A method and system for integrated sky, space and ground pipeline network security monitoring

Abstract

The present invention belongs to the technical field of pipeline leakage detection, and particularly relates to a method and system for integrated air-space-ground pipeline network safety monitoring. The method includes the following steps: S1. Obtain infrared thermal imaging data of the pipeline network in the area to be detected, pipeline crack distribution data of the pipeline network, and real-time pressure data of the pipeline head and end stations, and perform integrated air-space-ground multi-source data collection; S2. Establish a pipeline network risk assessment and early warning model based on the obtained data, perform numerical simulation calculations based on the pipeline network risk assessment and early warning model, delimit risk points and risk areas, and determine leakage points. The present invention adopts a "joint air-space-ground" monitoring method, builds a real-time pipeline network leakage monitoring system, and has relatively high monitoring accuracy. Among them, the measurement accuracy of the infrared thermal imager is millimeter-level, the crack monitoring accuracy of the unmanned aerial vehicle is centimeter-level, and the monitoring accuracy of the distributed optical fiber pressure sensor is micron-level.

Description

Technical Field

[0001] The present invention belongs to the technical field of pipeline leakage detection, and particularly relates to an integrated space-air-ground pipeline network safety monitoring method and system. Background Art

[0002] At present, pipeline transportation is the main way to transport crude oil. With the increase in pipeline construction, especially the growth of pipelines, during the transportation of crude oil, pipeline leakage accidents have become a serious social problem. Pipeline leakage mainly includes natural aging or accidental damage of pipelines and man-made damage. Pipeline leakage will not only pollute the environment, but also cause losses to people's lives and property. Pipeline defects are characterized by being difficult to predict, sudden, and highly destructive. Especially if a leakage accident occurs in a pipeline near a city, it will pose a threat to the lives of most people and result in serious resource losses. Therefore, how to prevent pipeline leakage has become a technical problem.

[0003] There are various existing anti-leakage detection devices, but the existing detection devices have relatively single means, and it is difficult to accurately detect leakage points with such a single detection means. Especially when a detection component is damaged, it is even more difficult to detect leakage points, which brings great difficulties to manual investigation. Summary of the Invention

[0004] In view of the above-mentioned shortcomings and deficiencies of the prior art, the present invention provides an integrated space-air-ground pipeline network safety monitoring method and system. The present invention solves the technical problems that there is no effective real-time monitoring technology for hidden defects in the current pipeline network, there is a lack of universal leakage criteria, and it is impossible to timely discover the leakage danger areas of the pipeline network. The present invention can realize multi-source data fusion monitoring and analysis, and conduct real-time monitoring and early warning of pipeline defects.

[0005] In order to achieve the above object, the main technical solutions adopted by the present invention include:

[0006] In the first aspect, an embodiment of the present invention provides an integrated space-air-ground pipeline network safety monitoring method, and the method includes the following steps:

[0007] S1. Obtain the infrared thermal imaging data of the pipeline network in the area to be detected, the pipeline crack distribution data of the pipeline network, and the real-time pressure data of the pipeline head and end stations, and perform integrated space-air-ground multi-source data collection;

[0008] S2. Establish a pipeline network risk assessment and early warning model based on the obtained integrated space-air-ground multi-source data, perform numerical simulation calculations based on the pipeline network risk assessment and early warning model, delimit risk points and risk areas, and determine leakage points.

[0009] Further, in the step S1, the specific steps for obtaining the infrared thermal imaging data of the pipeline network in the area to be detected are:

[0010] 1) Collect data from the infrared thermal imager for the area to be detected to obtain the pipeline network leakage image distribution data;

[0011] 2) Use an anomaly detection method based on a dual autoencoder and a transformation network to determine whether there is a leakage in the pipeline network;

[0012] 3) When a leakage occurs on the pipeline, the infrared thermal imager can present the image based on the abnormal temperature situation, and at the same time, the information detected by the infrared thermal imager processor is transmitted to the central processor of the monitoring terminal for processing.

