Intelligent management method and system for underground pipeline network integrating surveying and mapping geographic information and remote sensing

Abstract

The present invention discloses an intelligent management method and system for underground pipeline networks that integrate surveying and mapping geographic information and remote sensing, and relates to the technical field of intelligent management of underground pipeline networks. The method includes the following steps: obtaining the leakage type of the underground pipeline network and determining the pipeline aging leakage characteristic factor and the pipeline sudden leakage characteristic factor; using surveying and mapping and remote sensing to respectively obtain surveying and mapping geographic information and remote sensing data, constructing the underground pipeline network topology, and obtaining pipeline aging leakage characteristic factor data and pipeline sudden leakage characteristic factor data; combining Transformer and LSTM networks to respectively construct pipeline aging leakage prediction models and pipeline sudden leakage prediction models to obtain pipeline aging leakage prediction results and pipeline sudden leakage prediction results; using ant colony algorithm to generate inspection paths containing repair strategies to perform intelligent management of underground pipeline networks. The present invention achieves accurate analysis of leakage mechanisms, improves leakage prediction results, and thereby improves the real-time response.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent management of underground pipe networks, and in particular to an intelligent management method and system for underground pipe networks that integrates surveying and mapping geographic information and remote sensing. Background Art

[0002] With the acceleration of urbanization, underground pipeline networks, as an important part of urban lifelines, face increasingly severe challenges in their safe operation and maintenance. In recent years, the introduction of the Internet of Things, remote sensing, and artificial intelligence technologies has driven the evolution of pipeline network management towards intelligence. However, significant technical bottlenecks still exist in core aspects such as leakage risk prediction and operation and maintenance decision-making, as follows:

[0003] (1) Existing methods do not distinguish between pipeline aging leakage and sudden leakage, and use a single model to handle them uniformly. However, in practice, the physical mechanisms of the two types of leakage are significantly different: aging leakage is strongly related to soil moisture and pipe corrosion rate, while sudden leakage is mostly caused by sudden pressure changes and external construction disturbances. Therefore, existing methods use hybrid modeling for the above two types of situations, which will result in the model being unable to accurately capture differentiated characteristics, and the prediction results deviate from reality, which in turn leads to delayed response and increased operation and maintenance costs.

[0004] (2) Existing methods rely on isolated data sources, or only use GIS static topological data, or only analyze sensor time series signals, lacking the deep integration of surveying and mapping geographic information and remote sensing dynamic data. Data fragmentation leads to the loss of key environmental parameters and the limitation of model input dimensions, which in turn leads to increased prediction errors.

[0005] (3) Existing methods use a single network model to process leakage prediction, which cannot effectively model the actual spatiotemporal state of the underground pipeline network. The limitations of the model structure directly lead to low spatiotemporal resolution of the prediction results, which cannot support refined operation and maintenance. Summary of the Invention

[0006] In response to the above-mentioned deficiencies in the prior art, the present invention provides an intelligent management method and system for underground pipe networks that integrates surveying and mapping geographic information and remote sensing.

[0007] In order to achieve the above-mentioned object of the invention, the technical solution adopted by the present invention is:

[0008] The intelligent management method for underground pipe networks integrating surveying and mapping geographic information and remote sensing includes the following steps:

[0009] Collect historical leakage data of underground pipe networks, and obtain leakage types of underground pipe networks based on the historical leakage data. The leakage types of underground pipe networks include pipeline aging leakage and pipeline sudden leakage. Based on pipeline aging leakage and pipeline sudden leakage, determine pipeline aging leakage characteristic factors and pipeline sudden leakage characteristic factors respectively;

[0010] Using surveying and mapping and remote sensing to obtain surveying and mapping geographic information and remote sensing data respectively, the underground pipeline network topology structure is constructed with inspection wells as nodes and pipeline connection relationships as edges. Based on the underground pipeline network topology structure, the surveying and mapping geographic information and remote sensing data are integrated to obtain multi-source fusion data of the underground pipeline network. Based on the multi-source fusion data of the underground pipeline network, the pipeline aging leakage characteristic factor data and pipeline sudden leakage characteristic factor data are obtained;

[0011] Combining the Transformer and LSTM networks, a pipeline aging leakage prediction model and a pipeline sudden leakage prediction model were constructed respectively. The pipeline aging leakage prediction model and the pipeline sudden leakage prediction model were trained based on the historical leakage data of the underground pipeline network. The pipeline aging leakage prediction results were obtained using the trained pipeline aging leakage prediction model and pipeline aging leakage characteristic factor data, and the pipeline sudden leakage prediction results were obtained using the trained pipeline sudden leakage prediction model and pipeline sudden leakage characteristic factor data.

[0012] According to the underground pipeline network topology, pipeline aging leakage prediction results and pipeline sudden leakage prediction results, the ant colony algorithm is used to generate inspection routes including repair strategies to carry out intelligent management of the underground pipeline network.

[0013] Furthermore, the pipeline aging leakage characteristic factors include pipeline service time, pipeline corrosion rate, average soil moisture around the pipeline and the number of historical pipeline maintenance times; the pipeline sudden leakage characteristic factors include the surface deformation around the pipeline and the surface temperature change around the pipeline.

[0014] Furthermore, surveying and mapping and remote sensing are used to obtain surveying and mapping geographic information and remote sensing data respectively, and an underground pipe network topology structure is constructed with inspection wells as nodes and pipeline connection relationships as edges. The surveying and mapping geographic information and remote sensing data are fused according to the underground pipe network topology structure to obtain multi-source fusion data of the underground pipe network, including the following steps:

[0015] A1. Use surveying and mapping to obtain the location coordinates of inspection wells, pipeline connections, and pipeline attributes. Build an underground pipeline network topology using inspection wells as nodes and pipeline connections as edges. Pipeline attributes include pipeline material, service life, and historical maintenance counts.

[0016] A2. Use remote sensing to obtain time-series deformation data of the surface around the pipeline, thermal infrared remote sensing data, and microwave remote sensing data of the soil around the pipeline. Use the Kriging spatial interpolation method to match these data to the spatial accuracy of the underground pipeline network topology. Unify the timestamps of the surveyed geographic information and remote sensing data for spatiotemporal alignment. Based on the spatiotemporally aligned time-series deformation data and thermal infrared remote sensing data, obtain the surface deformation and surface temperature change around the pipeline. Invert the spatiotemporally aligned microwave remote sensing data of the soil around the pipeline to obtain the mean soil moisture around the pipeline.

