A rapid detection method for rainwater and sewage mixing

Abstract

The present invention relates to the technical field of rainwater-sewage mixing detection, and specifically, to a method for rapid detection of rainwater-sewage mixing. The detection points are preliminarily determined through the layout location information of the pipe network and the regional location information, and the original data set is obtained using a gas detection device; the original data is then processed through steps such as data cleaning, feature processing, and weight analysis; then it is input into the RF model based on GS optimization to determine the level of pipe mixing; finally, according to the results of the grade judgment, management and control work for different pipe grades is formulated. The advantage of the present invention is that external influencing factors and multiple gas concentrations are used as parameters to input the GS‑RF model to analyze and judge the phenomenon of rainwater-sewage mixing, thereby realizing accurate classification and control of the drainage network, improving the detection efficiency of the drainage network, and providing a basis for subsequent network management.

Description

Technical Field

[0001] The present invention relates to the technical field of rain-sewage mixing detection, and in particular to a method for rapid detection of rain-sewage mixing. Background Art

[0002] Due to the concealed nature of drainage pipes, their diversity, and the lack of standardized maintenance of pipe sections, it is inevitable that rainwater and sewage will mix, leading to the following problems: (1) Sewage mistakenly flows into rainwater pipes, causing a large amount of pollutants to be directly discharged into the water body, resulting in frequent black and smelly water phenomena in cities; (2) Rainwater mistakenly enters sewage pipes, increasing the transportation pressure of sewage pipes; (3) Rainwater mistakenly enters sewage pipes and enters sewage treatment plants, causing the sewage treatment plants to operate at low load.

[0003] Traditional rainwater and sewage mixing detection methods are difficult to meet the requirements in terms of practicality and timeliness. Currently, the commonly used rainwater and sewage mixing detection technologies include visual inspection, image inspection, electromagnetic inspection, water quality and quantity inspection, ultrasonic inspection, etc. However, these methods have many disadvantages, such as: although the visual inspection method is simple, fast and low-cost, its accuracy is low and the detection results are easily affected by external factors; the image inspection method has relatively high requirements for pipelines; the laboratory analysis of the water quality inspection method may affect the progress of the detection; the water quantity inspection method requires a lot of manpower and material resources, etc. Therefore, how to accurately and quickly detect the rainwater and sewage mixing in the drainage pipe is one of the current research hotspots in the management and maintenance of the drainage network.

[0004] Algorithm optimization is a method of solving problems through certain approaches or rules based on certain ideas and mechanisms. It can enable the algorithm to have performance and characteristics that it did not originally have, and provide a practical solution to practical problems that are difficult to handle with traditional algorithm models.

[0005] The selection of data types is an important preparation before building a prediction model. Most traditional prediction methods use water quality data to build models, which makes it difficult to quickly and timely determine the situation of rainwater and sewage mixing. In recent years, with the development of sensor technology, the collection of gas data has become more convenient and faster. The gas data type has begun to be used in the field of drainage network prediction, bringing new opportunities for rainwater and sewage mixing detection.

[0006] In view of this, the present invention provides a method for quickly detecting the mixing of rainwater and sewage. Summary of the invention

[0007] The purpose of the present invention is to provide a method for quickly detecting rainwater-sewage mixing, so as to solve the technical problem of how to quickly determine whether there is rainwater-sewage mixing in a pipe section without construction.

[0008] The specific steps are as follows:

[0009] Divide the pipeline data collection area and determine the pipe section that needs to be tested;

[0010] Collect gas concentration information as the original data set and divide it into training set and test set at a ratio of 4:1. The collected gases should include hydrogen sulfide, methane, oxygen, carbon monoxide, ammonia, nitrogen dioxide, carbon dioxide and organic volatile gases;

[0011] According to the actual detection results of closed-circuit television (CCTV), the drainage pipes (pipe sections) are divided into three levels of mixing: no mixing (Ⅰ), suspected mixing (Ⅱ) and confirmed mixing (Ⅲ);

[0012] Pearson correlation analysis was used to determine the factors affecting the correlation between the original data set and the hybrid level;

[0013] Then, the F statistic formula is used to evaluate the linear relationship between each feature and the target variable, and several factors with higher influencing factors are screened out. The calculation formula is as follows:

[0014] ;

[0015] Among them, TSS represents the total sum of squares, RSS represents the residual sum of squares, p represents the number of independent variables, and n represents the sample size;

[0016] Based on the training set data and CCTV detection results, the optimal parameter combination of the RF model found by the GS algorithm is used to establish the analysis and judgment model GS-RF;

[0017] Input the data into the established GS-RF model, output the judgment result of rainwater and sewage mixing in the test set, and compare it with the real value to determine the practicality of the model;

[0018] By establishing a good GS-RF model, the rainwater pipe sections to be tested in the city can be divided into levels, namely, sections that do not need maintenance (no mixed connection), sections to be repaired (suspected mixed connection) and sections that urgently need maintenance (confirmed mixed connection);

[0019] According to the grade of the pipe sections, the pipe sections of different grades are managed and maintained in order.