[0013] Furthermore, in the step S1, the specific steps for obtaining the pipeline network pipeline crack distribution data are as follows:

[0014] 1) Use a drone to take a digital orthophoto map of the pipeline network;

[0015] 2) Use a crack recognition algorithm to extract the pipeline crack information of a single image, and obtain the length and width of the single crack image;

[0016] 3) Complete the crack splicing at the edges of multiple images to obtain the length, width, and spatial position information of the overall pipeline network cracks;

[0017] 4) Conduct multi-dimensional analysis on the obtained pipeline crack length, width, and spatial position information, calculate the rate and distribution characteristics of crack development, establish a mapping relationship between crack development and pipeline service strength, and evaluate the safety status of the pipeline network pipeline based on the mapping relationship between pipeline crack information and pipeline service life.

[0018] Furthermore, in the step S1, the specific steps for obtaining the real-time pipeline head and end station pressure data are as follows:

[0019] 1) Place pressure sensors at the pipeline head and end stations respectively to monitor and obtain the pressure change information of the pipeline head and end stations in real time;

[0020] 2) Use the obtained pressure change information of the pipeline head and end stations to construct a negative pressure wave signal denoising model based on a residual noise reduction autoencoder;

[0021] 3) Conduct leakage location analysis on the obtained denoised pressure data, and calculate the pipeline leakage point using the negative pressure wave location formula;

[0022] 4) During the process of dealing with the leakage source, take different countermeasures according to the size of the leakage situation.

[0023] Furthermore, in the step S2, the specific steps for establishing a pipeline network risk assessment and early warning model are as follows:

[0024] 1) According to the obtained pipeline network pipeline crack distribution data, divide the suspected leakage area and the normal area;

[0025] 2) Divide the suspected leakage area and the normal area according to the obtained infrared thermal imaging data of the pipeline network;

[0026] 3) Extract the pressure data of the head and end stations of the suspected leakage pipeline, conduct leakage detection, and obtain the location of the suspected leakage point.

[0027] Further, in the step S2, the specific steps of numerical simulation calculation based on the pipeline network risk assessment and early warning model are as follows:

[0028] 1) Use an anomaly detection method based on a dual autoencoder and a transformation network to determine whether there is a leakage in the pipeline network;

[0029] 2) Use a crack recognition algorithm to extract the pipeline crack information of a single image, and obtain the length and width of the single crack image;

[0030] 3) Conduct leakage location analysis on the obtained denoised pressure data, and calculate the pipeline leakage point using the negative pressure wave location formula.

[0031] In a second aspect, an embodiment of the present invention provides an air-space-ground integrated pipeline network safety monitoring system, including an oil pipeline, a plurality of pressure sensors sequentially installed on the head and end stations of the oil pipeline, an unmanned aerial vehicle camera, and further including an infrared thermal imager for detecting abnormal temperature conditions of the oil pipeline. The infrared thermal imager is arranged on one side of the oil pipeline, and the relative position is located between adjacent on-line monitoring points. The detection information of the pressure sensors, the unmanned aerial vehicle camera, and the infrared thermal imager can all be processed and analyzed by a data processing center, and the detection results are output to determine the location of the leakage point and potential risk points on the oil pipeline.

[0032] The beneficial effects of the present invention are as follows: The present invention provides an air-space-ground integrated pipeline network safety monitoring method and system, which adopts an "air-space-ground" joint monitoring method to build a real-time pipeline network leakage monitoring system with high monitoring accuracy. Among them, the measurement accuracy of the infrared thermal imager is millimeter-level, the crack monitoring accuracy of the unmanned aerial vehicle is centimeter-level, and the monitoring accuracy of the distributed optical fiber pressure sensor is micron-level.