[0017] A3. Construct an indirect corrosion rate model for the underground pipeline network. Calculate the pipeline corrosion rate based on the pipeline material, the average soil moisture around the pipeline, and the indirect corrosion rate model for the underground pipeline network.

[0018] A4. Based on the topological structure of the underground pipeline network, the service life of the pipeline, the pipeline corrosion rate, the average soil moisture around the pipeline, the number of historical pipeline maintenance times, the surface deformation around the pipeline, and the surface temperature change around the pipeline are integrated to obtain multi-source fusion data of the underground pipeline network.

[0019] Furthermore, in step A3, the corrosion rate is indirectly solved by the model, which is expressed as:

[0020]

[0021] in: is the corrosion rate, is the pre-exponential factor of the corrosion rate, which is related to the pipe material. is the mean soil moisture around the pipeline, is the nonlinear sensitivity coefficient of humidity to corrosion rate, is the activation energy, which is related to the pipe material. is the gas constant, is the absolute temperature of the ground around the pipeline.

[0022] Furthermore, the Transformer and LSTM networks were combined to construct pipeline aging leakage prediction models and pipeline sudden leakage prediction models, respectively. The specific process is as follows: combining the Transformer and LSTM networks, using the sequentially connected input layer, position encoding layer, Transformer encoding layer, attention weight optimization layer, LSTM decoding layer and output layer to construct pipeline aging leakage prediction models and pipeline sudden leakage prediction models, respectively;

[0023] The input layer is used to input leakage feature factor data and transmit it to the position encoding layer;

[0024] The position encoding layer is used to receive the leakage feature factor data, position encode the leakage feature factor data, and transmit the leakage feature factor data with position encoding to the Transformer encoding layer;

[0025] The Transformer encoding layer is used to perform dot product operations on the leaky feature factor data with position encoding to obtain enhanced leaky feature factor data and transmit it to the attention weight optimization layer;

[0026] The attention weight optimization layer is used to store the memory item of the enhanced leakage feature factor data, obtain the optimized attention weight coefficient based on the memory item of the enhanced leakage feature factor data and the enhanced leakage feature factor data, and transmit the enhanced leakage feature factor data and the optimized attention weight coefficient to the LSTM decoding layer;

[0027] The LSTM decoder includes a first linear layer, a long short-term memory network layer, a second linear layer, and a fully connected layer connected in sequence. The LSTM decoding layer is used to serially process the enhanced leakage feature factor data according to the optimized attention weight coefficient to obtain the leakage prediction result and transmit it to the output layer.

[0028] The output layer is used to output the leakage prediction results.

[0029] Furthermore, the data processing process of the attention weight optimization layer includes the following steps:

[0030] B1. Based on the memory item of the enhanced leakage feature factor data and the enhanced leakage feature factor data, construct the initial attention weight coefficient, which is expressed as:

[0031]

[0032] in: For the The initial attention weight coefficient of time steps, is an exponential function, For the Transformer encoding layer in The enhanced leakage characteristic factor output in the first time step is time step to enhance the memory term of the leaky feature factor data The cosine similarity of For the Transformer encoding layer in The enhanced leakage characteristic factor output in the first time step is time step to enhance the memory term of the leaky feature factor data The cosine similarity of The number of time steps for enhancing the leakage characteristic factor data;

[0033] B2. Optimize the initial attention weight coefficient to obtain the optimized attention weight coefficient, expressed as:

[0034]

[0035] in: For the The optimized attention weight coefficient of time steps, is the RELU activation function, is the attention weight constraint threshold, set it to , is a constant greater than 0.

[0036] Furthermore, the long short-term memory network layer includes an input gate substructure, a forget gate substructure, an output gate substructure, and a unit state gate substructure;

[0037] The data processing process of the input gate substructure is expressed as:

[0038] ,

[0039]

[0040] in: For the input gate substructure The output of the time step, is the sigmoid activation function, For the The optimized input gate substructure has trainable weights for time steps, is the splicing symbol, For the The hidden state of time steps, For the Transformer encoding layer in The enhanced leakage characteristic factor output in time steps, is the trainable bias of the input gate substructure, For the Candidate values ​​of the cell state for time steps, is the hyperbolic tangent function, For the The trainable weights of the candidate optimized cell states for time steps, is the trainable bias of the candidate value of the unit state;

[0041] The data processing process of the forget gate substructure is expressed as:

[0042]

[0043] in: For the forget gate structure The output of the time step, For the The trainable weights of the optimized forget gate substructure for time steps, is the splicing symbol, A trainable bias for the forget gate substructure;

[0044] The data processing process of the output door sub-mechanism is expressed as:

[0045]

[0046] in: For the output gate structure The output of the time step, For the The optimized output of the gate mechanism is the trainable weights of the time step, is the trainable bias of the output gate sub-mechanism;

[0047] The data processing process of the unit state gate substructure is expressed as:

[0048] ,

[0049]

[0050] in: For the unit state gate structure time steps of long memory output, For the unit state gate structure The output of the time step, For the unit state gate structure Time steps of short memory output.

[0051] Furthermore, the pipeline aging leakage prediction model and the pipeline sudden leakage prediction model are trained respectively based on the historical leakage data of the underground pipeline network, including the following steps:

[0052] C1. Determine the training set of pipeline aging leakage characteristic factor data and pipeline sudden leakage characteristic factor data based on the historical leakage data of the underground pipeline network;

[0053] C2. Construct the loss function model of pipeline aging leakage prediction model and pipeline sudden leakage prediction model, which is expressed as:

[0054]

[0055] in: is the total loss of the leakage prediction model, is the number of time steps in the training set, For the The actual result of leakage in time steps, For the The leakage prediction results of time steps are: is the L2 norm, is the hyperparameter of the loss function, For the The optimized number of time steps is The trainable weights of =1, 2, 3, 4, respectively represent the forget gate structure, input gate structure, unit state candidate value, output gate structure, is a logarithmic function;

[0056] C3. Based on the training set of pipeline aging leakage characteristic factor data, the training set of pipeline sudden leakage characteristic factor data, and the loss function models of the pipeline aging leakage prediction model and the pipeline sudden leakage prediction model, the pipeline aging leakage prediction model and the pipeline sudden leakage prediction model are trained respectively.

