A method and system for detecting a leakage of a buried pipeline of a thermal power plant

Abstract

The application provides a kind of power plant buried pipe leakage detection method and system, method includes: obtaining historical buried pipeline pressure detection data set;The leaked pipeline pressure information is carried out data preprocessing, to be arranged into the pressure detection sequence with pressure leakage position as feature;Cycle neural network is constructed, and pressure detection sequence and un-leaked pipeline pressure information are input into cycle neural network;A plurality of pressure detection devices are set on target pipeline, and the pressure information collected by each pressure detection device is carried out data cleaning processing, to obtain cleaned pressure information;The cleaned pressure information is weighted using weighting coefficient, to obtain the weighted pressure information of entire target pipeline;Weighted pressure information is input into pressure detection model, to obtain detection result.The application can realize fast and accurate detection of buried pipeline leakage by combining cycle neural network and signal processing and data analysis technology, and provides a strong guarantee for safe operation of power plant.

Description

Technical Field

[0001] This invention relates to the field of leakage detection technology, and in particular to a method and system for detecting leaks in underground pipelines of thermal power plants. Background Technology

[0002] Buried pipelines in thermal power plants often transport high-temperature, high-pressure fluids. If a leak occurs, it will pose a serious threat to the environment and personnel safety.

[0003] A patent application with application number 202210163155.1 discloses a method for detecting leaks in buried tailings pipelines, including a front-end detection instrument, a back-end monitoring unit, and a controller. The front-end detection instrument is installed at a preset location on the pipeline to detect data at that location. The controller is connected to the front-end detection instrument and the back-end monitoring unit, receiving the data detected by the front-end detection instrument and calculating, analyzing, and judging the data using a preset algorithm within the controller. The back-end monitoring unit issues an alarm for data that does not meet the judgment conditions to ensure the normal operation of the pipeline. This method for detecting leaks in buried tailings pipelines achieves automatic early warning of leaks. However, by only installing the front-end detection instrument at a preset location, this design cannot comprehensively monitor the entire pipeline. Leaks may occur in other locations, but the system may not detect them in time, leading to the risk of missed detections. Summary of the Invention

[0004] In order to overcome the shortcomings of the prior art, the purpose of this invention is to provide a method and system for detecting leaks in underground pipelines of thermal power plants.

[0005] To achieve the above objectives, the present invention provides the following solution:

[0006] A method for detecting leaks in underground pipelines of thermal power plants includes:

[0007] Obtain a historical buried pipeline pressure detection dataset; the historical buried pipeline pressure detection dataset includes pressure information of pipelines that have not been leaked and pressure information of pipelines that have been leaked;

[0008] The leaked pipeline pressure information is preprocessed to form a pressure detection sequence characterized by the location of the pressure leak.

[0009] A recurrent neural network is constructed, and the pressure detection sequence and the pressure information of the non-leaking pipeline are input into the recurrent neural network to obtain a trained pressure detection model.

[0010] Multiple pressure detection devices are installed on the target pipeline, and the pressure information collected by each pressure detection device is cleaned to obtain cleaned pressure information.

[0011] The pressure information after cleaning is weighted using weighting coefficients to obtain the weighted pressure information of the entire target pipeline;

[0012] The weighted pressure information is input into the pressure detection model to obtain the detection result.

[0013] Preferably, the method for constructing the pressure detection model is as follows:

[0014] A fully connected network is connected after the output of the two-layer gated recurrent unit network to build a multi-layer GRU network model;

[0015] Initialize the model parameters of the multilayer GRU network model: set the number of hidden neurons in the first and second layers to 10, and the number of hidden neurons in the fully connected layer to 24; set the learning rate to 0.06, select sigmoid and tanh as activation functions, and set the number of network iterations to 100;

[0016] The pressure detection sequence and the pressure information of the non-leaking pipeline are input into the multi-layer GRU network model for training; the trained pressure detection model is obtained; the output of the pressure detection model is the location of the pressure leak.