[0033] The present invention realizes the monitoring time coordination of pipeline network defects. Different technical means are used in different stages to achieve continuous observation throughout the process. The DATN-ND method is used to identify and screen pipeline liquid leakage. UAVs are used to collect pipeline network images, and then the crack identification algorithm (FDDNet) is used to identify pipeline cracks. Distributed fiber optic pressure sensors are used to monitor the pressure data at the head and end stations of the pipeline in real time to accurately locate the leakage point, making the monitoring method adaptable to the normal operation of the pipeline network; it realizes the spatial coordinated measurement of the pipeline leakage monitoring point-line-surface combination, taking into account both the inside and outside. The linear measurement results monitored by the distributed fiber optic pressure sensors are combined with the planar measurement by UAVs, achieving the combination of the internal operating conditions of the pipeline detected by the distributed fiber optic and various pipeline surface defect detections; it realizes the parameter coordination of pipeline network defect detection. By obtaining the pipeline leakage liquid distribution data, crack distribution data, and pipeline internal pressure change data, the intelligent analysis of pipeline defects is realized. Based on the pipeline defect mechanism, the analytic hierarchy process is used to establish the weight information of multi-source monitoring data such as leakage liquid distribution, crack distribution, and pressure data to predict pipeline defects, which can efficiently and accurately evaluate the safety status of the pipeline network and provide real-time early warning of pipeline defect risks.

[0034] The present invention divides the early warning and control area, effectively monitors the location of the leakage source, and ensures the safe operation of the pipeline network. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] Figure 1 is the flow chart of an air-space-ground integrated pipeline network safety monitoring method of the present invention;

[0036] Figure 2 is the structural diagram of an air-space-ground integrated pipeline network safety monitoring system of the present invention;

[0037] Figure 3 is the flow chart of liquid leakage identification in the present invention;

[0038] Figure 4 is the flow chart of crack identification in the present invention;

[0039] Figure 5 is the flow chart of de-noising the negative pressure wave signal in the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0040] In order to better explain the present invention for easy understanding, the present invention will be described in detail below with reference to the accompanying drawings through specific embodiments.

[0041] Embodiment 1:

[0042] As Figure 1 shown, the present invention provides an air-space-ground integrated pipeline network safety monitoring method. The method includes the following steps:

[0043] S1. Obtain the infrared thermal imaging data of the pipeline network in the area to be detected, the pipeline crack distribution data of the pipe network, and the real-time pressure data of the pipeline head and end stations, and perform integrated multi-source data acquisition of space-air-ground;

[0044] S2. Establish a pipe network risk assessment and early warning model based on the obtained integrated multi-source data of space-air-ground, perform numerical simulation calculations based on the pipe network risk assessment and early warning model, delimit risk points and risk areas, and determine leakage points.

[0045] The present invention organically integrates the infrared thermal imaging for monitoring the liquid leakage distribution of the pipe network, the unmanned aerial vehicle aerial photogrammetry for detecting the development of pipeline cracks, and the distributed optical fiber pressure sensors at the pipeline head and end stations for detecting the internal pressure changes of the pipeline, realizing the integrated detection and analysis of multi-source data. It solves the technical problems that there is no effective real-time monitoring technology for the hidden defects of the current pipe network, lacks a universal leakage criterion, and cannot timely discover the leakage dangerous areas of the pipe network.

[0046] Specifically, when there is crude oil leakage on the oil pipeline, the infrared thermal imager can present the image according to the abnormal temperature condition and transmit the detected information to the central processor of the monitoring terminal through the infrared thermal imager processor for processing, adopting the anomaly detection method based on dual autoencoders and transformation network (DATN-ND). As Figure 3 shown, the model of DATN-ND consists of two parallel autoencoders and a transformation network. First, the proposed transformation network generates the bottleneck features with abnormal data information (referred to as pseudo-abnormal bottleneck features) through the hidden layer feature representation (referred to as bottleneck features) encoded by the input samples, so as to increase the abnormal data information in the training set; secondly, the dual autoencoders reconstruct the bottleneck features with abnormal data information as normal data as much as possible instead of reconstructing themselves, so that the abnormal data obtains a reconstruction error that is quite different from that of the normal data.