[0057] Furthermore, based on the underground pipeline network topology, pipeline aging leakage prediction results and pipeline sudden leakage prediction results, an ant colony algorithm is used to generate an inspection path containing a repair strategy to intelligently manage the underground pipeline network. The specific process is as follows: the pipeline aging leakage prediction results and the pipeline sudden leakage prediction results are weightedly fused to obtain the pipeline comprehensive leakage prediction results, and a repair strategy is formulated based on the pipeline comprehensive leakage prediction results. The inspection path data is modeled based on the pipeline comprehensive leakage prediction results and the underground pipeline network topology to generate a graph structure containing the inspection well location coordinates, pipeline distance and pipeline comprehensive leakage prediction results; the ant colony algorithm is initialized, pheromones and heuristic factors are set, the pheromones are associated with the comprehensive leakage prediction results, the heuristic factors are associated with the pipeline distance, and based on the repair strategy, an inspection path containing the repair strategy is generated.

[0058] An intelligent underground pipe network management system integrating surveying and mapping geographic information and remote sensing applied to the above method comprises a leakage analysis module, a data acquisition module, a leakage prediction module and an intelligent management module connected in sequence;

[0059] The leakage analysis module is used to collect historical leakage data of underground pipe networks and obtain the leakage types of underground pipe networks based on the historical leakage data. The leakage types of underground pipe networks include pipeline aging leakage and pipeline sudden leakage. The pipeline aging leakage characteristic factor and pipeline sudden leakage characteristic factor are determined based on the pipeline aging leakage and pipeline sudden leakage respectively.

[0060] The data acquisition module is used to use surveying and mapping and remote sensing to obtain surveying and mapping geographic information and remote sensing data respectively, to construct the underground pipe network topology structure with inspection wells as nodes and pipeline connection relationships as edges, to fuse surveying and mapping geographic information and remote sensing data based on the underground pipe network topology structure to obtain multi-source fusion data of the underground pipe network, and to obtain pipeline aging leakage characteristic factor data and pipeline sudden leakage characteristic factor data based on the multi-source fusion data of the underground pipe network;

[0061] The leakage prediction module is used to combine the Transformer and LSTM networks to respectively construct pipeline aging leakage prediction models and pipeline sudden leakage prediction models. The pipeline aging leakage prediction models and pipeline sudden leakage prediction models are trained based on the historical leakage data of the underground pipeline network. The pipeline aging leakage prediction results are obtained using the trained pipeline aging leakage prediction models and pipeline aging leakage characteristic factor data, and the pipeline sudden leakage prediction results are obtained using the trained pipeline sudden leakage prediction models and pipeline sudden leakage characteristic factor data.

[0062] The intelligent management module is used to generate an inspection route including a repair strategy based on the underground pipeline network topology, pipeline aging leakage prediction results, and pipeline sudden leakage prediction results, using an ant colony algorithm to intelligently manage the underground pipeline network.

[0063] The beneficial effects of the present invention are:

[0064] (1) The present invention determines the leakage types of underground pipeline networks, including pipeline aging leakage and pipeline sudden leakage, and determines pipeline aging leakage characteristic factors and pipeline sudden leakage characteristic factors respectively according to pipeline aging leakage and pipeline sudden leakage. Then, mapping and remote sensing are used to obtain mapping geographic information and remote sensing data respectively, and the mapping geographic information and remote sensing data are integrated to obtain multi-source fusion data of underground pipeline networks. In addition, the Transformer and LSTM networks are combined to construct pipeline aging leakage prediction models and pipeline sudden leakage prediction models respectively, so as to obtain pipeline aging leakage prediction results and pipeline sudden leakage prediction results respectively. This process realizes the accurate analysis of leakage mechanism, improves the leakage prediction results, thereby improving the real-time response and reducing operation and maintenance costs.

[0065] (2) The present invention combines surveying and mapping geographic information with remote sensing data to construct the underground pipeline network topology structure, integrates the pipeline service time, pipeline corrosion rate, average soil moisture around the pipeline, historical pipeline maintenance times, surface deformation around the pipeline, and surface temperature change around the pipeline, to obtain multi-source fusion data of the underground pipeline network, achieve deep fusion of spatial, attribute, and environmental data, and thus provide a data basis for the prediction of two types of leakage in the underground pipeline network;

[0066] (3) The present invention combines the Transformer and LSTM networks to construct a pipeline aging leakage prediction model and a pipeline sudden leakage prediction model respectively, and uses the trained pipeline aging leakage prediction model and pipeline aging leakage characteristic factor data to obtain pipeline aging leakage prediction results, and uses the trained pipeline sudden leakage prediction model and pipeline sudden leakage characteristic factor data to obtain pipeline sudden leakage prediction results, thereby improving the ability to capture time series features.

[0067] (4) By constructing an indirect corrosion rate solution model, the present invention can update the corrosion rate prediction in real time, breaking through the time and space limitations of traditional manual detection, and can realize dynamic monitoring of corrosion risks in large-scale underground pipeline networks, especially in hidden areas that are difficult to directly detect. BRIEF DESCRIPTION OF THE DRAWINGS

[0068] Figure 1 A flowchart of the intelligent management method for underground pipe networks that integrates surveying and mapping geographic information and remote sensing;

[0069] Figure 2 Schematic diagram of the structure of the underground pipeline network intelligent management system that integrates surveying and mapping geographic information and remote sensing. DETAILED DESCRIPTION

[0070] The specific embodiments of the present invention are described below to facilitate understanding of the present invention by those skilled in the art. However, it should be clear that the present invention is not limited to the scope of the specific embodiments. For those skilled in the art, as long as various changes are within the spirit and scope of the present invention as defined and determined by the appended claims, these changes are obvious, and all inventions and creations utilizing the concepts of the present invention are protected.