[0017] Preferably, the pressure information collected by each of the pressure detection devices is subjected to data cleaning processing to obtain cleaned pressure information, including:

[0018] Calculate the variance of pressure from the pressure monitoring device during each data acquisition cycle;

[0019] A pressure acquisition model is constructed using the variance.

[0020] The degree of connection between each pressure detection device is calculated based on the pressure acquisition model.

[0021] Construct a connection degree matrix based on the connection degree, and determine the weighted connection degree of each pressure detection device;

[0022] Determine whether the weighted correlation degree is greater than a preset threshold;

[0023] If the weighted correlation degree is greater than the preset threshold, the pressure information collected by the corresponding pressure detection device will be removed to obtain the cleaned pressure information.

[0024] Preferably, the formula for the pressure acquisition model is: Where, δ i δ represents the variance of the pressure values ​​of the pressure detection device during the i-th acquisition period. j x represents the variance of the pressure values ​​of the pressure detection device during the j-th data acquisition period. ip represents the average pressure value of the pressure detection device during the i-th acquisition period. i (x / x i ) represents the data acquisition model of the i-th pressure detection device, p j (x / x j ) represents the data acquisition model of the j-th pressure detection device.

[0025] Preferably, the correlation between each pressure detection device is calculated based on the pressure acquisition model, including:

[0026] The trust level between various pressure sensing devices is determined using a pressure acquisition model; wherein the trust level is calculated using the following formula: Where, d′ ij d′ represents the trust level between the i-th pressure testing device and the j-th pressure testing device. ji This represents the trust level between the j-th pressure testing device and the i-th pressure testing device;

[0027] The degree of connection between each pressure testing device is calculated based on the trust level between the various pressure testing devices.

[0028] Preferably, the formula for calculating the degree of connection between each pressure testing device is: Among them, s ij This represents the degree of connection between the i-th pressure testing device and the j-th pressure testing device.

[0029] Preferably, a connection degree matrix is ​​constructed based on the connection degree, and the weighted connection degree of each pressure detection device is determined, including:

[0030] A connection degree matrix is ​​determined based on the connection degree between each pressure testing device; wherein, the connection degree matrix is:

[0031] The weighted connection degree of each pressure testing device is determined using the connection degree matrix; wherein, the formula for calculating the weighted connection degree is: Among them, R i This represents the weighted correlation degree of the i-th pressure detection device.

[0032] Preferably, the pressure information after cleaning is weighted using weighting coefficients to obtain the weighted pressure information of the entire target pipeline, including:

[0033] The weighting coefficients are calculated based on the variance of the pressure information after cleaning within each collection cycle.

[0034] The pressure information after cleaning is weighted and averaged based on the weighting coefficients to obtain the weighted pressure information of the entire target pipeline.

[0035] Preferably, the formula for calculating the weighting coefficients is:

[0036]

[0037] Among them, W i Let σ′ be the i-th weighting coefficient. i This represents the variance of the pressure information after cleaning during the i-th acquisition period.

[0038] A leak detection system for underground pipelines in thermal power plants includes:

[0039] The dataset acquisition module is used to acquire historical buried pipeline pressure detection datasets; the historical buried pipeline pressure detection datasets include pressure information of pipelines that have not been leaked and pressure information of pipelines that have been leaked;

[0040] The preprocessing module is used to preprocess the leaked pipeline pressure information to organize it into a pressure detection sequence characterized by the location of the pressure leak.

[0041] The pressure detection model construction module is used to construct a recurrent neural network and input the pressure detection sequence and the pressure information of the non-leaking pipeline into the recurrent neural network to obtain a trained pressure detection model.

[0042] The data acquisition module is used to set up multiple pressure detection devices on the target pipeline and perform data cleaning processing on the pressure information collected by each pressure detection device to obtain cleaned pressure information.

[0043] The weighting module is used to weight the pressure information after cleaning using weighting coefficients to obtain the weighted pressure information of the entire target pipeline;

[0044] The detection module is used to input the weighted pressure information into the pressure detection model to obtain the detection result.