[0047] The algorithm implementation process of the anomaly detection method based on dual autoencoders and transformation network in the training stage and the testing stage is as follows. The training steps of DATN-ND:

[0048] Input: Training set

[0049] Output: Encoder f E (·) and decoder f D (·);

[0050] Initialization: Encoder f E (·), decoder f D (·) and transformation network f T (·) parameter set θ E , θ D , θ T .

[0051] For i = 1 to N do:

[0052] Obtain the training sample x through Equation 1: z = f E (x); i After encoding, obtain the bottleneck feature z;

[0053] Obtain the bottleneck feature z' after z is transformed by the transformation network through Equation 2: z t = f T (x); t ;

[0054] Through Equation 3: Obtain the t and and

[0055] Through the above Equation 1: z = f E (x) to calculate and obtain and the bottleneck features after re - encoding and

[0056] Through Equation 4: Randomly update the encoder, decoder, and the network parameter set θ E , θ D , θ T .

[0057] Algorithm 4: Anomaly Detection Based on DATN - ND

[0058] Input the test sample x test , the encoder f E (·), the decoder f D (·), and the threshold value.

[0059] Output the anomaly score S(x test ) and its category of x test .

[0060] Utilize Equation 5: to obtain the test sample x test and the reconstructed sample

[0061] Utilize Equation 6: to calculate the anomaly score S(x test ) of the test sample. If S(x test ) ≥ σ, it is concluded that the abnormal data is x test .

[0062] Specifically, when the digital orthophoto map of the pipeline network captured by the UAV camera transmits the image data to the central processor of the monitoring terminal for processing, the crack recognition algorithm (FDDNet) is used to identify the distribution of pipeline cracks and mark them. For example, Figure 4 As shown, the crack defect detection network is modified based on YOLOv5 (a single-stage object detection algorithm). The SPD-Conv module is added to the backbone network of YOLOv5. The SPD-Conv mainly consists of the Space-to-depth (SPD) layer and the non-strided convolution layer. Through this downsampling method, the problem of a large amount of missing fine-grained information of the target can be solved. Finally, a small target detection head is added to the detection head, which can make the network pay more attention to the detection of small targets, improve the detection effect, and at the same time, in order to improve the detection speed of the network, the lightweight network G-GhostNet is introduced into the backbone network.

[0063] In this embodiment, the industrial production environment has a computer CPU of i5-12500H, a GPU of RTX3060, a memory of 6G, uses the language of python3.7, and the deep learning framework is pytorch1.8. The pipeline image is input into the crack defect detection network for pipeline crack detection. A total of 24,633 pipeline crack images are used, of which 22,170 images are used as the training set and 2,463 images are used as the validation set. The parameter settings of the crack defect detection network are as follows: The optimizer selected is Stochastic Gradient Descent (SGD), the momentum is 0.9, the initial learning rate is 1e-2, the weight decay is set to 5e-4, and the learning rate decay method is selected as cos. The specific parameters of the backbone network are as follows in the table:

[0064] Operator Input Output SPD-Conv Conv 3×3 <![CDATA[640 2 ×3]]> <![CDATA[320 2 ×3]]> √ Conv 3×3 <![CDATA[320 2 ×3]]> <![CDATA[160 2 ×128]]> √ Block <![CDATA[160 2 ×128]]> <![CDATA[160 2 ×128]]> Block×3Cheap <![CDATA[160 2 ×128]]> <![CDATA[160 2 ×128]]> Concat <![CDATA[160 2 ×128]]> <![CDATA[160 2 ×128]]> Conv 3×3 <![CDATA[160 2 ×128]]> <![CDATA[80 2 ×256]]> √ Block <![CDATA[80 2 ×256]]> <![CDATA[80 2 ×256]]> Block×6Cheap <![CDATA[80 2 ×256]]> <![CDATA[80 2 ×256]]> Concat <![CDATA[80 2 ×256]]> <![CDATA[80 2 ×256]]> Conv 3×3 <![CDATA[80 2 ×256]]> <![CDATA[40 2 ×512]]> √ Block <![CDATA[40 2 ×512]]> <![CDATA[40 2 ×512]]> Block×9Cheap <![CDATA[40 2 ×512]]> <![CDATA[40 2 ×512]]> Concat <![CDATA[40 2 ×512]]> <![CDATA[40 2 ×512]]> Conv 3×3 <![CDATA[40 2 ×512]]> <![CDATA[20 2 ×1024]]> √ Block <![CDATA[20 2 ×1024]]> <![CDATA[20 2 ×1024]]> Block×3Cheap <![CDATA[20 2 ×1024]]> <![CDATA[20 2 ×1024]]> Concat <![CDATA[20 2 ×1024]]> <![CDATA[20 2 ×1024]]> SPPF <![CDATA[20 2 ×1024]]> <![CDATA[20 2 ×1024]]>

[0065] In the above table, Block represents the traditional residual network, Input represents the size of the input image, Output represents the size of the output image. After the feature extraction of the backbone network, the detection results can be marked on the image through the detection head.

[0066] The crack defect detection network can solve the problem of a large amount of missing fine-grained information of small targets through the downsampling method. Finally, a small target detection head is added to the detection head, which can make the network pay more attention to the detection of small targets, improve the detection effect, and thus determine whether the pipeline leaks.

[0067] If there is a possibility of pipeline leakage, the negative pressure wave signal detected by the pressure sensor is denoised based on the denoising autoencoder, and the leakage point is located using the negative pressure wave positioning formula.

[0068] In the embodiments of the present invention, since the propagation speed of the negative pressure wave signal is very fast, even a very small time error will cause a large positioning error. Therefore, accurately measuring the time difference between the negative pressure wave signals detected by the sensors when reaching the first and last stations is the key to effectively locating leaks. In an industrial production environment, noise will have an adverse impact on the arrival time of the negative pressure wave signal. Therefore, signal processing must be carried out first to denoise the negative pressure wave signal.

[0069] In the embodiments of the present invention, in order to improve the denoising accuracy of the negative pressure wave signal, a deep learning algorithm based on a residual denoising autoencoder is adopted. As Figure 5 shown, the negative pressure wave signal is input into the denoising autoencoder to obtain the denoised negative pressure wave signal; the time difference between the negative pressure wave signals received by the sensors at the first and last stations is determined through batch normalization (Batch Normalization), and the specific location of the pipeline leak is detected through the positioning formula.

[0070] Simulated signals or low-noise signals collected in an ideal experimental environment are generated, and these signals are used to construct the training set and test set of the deep learning algorithm.

[0071] Specifically, in the embodiments of the present invention, the amplitude of the time series of the negative pressure wave signal is used as a characteristic parameter, and a negative pressure wave pure signal without interference and a noise profile signal are constructed. The two are added to obtain a negative pressure wave noisy signal. The length of the negative pressure wave pure signal without interference is 330,000, and the amplitude is between -1000 and 1000; the sum of the noise profile signal and the negative pressure wave pure signal without interference is used as the negative pressure wave noisy signal. The obtained noisy signal and noise profile signal are divided into a training set and a test set according to a ratio of 7:3 for standby. Then, the constructed training set has a length of 231,000, and the test set has a length of 99,000.

[0072] Determine the number of network layers of the autoencoder and train layer by layer using the training data.

[0073] First, the calculation of the C-channel feature maps output after the first transposed convolution operation is:

[0074] z 1 =w' 0 ⊙z 0 +b' 0

[0075] Among them, b' 0 and w' 0 are respectively the bias and the transposed convolution kernel in the transposed convolution, ⊙ is the transposed convolution operation, and z 0 is the feature vector after the 5th convolution reconstruction. z 1 is the feature vector V with a length of C obtained after passing through the global average pooling layer. The feature vector V is specifically expressed as:

[0076]

[0077] Among them, H and W respectively represent the height and width of the input signal.

