data; Using an interpolation method of adjacent time series data to repair the abnormal data; The missing data are filled by using the moving average method of adjacent time series.

4. The method for predicting effluent quality of underground sewage treatment plants based on deep learning according to claim 1 is characterized in that: Before the step of obtaining sample data, it also includes: performing correlation analysis on the preprocessed data to determine features that affect water quality prediction; Obtaining data information corresponding to the feature; The data information is normalized to obtain sample data.

5. The method for predicting effluent quality of underground sewage treatment plants based on deep learning according to claim 4 is characterized in that: The characteristics include at least: water quality monitoring data, environmental data and treatment process parameters.

6. The method for predicting effluent quality of underground sewage treatment plants based on deep learning according to any one of claims 1 to 5, characterized in that: The preset prediction model includes: Prediction model based on random forest; Prediction model based on back-propagation neural network; Prediction model based on long short-term memory network; Transformer-based prediction model; Informer-based prediction model.

7. The method for predicting effluent quality of underground sewage treatment plants based on deep learning according to claim 6 is characterized in that: The predicting of the effluent water quality corresponding to the sample data by using a preset prediction model includes: Acquire a training data set from the sample data to train the preset prediction model; The prediction data set obtained from the sample data is predicted based on the trained preset prediction model.

8. The method for predicting effluent quality of underground sewage treatment plants based on deep learning according to any one of claims 1 to 5, characterized in that: The prediction performance of the preset prediction model is evaluated based on preset indicators, including: Evaluating the prediction performance of the prediction model based on the mean square error of the prediction results; and / or, Evaluating the prediction performance of the prediction model based on the root mean square error of the prediction results; and / or, Evaluating the prediction performance of the prediction model based on the mean absolute error of the prediction results; and / or, The prediction performance of the prediction model is evaluated based on the determination coefficient of the prediction results.

9. The method for predicting effluent quality of underground sewage treatment plants based on deep learning according to claim 8, characterized in that: Evaluating the prediction performance of the preset prediction model based on preset indicators also includes: Based on the cross-validation method, the preset prediction model is used to predict different prediction data sets; The prediction performance of the preset prediction model is evaluated based on the prediction results.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method for predicting effluent quality of an underground sewage treatment plant based on deep learning described in any one of claims 1 to 9 are implemented.

Images