System and method for monitoring drainage pipe network in industrial park

Abstract

A method for improving the accuracy of water quality prediction in monitoring drainage pipe networks in industrial parks is provided. The method periodically collects water quality data from monitoring points in a drainage pipe network, uses the obtained water quality data from a first period as training data to generate a water quality prediction model based on a neural network, inputs verification data into the water quality prediction model in chronological order to obtain a first prediction result, and if any difference is greater than a predetermined difference, determines the difference as an abnormal difference and sets the verification data corresponding to the abnormal difference as abnormal water quality data.The method further obtains water quality data from the verification data that is normal and closest to the abnormal water quality data, regenerates second training data, and compares the first prediction accuracy of the first model with the second prediction accuracy of the second model to obtain an optimal prediction model.

Description

[Technical Field]

[0001] The present invention relates to the technical field of drainage pipe networks, and in particular to a system and method for monitoring drainage pipe networks in industrial parks. [Background technology]

[0002] With the rapid development of industrial parks, drainage pipe networks have become an important piece of infrastructure, and their operation status is directly related to the production safety and environmental protection of the park. Traditional drainage pipe network monitoring methods often have problems such as delayed information and limited monitoring range, making it difficult to meet the demands of drainage system management in modern industrial parks.

[0003] Similar prior art includes a Chinese patent application with publication number CN104697587A that discloses a monitoring device and its operating mechanism for use in an underground drainage pipe network, and a Chinese patent application with publication number CN107426539A that discloses a drainage pipe network monitoring system and method based on narrowband IoT. However, these two inventions do not take into account the possibility of drainage pipe network clogging due to the quality of wastewater in the drainage pipe network. Summary of the Invention [Means for solving the problem]

[0004] In order to further solve the above technical problems, the present invention provides a method for monitoring a drainage pipe network in an industrial park, which is realized by carrying out the following steps: Step S1: periodically collecting water quality data from each monitoring point in the drainage pipe network, using the obtained first period water quality data as sample data, and dividing the preprocessed sample data into training data and validation data according to time series; Step S2: respectively marking the actual results corresponding to each of the training data and the validation data, generating a water quality prediction model based on a neural network, and training the water quality prediction model using the training data to obtain a trained water quality prediction model; Step S3: inputting the verification data into the water quality prediction model in chronological order, obtaining a first prediction result, calculating multiple differences between the first prediction result and the actual result corresponding to the verification data, and if any of the differences is larger than a predetermined difference, determining the difference as an abnormal difference and determining the verification data corresponding to the abnormal difference as abnormal water quality data; Step S4: obtaining normal water quality data from the verification data that is closest to the abnormal water quality data, and combining it with multiple types of wastewater data stored in a storage unit to regenerate second training data; and step S5 of obtaining a first model and a second model based on the water quality prediction model, training the first model using a first proportion of second training data, training the second model using a second proportion of second training data, and comparing the first prediction accuracy of the first model with the second prediction accuracy of the second model to obtain an optimal prediction model.

[0005] The present invention further provides a drainage pipe network monitoring system for an industrial park, the system comprising: a collection unit for periodically collecting water quality data from each monitoring point in the drainage pipe network, using the obtained water quality data of a first period as sample data, and dividing the preprocessed sample data into training data and validation data according to time series; a primary training unit for respectively marking actual results corresponding to each of the training data and the validation data, generating a water quality prediction model based on a neural network, and training the water quality prediction model using the training data to obtain a trained water quality prediction model; an optimization unit for inputting the verification data into the water quality prediction model in chronological order, obtaining a first prediction result, calculating a plurality of differences between the first prediction result and the actual result corresponding to the verification data, and if any of the differences is greater than a predetermined difference, determining the difference as an abnormal difference, and determining the verification data corresponding to the abnormal difference as abnormal water quality data; a secondary training unit for acquiring normal water quality data from the verification data that is closest to the abnormal water quality data, and combining the data with multiple types of wastewater data stored in a storage unit to regenerate second training data; and a selection unit for respectively obtaining a first model and a second model based on the water quality prediction model, training the first model using a first proportion of second training data, training the second model using a second proportion of second training data, and comparing the first prediction accuracy of the first model with the second prediction accuracy of the second model to obtain an optimal prediction model. [Effects of the Invention]