[0071] like Figure 1 As shown, the underground pipe network intelligent management method integrating surveying and mapping geographic information and remote sensing includes steps S1-S4, which are as follows:

[0072] S1. Collect historical leakage data of the underground pipeline network, and obtain the leakage type of the underground pipeline network based on the historical leakage data of the underground pipeline network. The leakage types of the underground pipeline network include pipeline aging leakage and pipeline sudden leakage. According to the pipeline aging leakage and pipeline sudden leakage, the pipeline aging leakage characteristic factor and the pipeline sudden leakage characteristic factor are determined respectively.

[0073] In an optional embodiment of the present invention, the historical leakage data of the underground pipeline network includes leakage type, pipeline service time, pipeline corrosion rate, average soil moisture around the pipeline, historical pipeline maintenance times, surface deformation around the pipeline, and surface temperature change around the pipeline.

[0074] Specifically, the pipeline aging leakage characteristic factors determined by the present invention include the pipeline service time, pipeline corrosion rate, average soil moisture around the pipeline and the number of historical pipeline maintenance times; the pipeline sudden leakage characteristic factors determined by the present invention include the surface deformation around the pipeline and the surface temperature change value around the pipeline.

[0075] S2. Use surveying and mapping and remote sensing to obtain surveying and mapping geographic information and remote sensing data respectively, construct the underground pipeline network topology structure with inspection wells as nodes and pipeline connection relationships as edges, fuse the surveying and mapping geographic information and remote sensing data according to the underground pipeline network topology structure to obtain multi-source fusion data of the underground pipeline network, and obtain pipeline aging leakage characteristic factor data and pipeline sudden leakage characteristic factor data based on the multi-source fusion data of the underground pipeline network.

[0076] In an optional embodiment of the present invention, the present invention utilizes surveying and mapping and remote sensing to respectively obtain surveying and mapping geographic information and remote sensing data, constructs an underground pipe network topology structure with inspection wells as nodes and pipeline connection relationships as edges, and fuses the surveying and mapping geographic information and remote sensing data according to the underground pipe network topology structure to obtain multi-source fused data of the underground pipe network, including the following steps:

[0077] A1. Use surveying and mapping to obtain the location coordinates of inspection wells, pipeline connection relationships, and pipeline attributes. Build an underground pipeline network topology with inspection wells as nodes and pipeline connection relationships as edges. Pipeline attributes include pipeline material, pipeline service time, and historical pipeline maintenance times.

[0078] A2. Use remote sensing to obtain time-series deformation data of the surface around the pipeline, thermal infrared remote sensing data, and microwave remote sensing data of the soil around the pipeline. Use the Kriging spatial interpolation method to match these data to the spatial accuracy of the underground pipeline network topology. Unify the timestamps of the surveyed geographic information and remote sensing data for spatiotemporal alignment. Based on the spatiotemporally aligned time-series deformation data and thermal infrared remote sensing data, obtain the surface deformation and surface temperature change around the pipeline. Invert the spatiotemporally aligned microwave remote sensing data of the soil around the pipeline to obtain the mean soil moisture around the pipeline.

[0079] A3. Construct an indirect corrosion rate model for the underground pipeline network. Calculate the pipeline corrosion rate based on the pipeline material, the average soil moisture around the pipeline, and the indirect corrosion rate model for the underground pipeline network.

[0080] The corrosion rate indirect solution model is expressed as:

[0081]

[0082] in: is the corrosion rate, It is the pre-exponential factor of the corrosion rate, which is related to the pipe material. When the pipe material is cast iron, the value is 0.15 mm per year. When the pipe material is low carbon steel, the value is The value range is 0.3 mm per year. When the pipe material is PE pipe The value range is 0.01 mm per year. is the mean soil moisture around the pipeline, is the nonlinear sensitivity coefficient of humidity to corrosion rate, which is obtained by fitting corrosion experiments under different humidity conditions. is the activation energy, which is related to the pipe material. is the gas constant, which is 8.314 in this invention. is the absolute temperature of the ground around the pipeline.

[0083] When determining the area surrounding the pipeline, the present invention takes the pipeline as the center and selects an area with a length of 10m and a thickness of 1m around the pipe wall as a surrounding area unit.

[0084] By constructing an indirect solution model for corrosion rate, the present invention can update the corrosion rate prediction in real time, breaking through the time and space limitations of traditional manual detection. It can realize dynamic monitoring of corrosion risks in large-scale underground pipeline networks, especially in hidden areas that are difficult to directly detect.

[0085] A4. Based on the topological structure of the underground pipeline network, the service life of the pipeline, the pipeline corrosion rate, the average soil moisture around the pipeline, the number of historical pipeline maintenance times, the surface deformation around the pipeline, and the surface temperature change around the pipeline are integrated to obtain multi-source fusion data of the underground pipeline network.

[0086] S3. Combine the Transformer and LSTM networks to construct pipeline aging leakage prediction models and pipeline sudden leakage prediction models respectively, and train the pipeline aging leakage prediction models and pipeline sudden leakage prediction models according to the historical leakage data of the underground pipeline network. Use the trained pipeline aging leakage prediction model and pipeline aging leakage characteristic factor data to obtain pipeline aging leakage prediction results, and use the trained pipeline sudden leakage prediction model and pipeline sudden leakage characteristic factor data to obtain pipeline sudden leakage prediction results.

[0087] In an optional embodiment of the present invention, the present invention combines Transformer and LSTM networks to construct pipeline aging leakage prediction models and pipeline sudden leakage prediction models, respectively. The specific process is as follows: Combining Transformer and LSTM networks, the pipeline aging leakage prediction model and pipeline sudden leakage prediction model are respectively constructed using a sequentially connected input layer, position encoding layer, Transformer encoding layer, attention weight optimization layer, LSTM decoding layer, and output layer. Both the pipeline aging leakage prediction model and the pipeline sudden leakage prediction model include a sequentially connected input layer, position encoding layer, Transformer encoding layer, attention weight optimization layer, LSTM decoding layer, and output layer.