[0045] According to specific embodiments provided by the present invention, the present invention discloses the following technical effects:

[0046] This invention provides a method and system for detecting leaks in buried pipelines of thermal power plants. The method includes: acquiring a historical buried pipeline pressure detection dataset; the historical buried pipeline pressure detection dataset includes pressure information of non-leaking pipelines and pressure information of leaking pipelines; preprocessing the pressure information of leaking pipelines to form a pressure detection sequence characterized by the location of the pressure leak; constructing a recurrent neural network and inputting the pressure detection sequence and the pressure information of non-leaking pipelines into the recurrent neural network to obtain a trained pressure detection model; setting up multiple pressure detection devices on the target pipeline and cleaning the pressure information collected by each pressure detection device to obtain cleaned pressure information; weighting the cleaned pressure information using weighting coefficients to obtain weighted pressure information for the entire target pipeline; and inputting the weighted pressure information into the pressure detection model to obtain the detection result. This invention, by combining recurrent neural networks and signal processing and data analysis techniques, can achieve rapid and accurate detection of leaks in buried pipelines, providing a strong guarantee for the safe operation of thermal power plants. Attached Figure Description

[0047] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0048] Figure 1 A flowchart of the method provided in an embodiment of the present invention;

[0049] Figure 2 This is a schematic diagram of the system structure provided in an embodiment of the present invention. Detailed Implementation

[0050] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0051] The purpose of this invention is to provide a method for detecting leaks in underground pipelines of thermal power plants. By combining recurrent neural networks and signal processing and data analysis techniques, it is possible to achieve rapid and accurate detection of leaks in underground pipelines, providing a strong guarantee for the safe operation of thermal power plants.

[0052] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0053] Figure 1 The method flowchart provided in the embodiments of the present invention is as follows: Figure 1 As shown, the present invention provides a method for detecting leaks in underground pipelines of thermal power plants, comprising:

[0054] Step 100: Obtain historical buried pipeline pressure detection dataset; the historical buried pipeline pressure detection dataset includes pressure information of pipelines that have not leaked and pressure information of pipelines that have leaked.

[0055] Step 200: Perform data preprocessing on the leaked pipeline pressure information to organize it into a pressure detection sequence characterized by the location of the pressure leak;

[0056] Step 300: Construct a recurrent neural network and input the pressure detection sequence and the pressure information of the leak-free pipeline into the recurrent neural network to obtain a trained pressure detection model;

[0057] Step 400: Set up multiple pressure detection devices on the target pipeline, and perform data cleaning processing on the pressure information collected by each pressure detection device to obtain cleaned pressure information;

[0058] Step 500: Use weighting coefficients to weight the pressure information after cleaning to obtain the weighted pressure information of the entire target pipeline;

[0059] Step 600: Input the weighted pressure information into the pressure detection model to obtain the detection results.

[0060] Preferably, the method for constructing the pressure detection model is as follows:

[0061] A fully connected network is connected after the output of the two-layer gated recurrent unit network to build a multi-layer GRU network model;

[0062] Initialize the model parameters of the multilayer GRU network model: set the number of hidden neurons in the first and second layers to 10, and the number of hidden neurons in the fully connected layer to 24; set the learning rate to 0.06, select sigmoid and tanh as activation functions, and set the number of network iterations to 100;

[0063] The pressure detection sequence and the pressure information of the non-leaking pipeline are input into the multi-layer GRU network model for training; the trained pressure detection model is obtained; the output of the pressure detection model is the location of the pressure leak.

[0064] The gated recurrent unit (GRU) in this embodiment is a type of recurrent neural network, which introduces an update gate vt Reset door r t Two gating units address the vanishing and exploding gradient problems in RNNs, and combining them with fully connected layers can more effectively solve data nonlinearity issues. The forward propagation function corresponding to the stress detection model is summarized below, where k is the number of layers in the network:

[0065] r t =σ(λ) r ·[h t-1 ,X t ]+b r )

[0066] v t =σ(λ) v ·[h t-1 ,X t ]+b v )

[0067] h t =tanh(λ) h ·[r t ·Y t-1 ,X t ]+b h )

[0068] y t =(1-v t )·Y t-1 +v t ·h t

[0069]

[0070] After forward propagation, a suitable loss function is selected for backpropagation to update the model parameters. Training of the stress detection model ends when the loss value tends to stabilize.