[0006] Compared with the prior art, the present invention has at least the following beneficial effects: 1. The technical means of the present invention generates a water quality prediction model based on a neural network, and trains the water quality prediction model using training data to learn latent rules and features in the water quality data and improve prediction accuracy. Furthermore, by calculating multiple differences between the first prediction result and the actual result corresponding to the verification data, the water quality data in the verification data that is closest to the abnormal water quality data is obtained, and this is combined with multiple wastewater data stored in the memory unit to regenerate second training data, and multiple different proportions of wastewater data are obtained to continue training the water quality prediction model, so that the water quality prediction model can more accurately predict the likelihood of drainage pipe blockages. 2. The technical means of the present invention also obtains a first model and a second model based on the water quality prediction model, where the first model is trained using a first proportion of second training data, and the second model is trained using a second proportion of second training data, and the first prediction accuracy of the first model is compared with the second prediction accuracy of the second model to obtain an optimal prediction model. This invention can more accurately predict the likelihood of blockages occurring in a drainage network. [Brief explanation of the drawings]

[0007] In order to more clearly describe the technical aspects of the embodiments of the present invention or the prior art, the following briefly describes the drawings that need to be used in the description of the embodiments or the prior art. Obviously, the drawings in the following description are only the embodiments of the present invention, and those with ordinary skill in the art can obtain other drawings based on the provided drawings without paying creative labor. [Figure 1] 1 is a flowchart of a method for monitoring a drainage pipe network of an industrial park in the present invention. [Figure 2] 1 is a configuration diagram of an industrial park drainage pipe network monitoring system according to the present invention; DETAILED DESCRIPTION OF THE INVENTION

[0008] In order to clarify the objectives, claims and advantages of the present invention, the present invention will be described in more detail below in conjunction with drawings and examples. It should be understood that the specific embodiments described in this specification are only used to explain the present invention, and are not used to limit the present invention.

[0009] As used herein, the terms "first," "second," and the like may be used to describe various elements, but it will be understood that these elements are not limited to these terms unless otherwise specified. These terms are used only to distinguish one component from another. For example, a first xx script may be referred to as a second xx script, and similarly, a second xx script may be referred to as a first xx script, without departing from the scope of this application.

[0010] The present invention proposes a method for monitoring drainage pipe networks in industrial parks, as shown in Figure 1, by installing a water quality sensor and a power supply device directly above the inlet of each drainage pipe in the drainage pipe network in the industrial park as a monitoring point, and continuously monitoring the water quality of each drainage pipe in the drainage pipe network, which is realized by carrying out the following steps: Step S1: Periodically collect water quality data from each monitoring point in the drainage pipe network, and use the obtained water quality data of the first period as sample data. The preprocessed sample data is divided into training data and validation data according to time series, where the sample data belongs to time series data.

[0011] Specifically, water quality data from the industrial park's drainage network changes over time. Periodic data collection ensures data continuity and provides a data source for model training and prediction. In the above technical solution, the water quality data from the first period is selected as sample data for model training. Furthermore, preprocessing is performed to improve the data quality of the raw data. The training data and validation data are used to improve the accuracy of the water quality prediction model.

[0012] Step S2: Mark the actual results corresponding to each training data and validation data, respectively, generate a water quality prediction model based on the neural network, and use the training data to train the water quality prediction model to obtain a trained water quality prediction model.

[0013] Specifically, by marking the actual results corresponding to each training data and validation data, accurate monitoring information is provided for training the water quality prediction model, and the water quality prediction model is generated based on a neural network. The water quality prediction model automatically learns the latent rules and features in the water quality data, improving the accuracy of the prediction. If the difference between the predicted result of the water quality prediction model and the actual result is greater than the predetermined difference, further processing is performed, as described in detail below. The above technical means improve the accuracy of water quality prediction through steps such as data marking, model construction and training.

[0014] Step S3: Input data into the water quality prediction model in chronological order, obtain a first prediction result, calculate multiple differences between the first prediction result and the actual result corresponding to the verification data, and if any difference is larger than the preset difference, determine the difference as an abnormal difference, and determine the data corresponding to the abnormal difference as abnormal water quality data.