[0078] Then, it passes through 2 fully connected layers and directly adds the result to the residual block transmitted from the convolutional layer. The output is expressed as:

[0079]

[0080] Among them, respectively represent the weight values of the first fully connected layer and the second fully connected layer, σ represents the activation operation using the sigmoid activation function, and e 4 is the residual block transmitted after passing through the fourth convolutional layer.

[0081] Then, it is activated using the ReLU activation function:

[0082] V' 1 = ReLU(V 1 )

[0083] Finally, it passes through 1 transposed convolution and ReLU and is expressed as:

[0084] z' 0 = ReLU(w' 1 ⊙V 1 '+ b' 1 )

[0085] Among them, b' 1 and w' 1 are respectively the bias and the transposed convolution kernel of the second transposed convolution layer in the attention decoder, and z' 0 is the input of the second transposed convolution residual block in the attention decoder. After the attention decoding process, the denoised negative pressure wave signal is obtained.

[0086] Referring to Figure 2 , after obtaining the denoised negative pressure wave signal, the output of the residual denoising autoencoder is passed through batch normalization (Batch Normalization) to extract the inflection point information of the denoised negative pressure wave signal features, calculate the arrival time difference of the upstream and downstream negative pressure waves, and finally calculate the accurate leakage point position L 1 , and the negative pressure wave positioning formula is:

[0087]

[0088] Among them, L is the total length of the pipeline from the head station to the end station, Δt is the time difference for the negative pressure wave to propagate to the head station and the end station, v is the wave speed of the negative pressure wave, t 1 is the duration for the negative pressure wave to reach the head station, t2 The duration to reach the terminal station.

[0089] In the DATN-ND network and FDDNet network in the embodiments of the present invention, they have the advantages of high detection accuracy and fast detection speed. The residual noise reduction autoencoder constructs a training model for negative pressure wave noisy signals, which can accurately predict the noise signal profile, achieve the purpose of noise elimination, ensure accurate extraction of the time difference, and the error of the negative pressure wave velocity correction formula is small. This method can fundamentally ensure the accuracy of the pipeline leakage real-time positioning method, has important theoretical and practical guiding values, and can also meet the effective detection under different pipeline operating conditions. It is a pipeline leakage real-time online positioning method with a brand-new idea.

[0090] Embodiment 2:

[0091] As Figure 2 shown, an air-space-ground integrated pipeline network safety monitoring system provided by the present invention. The system includes an oil pipeline 1, a pressure sensor 2, a drone camera 3 sequentially installed on the oil pipeline 1, and also includes an infrared thermal imager 4 for detecting the abnormal temperature condition of the oil pipeline 1. The infrared thermal imager 4 is arranged on one side of the oil pipeline 1 and its relative position is located between adjacent online monitoring points. The detection information of the pressure sensor 2, the drone camera 3, and the infrared thermal imager 4 can be processed and analyzed through a data processing center (processed according to a conventional software system) and then the detection result is output to determine the location L of the leakage point on the pipeline 1 and potential risk points.

[0092] The above are only the preferred embodiments of the present invention and are not used to limit the present invention. For those skilled in the art, the present invention can have various changes and modifications. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included within the protection scope of the present invention.