[0071] Furthermore, this embodiment takes into account the difficulty of predicting pipeline pressure in reality. When predicting pipeline pressure after a period of time, it is insufficient to only know the flow rate at each interface of the pipeline at the current moment. Pressure information at this moment and several moments before also has a significant impact on pressure prediction. Therefore, this embodiment uses a constructed pressure detection model to process the entire sequence connected by these flow rate information.

[0072] Preferably, the pressure information collected by each of the pressure detection devices is subjected to data cleaning processing to obtain cleaned pressure information, including:

[0073] Calculate the variance of pressure from the pressure monitoring device during each data acquisition cycle;

[0074] A pressure acquisition model is constructed using the variance.

[0075] The degree of connection between each pressure detection device is calculated based on the pressure acquisition model.

[0076] Construct a connection degree matrix based on the connection degree, and determine the weighted connection degree of each pressure detection device;

[0077] Determine whether the weighted correlation degree is greater than a preset threshold;

[0078] If the weighted correlation degree is greater than the preset threshold, the pressure information collected by the corresponding pressure detection device will be removed to obtain the cleaned pressure information.

[0079] Specifically, because the pressure testing equipment collects a large amount of data, data cleaning is required to remove abnormal data before weighting the pressure data measured by multiple pressure testing devices.

[0080] Preferably, the formula for the pressure acquisition model is: Where, δ i δ represents the variance of the pressure values ​​of the pressure detection device during the i-th acquisition period. j x represents the variance of the pressure values ​​of the pressure detection device during the j-th data acquisition period. i p represents the average pressure value of the pressure detection device during the i-th acquisition period. i (x / x i ) represents the data acquisition model of the i-th pressure detection device, p j (x / x j ) represents the data acquisition model of the j-th pressure detection device.

[0081] Furthermore, in addition to the above-mentioned processing of the pressure data, this embodiment also performs noise reduction processing on the collected pressure data, uses Fast Fourier Transform (FFT) to perform frequency analysis on the collected signal, and combines filtering technology to remove background noise and extract leakage signal features.

[0082] Preferably, the correlation between each pressure detection device is calculated based on the pressure acquisition model, including:

[0083] The trust level between various pressure sensing devices is determined using a pressure acquisition model; wherein the trust level is calculated using the following formula: Where, d′ ij d′ represents the trust level between the i-th pressure testing device and the j-th pressure testing device. ji This represents the trust level between the j-th pressure testing device and the i-th pressure testing device;

[0084] The degree of connection between each pressure testing device is calculated based on the trust level between the various pressure testing devices.

[0085] Preferably, the formula for calculating the degree of connection between each pressure testing device is: Among them, s ij This represents the degree of connection between the i-th pressure testing device and the j-th pressure testing device.

[0086] Preferably, a connection degree matrix is ​​constructed based on the connection degree, and the weighted connection degree of each pressure detection device is determined, including:

[0087] A connection degree matrix is ​​determined based on the connection degree between each pressure testing device; wherein, the connection degree matrix is:

[0088] The weighted connection degree of each pressure testing device is determined using the connection degree matrix; wherein, the formula for calculating the weighted connection degree is: Among them, R i This represents the weighted correlation degree of the i-th pressure detection device.

[0089] Preferably, the pressure information after cleaning is weighted using weighting coefficients to obtain the weighted pressure information of the entire target pipeline, including:

[0090] The weighting coefficients are calculated based on the variance of the pressure information after cleaning within each collection cycle.

[0091] The pressure information after cleaning is weighted and averaged based on the weighting coefficients to obtain the weighted pressure information of the entire target pipeline.