[0015] Step S4: Obtain the normal water quality data in the verification data that is closest to the abnormal water quality data, and combine it with the multiple types of wastewater data stored in the storage unit to regenerate the second training data.

[0016] Specifically, to timely detect and analyze the impact of water quality changes in a drainage network on the pipelines, the system compares the first prediction result with the actual result corresponding to the verification data, sets a predetermined difference as a threshold, and if any difference is greater than the predetermined difference, marks the difference as an abnormal difference and marks the corresponding verification data as abnormal water quality data. Furthermore, the system finds normal water quality data within the verification data that is closest to the abnormal water quality data, compares data points before and after the abnormal water quality data, and selects the data point closest to the abnormal point and whose difference is within the predetermined difference range as normal water quality data. The second training data is then generated by mixing the second training data with multiple wastewater data stored in the storage unit, and the water quality prediction model can continue to train with this data, allowing the water quality prediction model to more accurately predict the likelihood of drainage pipe clogs.

[0017] Step S5: Obtain a first model and a second model based on the water quality prediction model, train the first model using a first proportion of second training data, train the second model using a second proportion of second training data, and compare the first prediction accuracy of the first model with the second prediction accuracy of the second model to obtain an optimal prediction model.

[0018] Specifically, the present application divides the second training data into two parts with different proportions (e.g., 70% and 30%, or 80% and 20%), trains the first model and the second model respectively, performs predictions using the first model and the second model respectively on the same test set, and calculates their prediction accuracy.The prediction accuracy of the water quality prediction models using different proportions of training data is compared to select the optimal prediction model.

[0019] The coordination between the above steps allows for training a water quality prediction model with high prediction accuracy and high prediction efficiency, which can more accurately predict the likelihood of blockages occurring in the drainage pipe network.

[0020] Following step S5, The method includes step S6 of obtaining water quality data from a second cycle after the first cycle and inputting it into an optimal prediction model, the optimal prediction model outputting a second prediction result based on the input water quality data from the second cycle, issuing a warning if the second prediction result is greater than a preset threshold, and having relevant parties perform maintenance and verification of the drainage pipe network based on the warning information.

[0021] Specifically, traditional drainage network monitoring methods rely on manual or periodic inspections, resulting in slow response times. To achieve real-time water quality monitoring and early warning for industrial park drainage networks, the water quality data from the second cycle after the first cycle can be obtained and input into an optimal prediction model to predict the likelihood of blockages occurring in the drainage network.

[0022] Step S4 further comprises: Obtaining water quality data without abnormalities and water quality parameters of each wastewater data including pH value, dissolved oxygen, ammonia nitrogen, total phosphorus and total nitrogen; The method includes: mixing the water quality without abnormality with each wastewater in different proportions to obtain a plurality of mixed wastewaters; obtaining water quality parameters for each mixed wastewater; marking the actual results corresponding to each mixed wastewater; and using the water quality parameters and the corresponding actual results of all the mixed wastewaters as second training data.

[0023] Specifically, to improve the prediction accuracy of the water quality prediction model, more representative second training data was generated to simulate various water quality conditions that may be encountered in an actual drainage network. Normal water quality data was mixed with each wastewater data in different ratios, such as 1:1.2, 1:1.5, and 2:1.5. The actual results corresponding to each mixed wastewater, such as a 60% probability of causing drainage blockage, were also displayed. The water quality parameters of all mixed wastewaters and their corresponding actual results were compiled into the second training data. The second training data includes data for multiple cases of different water qualities, enabling the training of a more accurate water quality prediction model.

[0024] The step S5 further includes the following steps: Step S51: define the first ratio as i% and the second ratio as j%, respectively obtain second training data of the first ratio as the first data, obtain second training data of the second ratio as the second data, train water quality prediction models based on the first data and the second data, respectively, and obtain the trained first model and second model, where i is a number greater than or equal to 1 and less than j, and j is a number greater than or equal to i and less than 99.

[0025] Step S52: input the validation data into the first model and the second model in chronological order, obtain the first result and the second result, calculate the first prediction accuracy of the first model based on the first result and the actual result corresponding to the validation data, and similarly, calculate the second prediction accuracy of the second model based on the second result and the actual result corresponding to the validation data.