[0020] Preferably, in the method for rapid detection of rainwater and sewage mixing based on the GS-RF model, the data to be collected are hydrogen sulfide concentration, methane concentration, oxygen concentration, carbon monoxide concentration, ammonia, nitrogen dioxide, carbon dioxide, organic volatile gases and working day conditions (WD).

[0021] Preferably, the steps for collecting the gas concentration information of the pipe section to be detected are as follows:

[0022] First, determine the rainwater inspection wells where gas needs to be collected based on the pipe network layout location information and regional related information;

[0023] Turn on the portable gas detector, wait for the instrument to warm up, and then observe the specific values ​​in the instrument;

[0024] According to the actual depth of the rainwater well and the type of gas to be collected, determine the descent depth of the collection pipe, then connect the instrument and start recording data.

[0025] According to the detection results of the instrument, the concentrations of several gases on the instrument are recorded every 10 minutes.

[0026] Preferably, the collected gas concentration information is preprocessed, specifically including noise removal, abnormal data processing and feature extraction.

[0027] Preferably, a machine learning model is established, and the processed data is used for analysis, and based on the analysis results, it is determined whether there is a mixed connection phenomenon in the pipeline.

[0028] It should be noted that due to the fluidity of gas, the criteria for determining whether rainwater and sewage are mixed in rainwater inspection wells are CCTV and gas detection results. The purpose is to detect whether there are characteristic gases in sewage pipes in rainwater pipes. If not, there is no mixing (I). If there are characteristic gases but no mixing points, there may be mixing points in the nearby pipe network, which is suspected mixing (II). If there are mixing points and characteristic gases, it is confirmed mixing (III).

[0029] Preferably, the RF model is optimized using the GS algorithm as a basic model to eliminate low-correlation influencing factors, simplify the model, and improve prediction accuracy.

[0030] Preferably, the steps of optimizing the RF model using the GS algorithm as the basic model are as follows:

[0031] First determine the parameter range to be tuned;

[0032] Use the GridSearchCV class to create a grid search object and pass in the parameter range and scoring index to be tuned;

[0033] Call the fit method of the grid search object to train the model, train and evaluate the performance of each set of parameters according to the listed parameters;

[0034] Get the best parameter combination through the best_params_ attribute;

[0035] Use the best_estimator_ attribute to get the best model and evaluate the performance of the best model on the test set;

[0036] The parameter range is iteratively adjusted based on the evaluation results to find the most appropriate parameter combination to optimize the performance of the RF model.

[0037] Compared with the prior art, the present invention has the following beneficial effects:

[0038] By continuously monitoring the concentration of gas in the rainwater pipes and combining machine learning models to determine whether rainwater and sewage mixing occurs, it is possible to quickly detect and classify the mixing level as soon as the rainwater and sewage mixing occurs, thereby improving work efficiency. This is specifically explained in the following points:

[0039] The mixed connection of rainwater pipes is judged by multiple factors, which improves the accuracy of model judgment;

[0040] According to the different nature of land use, relevant parameters are established for optimization to enhance the applicability of the model;

[0041] The use of data preprocessing improves data accuracy and reduces the high requirements of machine learning models for historical data;

[0042] By using machine learning models to analyze and compare various data, we can not only greatly reduce the use of manpower and material resources, save costs and detection time, and improve work efficiency, but also reduce the probability of erroneous analysis. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] Figure 1 A flow chart for constructing the GS-RF model of the present invention;

[0044] Figure 2 It is a multi-gas correlation thermodynamic diagram of the present invention;

[0045] Figure 3 It is a schematic diagram of the prediction result of the support vector machine of the present invention;

[0046] Figure 4 This is a schematic diagram of the random forest prediction results of the present invention;

[0047] Figure 5 It is a schematic diagram of the Bayesian prediction results of the present invention;

[0048] Figure 6 It is a schematic diagram of the decision tree prediction results of the present invention;

[0049] Figure 7 It is a schematic diagram of the prediction results of the artificial neural network of the present invention;

[0050] Figure 8 This is a schematic diagram of the prediction results of the gradient boosting tree of the present invention;

[0051] Fig. 9This is a schematic diagram of the grid search optimization random forest prediction results of the present invention. DETAILED DESCRIPTION

[0052] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0053] See also Figure 1 As shown, a method for rapid detection of rainwater and sewage mixing is provided, comprising the following steps:

[0054] S1. Divide the industrial areas of a city and determine the pipe sections that need to be tested.