[0092] Preferably, the formula for calculating the weighting coefficients is:

[0093]

[0094] Among them, W i Let σ′ be the i-th weighting coefficient. i This represents the variance of the pressure information after cleaning during the i-th acquisition period.

[0095] Specifically, this invention deploys multiple pressure detection devices on the target pipeline and uses weighting coefficients to weight the pressure information to obtain the weighted pressure of the entire pipeline. This weighted pressure information can reflect the pressure leakage situation of the entire pipeline.

[0096] Corresponding to the above methods, such as Figure 2 As shown, this embodiment also provides a leak detection system for underground pipelines in thermal power plants, including:

[0097] The dataset acquisition module is used to acquire historical buried pipeline pressure detection datasets; the historical buried pipeline pressure detection datasets include pressure information of pipelines that have not been leaked and pressure information of pipelines that have been leaked;

[0098] The preprocessing module is used to preprocess the leaked pipeline pressure information to organize it into a pressure detection sequence characterized by the location of the pressure leak.

[0099] The pressure detection model construction module is used to construct a recurrent neural network and input the pressure detection sequence and the pressure information of the non-leaking pipeline into the recurrent neural network to obtain a trained pressure detection model.

[0100] The data acquisition module is used to set up multiple pressure detection devices on the target pipeline and perform data cleaning processing on the pressure information collected by each pressure detection device to obtain cleaned pressure information.

[0101] The weighting module is used to weight the pressure information after cleaning using weighting coefficients to obtain the weighted pressure information of the entire target pipeline;

[0102] The detection module is used to input the weighted pressure information into the pressure detection model to obtain the detection result.

[0103] The beneficial effects of this invention are as follows:

[0104] This invention, by combining recurrent neural networks and signal processing and data analysis techniques, enables rapid and accurate detection of leaks in underground pipelines, providing strong protection for the safe operation of thermal power plants.

[0105] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For the systems disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the descriptions are relatively simple; relevant parts can be referred to the method section.

[0106] This document uses specific examples to illustrate the principles and implementation methods of the present invention. The descriptions of the above embodiments are only for the purpose of helping to understand the method and core ideas of the present invention. Furthermore, those skilled in the art will recognize that, based on the ideas of the present invention, there will be changes in the specific implementation methods and application scope. Therefore, the content of this specification should not be construed as a limitation of the present invention.

Claims

1. A method for detecting leaks in underground pipelines of thermal power plants, characterized in that, include: Obtain a historical buried pipeline pressure detection dataset; the historical buried pipeline pressure detection dataset includes pressure information of pipelines that have not been leaked and pressure information of pipelines that have been leaked; The leaked pipeline pressure information is preprocessed to form a pressure detection sequence characterized by the location of the pressure leak. A recurrent neural network is constructed, and the pressure detection sequence and the pressure information of the non-leaking pipeline are input into the recurrent neural network to obtain a trained pressure detection model. Multiple pressure detection devices are installed on the target pipeline, and the pressure information collected by each pressure detection device is cleaned to obtain cleaned pressure information. The pressure information after cleaning is weighted using weighting coefficients to obtain the weighted pressure information of the entire target pipeline; The weighted pressure information is input into the pressure detection model to obtain the detection result; The pressure information collected by each of the pressure detection devices is cleaned to obtain cleaned pressure information, including: Calculate the variance of pressure from the pressure monitoring device during each data acquisition cycle; A pressure acquisition model is constructed using the variance. The degree of connection between each pressure detection device is calculated based on the pressure acquisition model. Construct a connection degree matrix based on the connection degree, and determine the weighted connection degree of each pressure detection device; Determine whether the weighted correlation degree is greater than a preset threshold; If the weighted correlation degree is greater than the preset threshold, the pressure information collected by the corresponding pressure detection device is removed to obtain the pressure information after cleaning. The formula for the pressure acquisition model is: ;in, This represents the variance of the pressure values ​​of the pressure detection device during the i-th data acquisition period. This represents the variance of the pressure values ​​measured by the pressure detection device during the j-th data acquisition period. This represents the average pressure value of the pressure detection device during the i-th data acquisition period. Let represent the data acquisition model of the i-th pressure detection device. This represents the data acquisition model of the j-th pressure detection device; The correlation between each pressure detection device is calculated based on the pressure acquisition model, including: The trust level between various pressure sensing devices is determined using a pressure acquisition model; wherein the trust level is calculated using the following formula: ;in, Indicates the first i The pressure testing device and the first j Trust between pressure testing devices Indicates the first j The pressure testing device and the first i Trust between pressure testing devices; The degree of connection between each pressure detection device is calculated based on the trust level between the various pressure detection devices. The formula for calculating the connectivity between each pressure testing device is: ;in, Indicates the first i The pressure testing device and the first j The degree of interconnectivity between pressure testing devices; Based on the aforementioned connection degree, a connection degree matrix is ​​constructed, and the weighted connection degree of each pressure detection device is determined, including: A connection degree matrix is ​​determined based on the connection degree between each pressure testing device; wherein, the connection degree matrix is: ; The weighted connection degree of each pressure testing device is determined using the connection degree matrix; wherein, the formula for calculating the weighted connection degree is: ;in, Indicates the first i Weighted correlation degree of each pressure testing device; The collected pressure data was denoised, and the collected signals were analyzed by using Fast Fourier Transform (FFT). Background noise was removed by combining filtering techniques, and leakage signal features were extracted.