[0088] The input layer is used to input leakage characteristic factor data and transmit it to the position encoding layer. Since the pipeline aging leakage prediction model and the pipeline sudden leakage prediction model have the same structure, the present invention simplifies the description and summarizes the pipeline aging leakage characteristic factor data and the pipeline sudden leakage characteristic factor data as leakage characteristic factor data. The present invention inputs the pipeline aging leakage characteristic factor data and the pipeline sudden leakage characteristic factor data into the input layer of the pipeline aging leakage prediction model and the input layer of the pipeline sudden leakage prediction model respectively. The pipeline aging leakage characteristic factor data includes data corresponding to the pipeline service time, pipeline corrosion rate, average soil moisture around the pipeline, and the number of historical pipeline maintenance times. The pipeline sudden leakage characteristic factor data includes data corresponding to the surface deformation around the pipeline and the surface temperature change value around the pipeline.

[0089] The position encoding layer is used to receive the leaky feature factor data, position-encode the leaky feature factor data, and transmit the position-encoded leaky feature factor data to the Transformer encoding layer. The specific processing process of the position encoding layer is as follows: calculating the odd-numbered position encoding data of the leaky feature factor data; calculating the even-numbered position encoding data of the leaky feature factor data; obtaining the position-encoded leaky feature factor data based on the leaky feature factor data, the odd-numbered position encoding data of the leaky feature factor data, and the even-numbered position encoding data of the leaky feature factor data, and transmitting it to the Transformer encoding layer.

[0090] The Transformer encoding layer is used to perform a dot product operation on the leaky feature factor data with position encoding to obtain enhanced leaky feature factor data and transmit it to the attention weight optimization layer. The Transformer encoder includes a multi-head attention mechanism, a first normalization layer, a first linear addition layer, a feedforward layer, a second normalization layer, and a second linear addition layer connected in sequence; the first linear addition layer is connected to the input of the Transformer encoder; and the feedforward layer is connected to the second linear addition layer. In the multi-head attention mechanism, the present invention performs a dot product operation on the leaky feature factor data with position encoding. The specific process is as follows: calculating the keyword vector, value vector, and query vector of the leaky feature factor data with position encoding; performing a dot product operation on the keyword vector, value vector, and query vector of the leaky feature factor data with position encoding to obtain enhanced leaky feature factor data of a single attention head; and performing parallel calculation on the enhanced leaky feature factor data of a single attention head to obtain the initial enhanced leaky feature factor data output by the multi-head attention mechanism. The initial enhanced leakage feature factor data output by the multi-head attention mechanism is then processed in sequence using the first normalization layer, the first linear addition layer, the feedforward layer, the second normalization layer, and the second linear addition layer to obtain enhanced leakage feature factor data. The feedforward layer includes two linear transformation layers connected by ReLU activation functions.

[0091] The attention weight optimization layer is used to store the memory item of the enhanced leakage feature factor data, obtain the optimized attention weight coefficient according to the memory item of the enhanced leakage feature factor data and the enhanced leakage feature factor data, and transmit the enhanced leakage feature factor data and the optimized attention weight coefficient to the LSTM decoding layer.

[0092] The data processing process of the attention weight optimization layer includes the following steps:

[0093] B1. Based on the memory item of the enhanced leakage feature factor data and the enhanced leakage feature factor data, construct the initial attention weight coefficient, which is expressed as:

[0094]

[0095] in: For the The initial attention weight coefficient of time steps, is an exponential function, For the Transformer encoding layer in The enhanced leakage characteristic factor output in the first time step is time step to enhance the memory term of the leaky feature factor data The cosine similarity of For the Transformer encoding layer in The enhanced leakage characteristic factor output in the first time step is time step to enhance the memory term of the leaky feature factor data The cosine similarity of The number of time steps for enhancing the leakage characteristic factor data.

[0096] B2. Optimize the initial attention weight coefficient to obtain the optimized attention weight coefficient, expressed as:

[0097]

[0098] in: For the The optimized attention weight coefficient of time steps, is the RELU activation function, is the attention weight constraint threshold, set it to , is a constant greater than 0.

[0099] The LSTM decoder includes a first linear layer, a long short-term memory network layer, a second linear layer, and a fully connected layer connected in sequence; the LSTM decoding layer is used to serially process the enhanced leakage feature factor data according to the optimized attention weight coefficient to obtain the leakage prediction result and transmit it to the output layer.

[0100] The long short-term memory network layer includes an input gate substructure, a forget gate substructure, an output gate substructure, and a unit state gate substructure.

[0101] The data processing process of the input gate substructure is expressed as:

[0102] ,

[0103]

[0104] in: For the input gate substructure The output of the time step, is the sigmoid activation function, For the The optimized input gate substructure has trainable weights for time steps, is the splicing symbol, For the The hidden state of time steps, For the Transformer encoding layer in The enhanced leakage characteristic factor output in time steps, is the trainable bias of the input gate substructure, For the Candidate values ​​of the cell state for time steps, is the hyperbolic tangent function, For the The trainable weights of the candidate optimized cell states for time steps, is a trainable bias for candidate values ​​of the cell state.

[0105] The data processing process of the forget gate substructure is expressed as:

[0106]

[0107] in: For the forget gate structure The output of the time step, For the The trainable weights of the optimized forget gate substructure for time steps, is the splicing symbol, is the trainable bias of the forget gate substructure.

[0108] The data processing process of the output door sub-mechanism is expressed as:

[0109]

[0110] in: For the output gate structure The output of the time step, For the The optimized output of the gate mechanism is the trainable weights of the time step, is the trainable bias of the output gate sub-mechanism.

[0111] The data processing process of the unit state gate substructure is expressed as:

[0112] ,

[0113]

[0114] in: For the unit state gate structure time steps of long memory output, For the unit state gate structure The output of the time step, For the unit state gate structure Time steps of short memory output.

[0115] Before the reconstructed LSTM decoding layer outputs the result, the present invention introduces a fully connected layer into the pipeline aging leakage prediction model and the pipeline sudden leakage prediction model to improve the robustness of the pipeline aging leakage prediction model and the pipeline sudden leakage prediction model and the reliability of the prediction data results.

[0116] The output layer is used to output the leakage prediction results.

[0117] The present invention trains a pipeline aging leakage prediction model and a pipeline sudden leakage prediction model based on historical leakage data of the underground pipeline network, including the following steps:

[0118] C1. Based on the historical leakage data of the underground pipeline network, a training set of pipeline aging leakage characteristic factor data and a training set of pipeline sudden leakage characteristic factor data are determined respectively.