[0026] Step S53: Compare the first prediction accuracy with the second prediction accuracy. If the second prediction accuracy is equal to or greater than the first prediction accuracy and equal to or greater than a predetermined accuracy, calculate a floating value, re-acquire data from the second training data based on the first ratio and the floating value, train a first model, and re-acquire data from the second training data based on the second ratio and the floating value, train a second model, update the trained first model and the second model, and simultaneously set the sum of the first ratio and the floating value as the updated first ratio, and the difference between the second ratio and the floating value as the updated second ratio. If the second prediction accuracy is smaller than the predetermined accuracy, calculate a floating value, acquire data from the second training data based on the first ratio and the floating value, train a first model, and update the trained first model, and simultaneously set the sum of the first ratio and the floating value as the updated first ratio. The j-th data corresponding to the smallest j-value at which the second prediction accuracy is equal to or greater than the predetermined accuracy is used as training data for the second model.

[0027] In step S54, steps S52 and S53 are repeated until the accuracy of the first prediction accuracy is equal to or greater than the accuracy of the second prediction accuracy, thereby obtaining an optimal prediction model.

[0028] To further improve the predictive accuracy of the water quality prediction model, the first training run involved obtaining 1,000 pieces of data with a first proportion of 10% and a second proportion of 90% from each of the 1,000 second training data sets, training the water quality prediction model and obtaining Model 1 and Model 2. Using the validation data, the first prediction accuracy of Model 1 and the second prediction accuracy of Model 2 were calculated to be 40% and 99%, respectively. It was assumed that the water quality prediction model could meet actual demand with just a 90% prediction accuracy; using too much training data would actually reduce the efficiency of the water quality prediction model's execution. In this case, reducing the amount of training data would improve the execution efficiency of the water quality prediction model. Comparing the first and second prediction accuracies, the amount of training data used to train the second model is generally much larger than the amount of training data used to train the first model, resulting in a higher second prediction accuracy. However, if the second prediction accuracy is higher than the preset accuracy, the second prediction accuracy can be reduced by 90% and the second model retrained. For example, 82% of the data is obtained from the second training data. Specifically, the amount of data corresponding to the water quality prediction model trained with a small amount of data is low, resulting in high accuracy. The first model can be retrained by increasing the amount of training data corresponding to 40% of the first prediction accuracy. For example, the first proportion is increased from 10% to 10% + 8%. By calculating a floating value based on the proportion of data initially acquired and increasing or decreasing the range of training data, training data that matches the amount of data can be accurately found. On the other hand, if the second prediction accuracy is higher than the preset accuracy, training data can be reacquired with a new first proportion and the first model retrained until the accuracy of the first prediction accuracy is equal to or greater than the accuracy of the second prediction accuracy or the accuracy of the first prediction accuracy is equal to or greater than the preset accuracy.

[0029] Furthermore, the first prediction accuracy of the first model is calculated by the following formula:

[0030]

number

[0031] Here, e1 represents the first prediction accuracy of the first model, n represents the number of the first data, S1 represents the first result, and S2 represents the actual result corresponding to the validation data.

[0032] Furthermore, the calculation of the floating value f is realized by the following formula:

[0033]

number

[0034] where f represents the floating value and α represents the floating coefficient.

[0035] Specifically, each time the first prediction accuracy is compared with the second prediction accuracy, the floating value is recalculated, and as the prediction accuracy improves, the amount of change in the acquired data decreases, and the prediction accuracy of the water quality prediction model increases.