[0055] S2. Collect the gas concentration information of the rainwater inspection well and determine the original data set. The collected gases should include hydrogen sulfide, methane, oxygen, carbon monoxide, ammonia, nitrogen dioxide, carbon dioxide and organic volatile gases;

[0056] Among them, it was found through preliminary measurements that the gas concentration at a depth of 1m in most rainwater wells can basically reflect the gas conditions in the rainwater wells, so the gases in this embodiment are all collected at a depth of about 1m in the rainwater wells.

[0057] It should be noted that because most harmful gases (such as H2S and CH4) are heavier than air, they tend to sink and gather at the bottom; while light gases such as O2 are more likely to be distributed in the upper area of ​​the space. Therefore, sampling at a height of about 1 meter can take into account the behavioral characteristics of these two types of gases, thereby providing more comprehensive and accurate information.

[0058] S3. Based on the CCTV and gas detection results, three levels are divided into no mixing (I), suspected mixing (II) and confirmed mixing (III).

[0059] S4, through the Pearson correlation analysis, the correlation between the influencing factor dataset and the mixing level, e.g. Figure 2 As shown, among the gas factors, the O2 gas concentration is strongly correlated with the judgment of the mixing level. In addition, CO2, NH3, H2S, CH4 and WD are also correlated with the judgment of the mixing level to a certain extent.

[0060] S5. According to the result of step S4, the linear relationship between each feature and the target variable is evaluated by the F statistic formula, and the six factors with higher influencing factors, namely CO2, O2, NH3, H2S, CH4 and WD, are screened out. The calculation formula is as follows:

[0061] ;

[0062] Among them, TSS represents the total sum of squares, RSS represents the residual sum of squares, p represents the number of independent variables, and n represents the sample size.

[0063] S6. Based on the training set data and the closed-circuit television (CCTV) detection results, the grid search algorithm (GS) algorithm was used to find the best parameter combination of the random forest (RF) model to establish the analysis and judgment model GS-RF;

[0064] S7, input the data into the established GS-RF model, output the judgment result of rainwater and sewage mixing in the test set, and compare it with the real value to determine the practicality of the model;

[0065] S8. Through the established GS-RF model, the rainwater pipe sections to be tested in the city can be divided into levels, namely, pipe sections that do not need maintenance (no mixed connection), pipe sections to be repaired (suspected mixed connection) and pipe sections that urgently need maintenance (confirmed mixed connection);

[0066] S9. According to the pipe segment grades divided in step S8, the pipe segments of different grades are managed and maintained in order.

[0067] It should be noted that when performing management and maintenance, it is necessary to refer to the calendar chart to distinguish whether the corresponding date is a working day. If not, skip it.

[0068] Table 1 Original dataset

[0069]

[0070] All the data in Table 1 are divided into training set and test set at a ratio of 4:1.

[0071] The input layer of the random forest established in this embodiment is six influencing factor variables selected according to correlation, namely O2, H2S, CH4, NH3, CO2, and WD, and the output layer is the rain and sewage mixing level.

[0072] The processed data set was input into the optimized random forest model and the traditional machine learning algorithm model for training respectively to predict the mixing of rainwater and sewage in the rainwater pipe network. The mean absolute error (MAE) and mean square error (MSE) were selected as error indicators to confirm the prediction accuracy of the model.

[0073] Table 2 compares the prediction accuracy of several model test sets in the example. Among the basic algorithms, the random forest model has the best prediction effect, with MAE and MSE of 0.05473 and 0.05473 respectively, which are the lowest among all basic models, and R2 is also the highest among all basic models, reaching 0.90108. In order to further improve the prediction accuracy, the parameters in the random forest model are optimized by grid search. Based on its performance, MAE decreased by 0.498%, MSE decreased by 0.498%, and R2 increased by 0.9%, further improving the prediction performance of the model.

[0074] Table 2 Prediction results accuracy index of test set

[0075]

[0076] The specific prediction results of each prediction model are as follows: Figures 3 to 9 shown.