[0119] C2. Construct the loss function model of pipeline aging leakage prediction model and pipeline sudden leakage prediction model, which is expressed as:

[0120]

[0121] in: is the total loss of the leakage prediction model, is the number of time steps in the training set, For the The actual result of leakage in time steps, For the The leakage prediction results of time steps are: is the L2 norm, is the hyperparameter of the loss function, For the The optimized number of time steps is The trainable weights of =1, 2, 3, 4, respectively represent the forget gate structure, input gate structure, unit state candidate value, output gate structure, is a logarithmic function.

[0122] C3. Based on the training set of pipeline aging leakage characteristic factor data, the training set of pipeline sudden leakage characteristic factor data, and the loss function models of the pipeline aging leakage prediction model and the pipeline sudden leakage prediction model, the pipeline aging leakage prediction model and the pipeline sudden leakage prediction model are trained respectively.

[0123] S4. Based on the underground pipeline network topology, pipeline aging leakage prediction results, and pipeline sudden leakage prediction results, an ant colony algorithm is used to generate an inspection path including a repair strategy to perform intelligent management of the underground pipeline network.

[0124] In an optional embodiment of the present invention, the present invention uses an ant colony algorithm to generate an inspection path including a repair strategy based on the underground pipeline network topology, pipeline aging leakage prediction results, and pipeline sudden leakage prediction results, so as to intelligently manage the underground pipeline network. The specific process is: weighted fusion of the pipeline aging leakage prediction results and the pipeline sudden leakage prediction results is performed to obtain a comprehensive pipeline leakage prediction result, and a repair strategy is formulated based on the comprehensive pipeline leakage prediction result, and inspection path data is modeled based on the comprehensive pipeline leakage prediction result and the underground pipeline network topology to generate a graph structure including the inspection well location coordinates, pipeline distance, and comprehensive pipeline leakage prediction result; the ant colony algorithm is initialized, pheromones and heuristic factors are set, pheromones are associated with the comprehensive leakage prediction results, the heuristic factors are associated with the pipeline distance, and based on the repair strategy, an inspection path including the repair strategy is generated.

[0125] The present invention initializes the ant colony algorithm, sets pheromones and heuristic factors, associates the pheromones with the comprehensive leakage prediction results, associates the heuristic factors with the pipeline distance, and generates an inspection path including the repair strategy based on the repair strategy. The specific process is as follows: initialize the ant colony algorithm, set the pheromones and heuristic factors, the higher the risk in the comprehensive leakage prediction result, the more concentrated the pheromones, thereby associating the pheromones with the comprehensive leakage prediction result, the closer the pipeline distance, the greater the heuristic value, thereby associating the heuristic factors with the pipeline distance, and establish a correlation function between the node selection probability and the pipeline distance and the comprehensive leakage prediction result to obtain the node selection probability, then the ants in the ant colony algorithm select the next inspection well according to the node selection probability to generate the inspection path, and based on the repair strategy, generate the inspection path including the repair strategy.

[0126] like Figure 2 As shown, the underground pipe network intelligent management system that integrates surveying and mapping geographic information and remote sensing and is applied to the above method includes a leakage analysis module, a data acquisition module, a leakage prediction module and an intelligent management module that are connected in sequence.

[0127] In an optional embodiment of the present invention, the leakage analysis module is used to collect historical leakage data of the underground pipeline network, and obtain the leakage type of the underground pipeline network based on the historical leakage data of the underground pipeline network. The leakage types of the underground pipeline network include pipeline aging leakage and pipeline sudden leakage. The pipeline aging leakage characteristic factor and the pipeline sudden leakage characteristic factor are determined respectively based on the pipeline aging leakage and pipeline sudden leakage.

[0128] In an optional embodiment of the present invention, the data acquisition module is used to use surveying and remote sensing to respectively obtain surveying and mapping geographic information and remote sensing data, construct an underground pipeline network topology structure with inspection wells as nodes and pipeline connection relationships as edges, fuse the surveying and mapping geographic information and remote sensing data according to the underground pipeline network topology structure to obtain multi-source fusion data of the underground pipeline network, and obtain pipeline aging leakage characteristic factor data and pipeline sudden leakage characteristic factor data based on the multi-source fusion data of the underground pipeline network.

[0129] In an optional embodiment of the present invention, the leakage prediction module is used to combine the Transformer and LSTM networks to respectively construct a pipeline aging leakage prediction model and a pipeline sudden leakage prediction model, and train the pipeline aging leakage prediction model and the pipeline sudden leakage prediction model according to the historical leakage data of the underground pipeline network, and obtain the pipeline aging leakage prediction result using the trained pipeline aging leakage prediction model and the pipeline aging leakage characteristic factor data, and obtain the pipeline sudden leakage prediction result using the trained pipeline sudden leakage prediction model and the pipeline sudden leakage characteristic factor data.

[0130] In an optional embodiment of the present invention, the intelligent management module is used to generate an inspection path including a repair strategy using an ant colony algorithm based on the underground pipeline network topology, pipeline aging leakage prediction results, and pipeline sudden leakage prediction results, so as to perform intelligent management of the underground pipeline network.

[0131] Those skilled in the art will appreciate that the embodiments described herein are intended to help readers understand the principles of the present invention, and it should be understood that the scope of protection of the present invention is not limited to such specific descriptions and embodiments. Those skilled in the art can make various other specific variations and combinations based on the technical teachings disclosed in the present invention without departing from the essence of the present invention, and such variations and combinations are still within the scope of protection of the present invention.