[0036] The present invention provides a drainage pipe network monitoring system for an industrial park as shown in Figure 2, which is installed as a monitoring point by installing a water quality sensor and a power supply device at the position directly above the inlet of each drainage pipe inside the drainage pipe network of the industrial park. a collection unit that periodically collects water quality data from each monitoring point in the drainage pipe network, uses the obtained water quality data of the first period as sample data, and divides the preprocessed sample data into training data and validation data according to time series; a primary training unit that respectively marks the actual results corresponding to each training data and validation data, generates a water quality prediction model based on a neural network, and uses the training data to train the water quality prediction model to obtain a trained water quality prediction model; an optimization unit for inputting the verification data into the water quality prediction model in chronological order, obtaining a first prediction result, calculating multiple differences between the first prediction result and the actual result corresponding to the verification data, and determining that any difference is greater than a predetermined difference as an abnormal difference, and determining that the verification data corresponding to the abnormal difference is abnormal water quality data; a secondary training unit for acquiring normal water quality data that is closest to the abnormal water quality data in the verification data, and combining the data with the multiple types of wastewater data stored in the storage unit to regenerate second training data; and a selection unit for respectively obtaining a first model and a second model based on the water quality prediction model, training the first model using a first proportion of second training data, training the second model using a second proportion of second training data, and comparing the first prediction accuracy of the first model with the second prediction accuracy of the second model to obtain an optimal prediction model.

[0037] For ease of explanation and conciseness, those skilled in the art can clearly understand that the specific operation processes of the above-mentioned systems, systems, and units can refer to the corresponding processes in the above-mentioned method embodiments, and will not be further described here.

[0038] The integrated unit may be implemented as a software functional unit and stored in a computer-readable storage medium when sold or used as an independent product. Based on this understanding, the technical approach of the present application may be essentially embodied or, in whole or in part, contribute to the prior art in the form of a software product stored in a storage medium containing instructions for performing all or part of the steps of the methods described in each embodiment of the present application. Meanwhile, the storage medium may include various media capable of storing program code, such as a U disk, a removable hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk.

[0039] As explained above, the above examples are only used to explain the technical solutions of the present application, and are not intended to limit the same. The present application has been described in detail with reference to the above embodiments, but those skilled in the art may modify the technical solutions described in each of the above embodiments or replace some of the technical features with equivalents, and these modifications or substitutions will not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of each of the examples of the present application.

Claims

1. Step S1: periodically collecting water quality data from each monitoring point in a drainage pipe network, using the obtained water quality data of the first period as sample data, and dividing the pre-processed sample data into training data and validation data according to time series; Step S2: respectively marking the actual results corresponding to each of the training data and the validation data, generating a water quality prediction model based on a neural network, and training the water quality prediction model using the training data to obtain a trained water quality prediction model; Step S3: inputting the verification data into the water quality prediction model in chronological order, obtaining a first prediction result, calculating multiple differences between the first prediction result and the actual result corresponding to the verification data, and if any of the differences is larger than a predetermined difference, determining the difference as an abnormal difference and determining the verification data corresponding to the abnormal difference as abnormal water quality data; Step S4: obtain parameters of normal water quality data in the verification data that are closest to the abnormal water quality data, mix the parameters of the normal water quality data with water quality parameters of multiple types of wastewater data stored in a storage unit in different proportions to obtain water quality parameters of multiple mixed wastewaters, mark actual results corresponding to each of the mixed wastewaters, and generate water quality parameters of all the mixed wastewaters and the corresponding actual results as second training data; and (S5) obtaining a first model and a second model based on the water quality prediction model, training the first model using a first proportion of second training data, training the second model using a second proportion of second training data, and comparing the first prediction accuracy of the first model with the second prediction accuracy of the second model to obtain an optimal prediction model.

2. Following step S5, 2. The method for monitoring a drainage pipe network in an industrial park according to claim 1, further comprising the step S6 of obtaining water quality data from a second period after the first period and inputting it into the optimal prediction model, the optimal prediction model outputting a second prediction result based on the input water quality data from the second period, issuing a warning if the second prediction result is greater than a preset threshold, and having relevant parties maintain and verify the drainage pipe network based on the warning information.

3. Step S4 further comprises: Acquiring the water quality data without abnormalities and water quality parameters of each wastewater data including pH value, dissolved oxygen, ammonia nitrogen, total phosphorus and total nitrogen; 2. The method for monitoring a drainage pipe network in an industrial park according to claim 1, further comprising: mixing the parameters of the water quality data without abnormalities and the water quality parameters of each wastewater in different proportions to obtain water quality parameters of multiple mixed wastewaters; marking the actual results corresponding to each mixed wastewater; and using the water quality parameters of all the mixed wastewaters and the corresponding actual results as second training data.

Images