[0077] The above is a detailed description of an embodiment of the present invention, but the content is only a preferred embodiment of the present invention and cannot be considered to limit the scope of implementation of the present invention. All equivalent changes and improvements made within the scope of the present invention should still fall within the scope of the patent coverage of the present invention.

[0078] The above shows and describes the basic principles, main features and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited by the above embodiments. The above embodiments and descriptions are only preferred examples of the present invention and are not intended to limit the present invention. Without departing from the spirit and scope of the present invention, the present invention may have various changes and improvements, which fall within the scope of the present invention. The scope of protection of the present invention is defined by the attached claims and their equivalents.

Claims

1. A method for rapid detection of rainwater and sewage mixing, characterized in that: It includes the following steps: Collecting gas concentration information of the pipe section to be detected as an original data set, wherein the original data set includes a training set and a test set; According to the actual test results of CCTV detection, each pipe section is divided into three mixed connection levels, namely: No mixed connection, suspected mixed connection and confirmed mixed connection; The influencing factors of the correlation between the original data set and the hybrid level were determined by Pearson correlation analysis. Specifically: The F statistic formula is used to evaluate the linear relationship between each feature and the target variable and screen out representative influencing factors. The F statistic formula is as follows: ； In the formula, TSS represents the total sum of squares, RSS represents the residual sum of squares, p represents the number of independent variables, and n represents the sample size; Based on the training set data and the actual detection results of CCTV detection, the analysis and judgment model GS-RF is established by using the grid search GS algorithm to find the best parameter combination of the random forest RF model; Input the data into the established GS-RF model, output the judgment result of rainwater and sewage mixing in the test set, and compare it with the real value to determine the practicality of the model; Through the established GS-RF model, each pipe section is graded, where: The pipe sections without mixed connections are classified as those that do not require maintenance, the pipe sections with suspected mixed connections are classified as those that need maintenance, and the pipe sections with confirmed mixed connections are classified as those that urgently need maintenance; Among them, the gases collected by the detection pipe section include hydrogen sulfide, methane, oxygen, carbon monoxide, ammonia, nitrogen dioxide, carbon dioxide and organic volatile gases.

2. The rapid detection method for rainwater and sewage mixing according to claim 1 is characterized in that: The original data set is divided into a training set and a test set at a ratio of 4:

1.

3. The rapid detection method for mixed rain and sewage according to claim 1 is characterized in that: Pipe sections of different levels are managed and maintained in order of priority.

4. The rapid detection method for rainwater and sewage mixing according to claim 1 is characterized in that: The steps to collect gas concentration information of the pipe section to be tested are as follows: First, determine the rainwater inspection wells where gas needs to be collected based on the pipe network layout location information and regional related information; Turn on the portable gas detector, wait for the instrument to warm up, and then observe the specific values ​​in the instrument; According to the actual depth of the rainwater well and the type of gas to be collected, determine the descent depth of the collection pipe, then connect the instrument and start recording data; According to the detection results of the instrument, the concentrations of several gases on the instrument are recorded every 10 minutes.

5. The rapid detection method for mixed rain and sewage according to claim 4 is characterized in that: The collected gas concentration information is preprocessed, including noise removal, abnormal data processing and feature extraction.

6. The method for rapid detection of rainwater and sewage mixing according to claim 4 is characterized in that: The standard for dividing the mixing level is to detect whether there are characteristic gases in the sewage pipe in the rainwater inspection well. If not, there is no mixing. If there are characteristic gases but no mixing points, it is suspected mixing. If there are mixing points and characteristic gases, it is confirmed mixing.

7. The rapid detection method for mixed rainwater and sewage according to claim 1 is characterized in that: The random forest RF model uses the grid search GS algorithm as the basic model for optimization to eliminate factors with low correlation.

8. The method for rapid detection of rainwater and sewage mixing according to claim 7 is characterized in that: The steps for optimizing the random forest RF model using the grid search GS algorithm as the basic model are as follows: First determine the parameter range to be tuned; Use the GridSearchCV class to create a grid search object and pass in the parameter range and scoring index to be tuned; Call the fit method of the grid search object to train the model, train and evaluate the performance of each set of parameters according to the listed parameters; Get the best parameter combination through the best_params_ attribute; Use the best_estimator_ attribute to get the best model and evaluate the performance of the best model on the test set; The parameter range is iteratively adjusted according to the evaluation results to find the most appropriate parameter combination to optimize the performance of the random forest RF model.

9. The method for rapid detection of rainwater and sewage mixing according to claim 1, characterized in that: Collect information on working days. Management and maintenance must be performed on working days.

Images