Claims

1. An intelligent underground pipe network management method integrating surveying and mapping geographic information and remote sensing is characterized by: The following steps are involved: Collect historical leakage data of underground pipe networks, and obtain leakage types of underground pipe networks based on the historical leakage data. The leakage types of underground pipe networks include pipeline aging leakage and pipeline sudden leakage. Based on pipeline aging leakage and pipeline sudden leakage, determine pipeline aging leakage characteristic factors and pipeline sudden leakage characteristic factors respectively; Using surveying and mapping and remote sensing to obtain surveying and mapping geographic information and remote sensing data respectively, the underground pipeline network topology structure is constructed with inspection wells as nodes and pipeline connection relationships as edges. Based on the underground pipeline network topology structure, the surveying and mapping geographic information and remote sensing data are integrated to obtain multi-source fusion data of the underground pipeline network. Based on the multi-source fusion data of the underground pipeline network, the pipeline aging leakage characteristic factor data and pipeline sudden leakage characteristic factor data are obtained; Combining the Transformer and LSTM networks, a pipeline aging leakage prediction model and a pipeline sudden leakage prediction model were constructed respectively using the sequentially connected input layer, position encoding layer, Transformer encoding layer, attention weight optimization layer, LSTM decoding layer, and output layer. The pipeline aging leakage prediction model and the pipeline sudden leakage prediction model were trained based on the historical leakage data of the underground pipeline network. The pipeline aging leakage prediction results were obtained using the trained pipeline aging leakage prediction model and pipeline aging leakage characteristic factor data, and the pipeline sudden leakage prediction results were obtained using the trained pipeline sudden leakage prediction model and pipeline sudden leakage characteristic factor data. According to the underground pipeline network topology, pipeline aging leakage prediction results and pipeline sudden leakage prediction results, the ant colony algorithm is used to generate inspection routes including repair strategies to carry out intelligent management of the underground pipeline network.

2. The intelligent management method for underground pipe networks integrating surveying and mapping geographic information and remote sensing according to claim 1 is characterized in that: The characteristic factors of pipeline aging leakage include pipeline service time, pipeline corrosion rate, average soil moisture around the pipeline and the number of historical pipeline maintenance times; the characteristic factors of pipeline sudden leakage include surface deformation around the pipeline and surface temperature change around the pipeline.

3. The intelligent management method for underground pipe networks integrating surveying and mapping geographic information and remote sensing according to claim 1 is characterized in that: Using surveying and mapping and remote sensing to obtain surveying and mapping geographic information and remote sensing data respectively, the underground pipeline network topology structure is constructed with inspection wells as nodes and pipeline connection relationships as edges. Based on the underground pipeline network topology structure, the surveying and mapping geographic information and remote sensing data are integrated to obtain multi-source fusion data of the underground pipeline network, including the following steps: A1. Use surveying and mapping to obtain the location coordinates of inspection wells, pipeline connections, and pipeline attributes. Build an underground pipeline network topology using inspection wells as nodes and pipeline connections as edges. Pipeline attributes include pipeline material, service life, and historical maintenance counts. A2. Use remote sensing to obtain time-series deformation data of the surface around the pipeline, thermal infrared remote sensing data, and microwave remote sensing data of the soil around the pipeline. Use the Kriging spatial interpolation method to match these data to the spatial accuracy of the underground pipeline network topology. Unify the timestamps of the surveyed geographic information and remote sensing data for spatiotemporal alignment. Based on the spatiotemporally aligned time-series deformation data and thermal infrared remote sensing data, obtain the surface deformation and surface temperature change around the pipeline. Invert the spatiotemporally aligned microwave remote sensing data of the soil around the pipeline to obtain the mean soil moisture around the pipeline. A3. Construct an indirect corrosion rate model for the underground pipeline network. Calculate the pipeline corrosion rate based on the pipeline material, the average soil moisture around the pipeline, and the indirect corrosion rate model for the underground pipeline network. A4. Based on the topological structure of the underground pipeline network, the service life of the pipeline, the pipeline corrosion rate, the average soil moisture around the pipeline, the number of historical pipeline maintenance times, the surface deformation around the pipeline, and the surface temperature change around the pipeline are integrated to obtain multi-source fusion data of the underground pipeline network.

4. The intelligent management method for underground pipe networks integrating surveying and mapping geographic information and remote sensing according to claim 3 is characterized in that: In step A3, the corrosion rate is solved indirectly by the model, which is expressed as: Where: C rate is the corrosion rate, A is the pre-exponential factor of the corrosion rate, which is related to the pipe material, S mois is the mean soil moisture around the pipeline, m is the nonlinear sensitivity coefficient of humidity to corrosion rate, E is the activation energy, which is related to the pipeline material, R is the gas constant, and T is the absolute temperature of the ground around the pipeline.

5. The underground pipe network intelligent management method integrating surveying and mapping geographic information and remote sensing according to claim 1 is characterized in that: The input layer is used to input leakage feature factor data and transmit it to the position encoding layer; The position encoding layer is used to receive the leakage feature factor data, position encode the leakage feature factor data, and transmit the leakage feature factor data with position encoding to the Transformer encoding layer; The Transformer encoding layer is used to perform dot product operations on the leaky feature factor data with position encoding to obtain enhanced leaky feature factor data and transmit it to the attention weight optimization layer; The attention weight optimization layer is used to store the memory item of the enhanced leakage feature factor data, obtain the optimized attention weight coefficient based on the memory item of the enhanced leakage feature factor data and the enhanced leakage feature factor data, and transmit the enhanced leakage feature factor data and the optimized attention weight coefficient to the LSTM decoding layer; The LSTM decoder includes a first linear layer, a long short-term memory network layer, a second linear layer, and a fully connected layer connected in sequence. The LSTM decoding layer is used to serially process the enhanced leakage feature factor data according to the optimized attention weight coefficient to obtain the leakage prediction result and transmit it to the output layer. The output layer is used to output the leakage prediction results.

6. The intelligent management method for underground pipe networks integrating surveying and mapping geographic information and remote sensing according to claim 5 is characterized in that: The data processing process of the attention weight optimization layer includes the following steps: B1. Based on the memory item of the enhanced leakage feature factor data and the enhanced leakage feature factor data, construct the initial attention weight coefficient, which is expressed as: Where: W t is the initial attention weight coefficient at the tth time step, exp is the exponential function, d(X t ,r j ) is the memory item r of the enhanced leakage feature factor output by the Transformer encoding layer at the tth time step and the enhanced leakage feature factor data at the jth time step j The cosine similarity of d(X t ,r t ) is the enhanced leakage feature factor output by the Transformer encoding layer at the tth time step and the memory item r of the enhanced leakage feature factor data at the tth time step t The cosine similarity of , N is the number of time steps of the enhanced leakage feature factor data; B2. Optimize the initial attention weight coefficient to obtain the optimized attention weight coefficient, expressed as: Where: W t g is the optimized attention weight coefficient at the tth time step, RELU(,) is the RELU activation function, φ is the attention weight constraint threshold, which is set to [1 / N,3 / N], and λ is a constant greater than 0.