2. The method for detecting leaks in underground pipelines of thermal power plants according to claim 1, characterized in that, The method for constructing the pressure detection model is as follows: A fully connected network is connected after the output of the two-layer gated recurrent unit network to build a multi-layer GRU network model; Initialize the model parameters of the multilayer GRU network model: set the number of hidden neurons in the first and second layers to 10, and the number of hidden neurons in the fully connected layer to 24; set the learning rate to 0.06, select sigmoid and tanh as activation functions, and set the number of network iterations to 100; The pressure detection sequence and the pressure information of the leak-free pipeline are input into the multi-layer GRU network model for training; The trained pressure detection model is obtained; the output of the pressure detection model is the location of the pressure leak.

3. The method for detecting leaks in underground pipelines of thermal power plants according to claim 1, characterized in that, The pressure information after cleaning is weighted using weighting coefficients to obtain the weighted pressure information of the entire target pipeline, including: The weighting coefficients are calculated based on the variance of the pressure information after cleaning within each collection cycle. The pressure information after cleaning is weighted and averaged based on the weighting coefficients to obtain the weighted pressure information of the entire target pipeline.

4. The method for detecting leaks in underground pipelines of thermal power plants according to claim 3, characterized in that, The formula for calculating the weighting coefficients is as follows: in, For the first i Each weighting coefficient This represents the variance of the pressure information after cleaning during the i-th acquisition period.

5. A leak detection system for underground pipelines in thermal power plants, characterized in that, The system for implementing the method for detecting leaks in underground pipelines of thermal power plants as described in any one of claims 1 to 4 includes: The dataset acquisition module is used to acquire historical buried pipeline pressure detection datasets; the historical buried pipeline pressure detection datasets include pressure information of pipelines that have not been leaked and pressure information of pipelines that have been leaked; The preprocessing module is used to preprocess the leaked pipeline pressure information to organize it into a pressure detection sequence characterized by the location of the pressure leak. A pressure detection model construction module is used to construct a recurrent neural network and input the pressure detection sequence and the pressure information of the non-leaking pipeline into the recurrent neural network to obtain a trained pressure detection model. The data acquisition module is used to set up multiple pressure detection devices on the target pipeline and perform data cleaning processing on the pressure information collected by each pressure detection device to obtain cleaned pressure information. The weighting module is used to weight the pressure information after cleaning using weighting coefficients to obtain the weighted pressure information of the entire target pipeline; The detection module is used to input the weighted pressure information into the pressure detection model to obtain the detection result.

Images