Claims

1. An integrated space-air-ground pipeline network safety monitoring method, characterized in that, the method comprises the following steps: S1. Obtain the infrared thermal imaging data of the pipeline network in the area to be detected, the pipeline crack distribution data of the pipeline network, and the real-time pressure data of the pipeline head and tail stations, and perform integrated space-air-ground multi-source data acquisition; S2. Establish a pipeline network risk assessment and early warning model based on the obtained integrated space-air-ground multi-source data, perform numerical simulation calculations based on the pipeline network risk assessment and early warning model, delimit risk points and risk areas, and determine leakage points; In the step S1, the specific steps for obtaining the infrared thermal imaging data of the pipeline network in the area to be detected are: 1) Collect data from the infrared thermal imager in the area to be detected to obtain the pipeline network leakage image distribution data; 2) Use the anomaly detection method based on the dual autoencoder and transformation network to judge whether there is leakage in the pipeline network; 3) When leakage occurs on the pipeline, the infrared thermal imager can present the image according to the abnormal temperature situation, and at the same time, transmit the detected information to the central processor of the monitoring terminal through the infrared thermal imager processor for processing; In the step S1, the specific steps for obtaining the pipeline crack distribution data of the pipeline network are: 1) Use an unmanned aerial vehicle to photograph to obtain the digital orthophoto map of the pipeline network; 2) Use the crack recognition algorithm to extract the pipeline crack information of a single image, and obtain the length and width of the single crack image; 3) Complete the crack splicing at the edges of multiple images to obtain the length, width, and spatial position information of the overall pipeline network cracks; 4) Perform multi-dimensional analysis on the obtained pipeline crack length, width, and spatial position information, calculate the rate and distribution characteristics of crack development, establish the mapping relationship between crack development and pipeline service strength, and evaluate the safety status of the pipeline network according to the mapping relationship between pipeline crack information and pipeline service life; In the step S1, the specific steps for obtaining the real-time pressure data of the pipeline head and tail stations are: 1) Place pressure sensors at the pipeline head and tail stations respectively to monitor and obtain the pressure change information of the pipeline head and tail stations in real time; 2) Use the obtained pressure change information of the pipeline head and tail stations to construct a negative pressure wave signal denoising model based on the residual noise reduction autoencoder; 3) Perform leakage location analysis on the obtained denoised pressure data, and calculate the pipeline leakage point using the negative pressure wave location formula; 4) During the process of disposing the leakage source, take different countermeasures according to the size of the leakage situation.

2. The integrated space-air-ground pipeline network safety monitoring method according to claim 1, characterized in that, in the step S2, the specific steps for establishing the pipeline network risk assessment and early warning model are: 1) According to the obtained pipeline crack distribution data of the pipeline network, divide the suspected leakage area and the normal area; 2) According to the obtained infrared thermal imaging data of the pipeline network, divide the suspected leakage area and the normal area; 3) Extract the pressure data of the head and tail stations of the suspected leakage pipeline for leakage detection to obtain the position of the suspected leakage point.

3. The integrated space-air-ground pipeline network safety monitoring method according to claim 1, characterized in that, in the step S2, the specific steps for performing numerical simulation calculations based on the pipeline network risk assessment and early warning model are: 1) An anomaly detection method based on a dual autoencoder and a transformation network is used to determine whether there is a leak in the pipeline network; 2) A crack recognition algorithm is used to extract the pipeline crack information from a single image, and the length and width of the single crack image are obtained; 3) Leak location analysis is performed on the obtained denoised pressure data, and the pipeline leak point is calculated using the negative pressure wave location formula.

4. An integrated space-air-ground pipeline network safety monitoring system monitors according to the integrated space-air-ground pipeline network safety monitoring method described in claim 1, characterized in that: It includes: An oil pipeline, and a plurality of pressure sensors installed at the head and end stations of the oil pipeline in sequence, an unmanned aerial vehicle camera, and an infrared thermal imager for detecting abnormal temperature conditions of the oil pipeline. The infrared thermal imager is placed on one side of the oil pipeline, and the relative position where it is placed is between adjacent on-line monitoring points. The detection information of the pressure sensor, the unmanned aerial vehicle camera, and the infrared thermal imager can all be processed and analyzed by a data processing center, and the detection result is output, so as to judge the location of the leak point and potential risk points on the oil pipeline.

Images