7. The intelligent management method for underground pipe networks integrating surveying and mapping geographic information and remote sensing according to claim 6 is characterized in that: The long short-term memory network layer includes an input gate substructure, a forget gate substructure, an output gate substructure, and a unit state gate substructure; The data processing process of the input gate substructure is expressed as: Among them: I t is the output of the input gate substructure at the tth time step, σ is the sigmoid activation function, is the trainable weight of the gate structure after optimization at the tth time step, [·,·] is the splicing symbol, H t-1 is the hidden state of the t-1th time step, X t is the enhanced leakage feature factor output by the Transformer encoding layer at the tth time step, b2 is the trainable bias of the input gate substructure, is the candidate value of the cell state at the tth time step, tanh is the hyperbolic tangent function, is the trainable weight of the optimized unit state candidate value at the tth time step, b3 is the trainable bias of the unit state candidate value; The data processing process of the forget gate substructure is expressed as: Where: Y t is the output of the forget gate substructure at the tth time step, is the trainable weight of the forget gate substructure after optimization at the tth time step, [·,·] is the concatenation symbol, and b1 is the trainable bias of the forget gate substructure; The data processing process of the output door sub-mechanism is expressed as: Among them: t is the output of the output gate substructure at the tth time step, is the trainable weight of the output gate sub-mechanism after optimization at the tth time step, b4 is the trainable bias of the output gate sub-mechanism; The data processing process of the unit state gate substructure is expressed as: O t =o t ·tanh(C t ) Where: C t is the long memory output of the cell state gate substructure at the tth time step, C t-1 is the output of the cell state gate substructure at the t-1th time step, O t is the short memory output of the cell state gate substructure at the tth time step.

8. The intelligent management method for underground pipe networks integrating surveying and mapping geographic information and remote sensing according to claim 1 is characterized in that: The pipeline aging leakage prediction model and pipeline sudden leakage prediction model are trained based on the historical leakage data of the underground pipeline network, including the following steps: C1. Determine the training set of pipeline aging leakage characteristic factor data and pipeline sudden leakage characteristic factor data based on the historical leakage data of the underground pipeline network; C2. Construct the loss function model of pipeline aging leakage prediction model and pipeline sudden leakage prediction model, which is expressed as: Where: L total is the total loss of the leakage prediction model, Z is the number of time steps in the training set, and x j is the actual leakage result at the jth time step, is the leakage prediction result at the jth time step, is the L2 norm, α is the hyperparameter of the loss function, is the trainable weight numbered i after optimization at the jth time step, i = 1, 2, 3, 4, representing the forget gate substructure, input gate substructure, unit state candidate value, and output gate substructure, respectively, and log is the logarithmic function; C3. Based on the training set of pipeline aging leakage characteristic factor data, the training set of pipeline sudden leakage characteristic factor data, and the loss function models of the pipeline aging leakage prediction model and the pipeline sudden leakage prediction model, the pipeline aging leakage prediction model and the pipeline sudden leakage prediction model are trained respectively.

9. The intelligent management method for underground pipe networks integrating surveying and mapping geographic information and remote sensing according to claim 1 is characterized in that: Based on the underground pipeline network topology, pipeline aging leakage prediction results, and pipeline sudden leakage prediction results, an ant colony algorithm is used to generate an inspection path containing a repair strategy for intelligent management of the underground pipeline network. The specific process is as follows: the pipeline aging leakage prediction results and the pipeline sudden leakage prediction results are weightedly fused to obtain the pipeline comprehensive leakage prediction results, and a repair strategy is formulated based on the pipeline comprehensive leakage prediction results. The inspection path data is modeled based on the pipeline comprehensive leakage prediction results and the underground pipeline network topology to generate a graph structure containing the inspection well location coordinates, pipeline distance, and pipeline comprehensive leakage prediction results; the ant colony algorithm is initialized, pheromones and heuristic factors are set, the pheromones are associated with the comprehensive leakage prediction results, the heuristic factors are associated with the pipeline distance, and based on the repair strategy, an inspection path containing the repair strategy is generated.

10. An intelligent underground pipe network management system integrating surveying and mapping geographic information and remote sensing, applied to the method according to any one of claims 1 to 9, characterized in that: It includes a leakage analysis module, a data acquisition module, a leakage prediction module and an intelligent management module connected in sequence; The leakage analysis module is used to collect historical leakage data of underground pipe networks and obtain the leakage types of underground pipe networks based on the historical leakage data. The leakage types of underground pipe networks include pipeline aging leakage and pipeline sudden leakage. The pipeline aging leakage characteristic factor and pipeline sudden leakage characteristic factor are determined based on the pipeline aging leakage and pipeline sudden leakage respectively. The data acquisition module is used to use surveying and mapping and remote sensing to obtain surveying and mapping geographic information and remote sensing data respectively, to construct the underground pipe network topology structure with inspection wells as nodes and pipeline connection relationships as edges, to fuse surveying and mapping geographic information and remote sensing data based on the underground pipe network topology structure to obtain multi-source fusion data of the underground pipe network, and to obtain pipeline aging leakage characteristic factor data and pipeline sudden leakage characteristic factor data based on the multi-source fusion data of the underground pipe network; The leakage prediction module is used to combine the Transformer and LSTM networks to respectively construct pipeline aging leakage prediction models and pipeline sudden leakage prediction models. The pipeline aging leakage prediction models and pipeline sudden leakage prediction models are trained based on the historical leakage data of the underground pipeline network. The pipeline aging leakage prediction results are obtained using the trained pipeline aging leakage prediction models and pipeline aging leakage characteristic factor data, and the pipeline sudden leakage prediction results are obtained using the trained pipeline sudden leakage prediction models and pipeline sudden leakage characteristic factor data. The intelligent management module is used to generate an inspection route including a repair strategy based on the underground pipeline network topology, pipeline aging leakage prediction results, and pipeline sudden leakage prediction results, using an ant colony algorithm to intelligently manage the underground pipeline network.

Images