Multi-parameter sensing urban underground pipe network monitoring method, system and related equipment

Abstract

The present invention discloses a multi-parameter sensing urban underground pipe network monitoring method, system and related equipment, including: obtaining the state of a manhole cover at a current time node, wherein the state of the manhole cover is characterized by whether a mechanical micro-displacement switch is in a closed state or in an open state; if the mechanical micro-displacement switch is in an open state, applying a multi-parameter monitoring instruction in real time, executing the multi-parameter monitoring instruction, and obtaining M downhole state parameters; the downhole state parameters include: the opening and closing state of the manhole cover, oxygen content, carbon monoxide content, hydrogen sulfide content, methane content, temperature, and liquid level; using the pre-processed M downhole state parameters to construct an downhole environment state vector; inputting the downhole environment state vector of the current time node into an underground pipe network monitoring model trained by an LSTM general model, and obtaining a potential risk category prediction and a risk level prediction of the underground pipe network at a future time node; the present invention can synchronously predict the risk category and risk level in the underground pipe network.

Description

Technical Field

[0001] The present application relates to the field of underground pipe network operation and maintenance, and specifically to a method, system and related equipment for monitoring urban underground pipe networks using multi-parameter sensing. Background Art

[0002] With the acceleration of urbanization, the scale of urban underground pipelines (such as water supply, drainage, natural gas, electricity, etc.) is getting larger and larger, and the complexity of their operation and maintenance management is also increasing. The operating environment of underground pipelines is complex. Affected by external factors (such as environmental pollution, gas leakage, water accumulation, etc.), underground gas concentration exceeds the standard, pipeline blockage, liquid level abnormality, etc. often occur, which may lead to safety accidents such as poisoning and explosion. Therefore, it is very important to ensure real-time monitoring and early warning of underground pipelines.

[0003] However, existing underground pipeline monitoring technologies mostly rely on regular manual inspections or single sensor monitoring. They not only lack the function of monitoring in multiple dimensions, but also lack simultaneous real-time monitoring and simultaneous early warning of risk categories and risk levels. Summary of the invention

[0004] This embodiment provides a multi-parameter sensing urban underground pipeline monitoring method, system and related equipment, and explores how to solve the problem of lack of monitoring in multiple dimensions in the existing underground pipeline monitoring, and lack of simultaneous real-time monitoring and simultaneous early warning of risk categories and risk levels.

[0005] In a first aspect, the present invention provides a multi-parameter sensing method for monitoring an urban underground pipe network, which is applied to a server side and includes:

[0006] Obtaining the state of the manhole cover at the current time node, wherein the state of the manhole cover is characterized by whether the mechanical micro-displacement switch is in a closed state or in an open state;

[0007] If the mechanical micro-displacement switch is in the disconnected state, the multi-parameter monitoring instruction is applied in real time, otherwise, the multi-parameter monitoring instruction is applied according to the preset time frequency;

[0008] Execute multi-parameter monitoring instructions to obtain M downhole status parameters;

[0009] Use the preprocessed M downhole state parameters to construct a downhole environment state vector;

[0010] The underground environment state vector at the current time node is input into the underground pipeline network monitoring model trained by the LSTM general model to obtain the potential risk category prediction and risk level prediction of the underground pipeline network at future time nodes;

[0011] The training samples of the underground pipe network monitoring model are downhole environmental state vectors at historical time nodes, and the downhole environmental state vectors are arranged in time series to form a downhole environmental state training set;

[0012] Wherein, the method for acquiring the underground environment state training set includes:

[0013] S1-1, obtaining N training samples along continuous historical time nodes in a time series; wherein the input feature of the training sample is the downhole environment state vector, and the target label of the training sample is the risk category and risk level corresponding to the downhole environment state vector;

[0014] The risk category and its risk level are used as dual target variables, and the downhole environment state vector is used as a characteristic variable to construct the training sample;

[0015] S1-2, collecting the training samples of N consecutive time nodes along the time series to construct the downhole environment state training set.

[0016] In some of the embodiments, the method for constructing the underground pipe network monitoring model further includes:

[0017] S2, train the LSTM model using the downhole environment state training set until the LSTM model converges;

[0018] S3, taking the model parameters when the convergence conditions are reached as the optimal parameters and determining them as the underground pipe network monitoring model.

[0019] In some embodiments, acquiring N training samples along consecutive historical time nodes in the time series includes:

[0020] S1-1-1, collecting M downhole state parameters at each historical time node;

[0021] S1-1-2, substituting the M underground state parameters into the risk category and level determination formula to determine the risk category and risk level of the historical time node;

[0022] S1-1-3, preprocessing M downhole state parameters;

[0023] S1-1-4, construct the M pre-processed downhole state parameters into the downhole environment state vector.

[0024] In some of the embodiments, preprocessing the M downhole state parameters includes:

[0025] Normalizing the original gas parameters including oxygen content, carbon monoxide content, hydrogen sulfide content, and methane content to obtain pre-processed gas parameters;

[0026] After the original liquid parameters including the liquid level are dimensionless, they are normalized again to obtain the pre-processed liquid parameters;

[0027] Standardize the temperature to obtain the pre-processed temperature parameters;

[0028] The pre-processed gas parameters, pre-processed liquid parameters and pre-processed temperature parameters are uniformly normalized to obtain parameter features with equal data scales.

[0029] In some embodiments, the LSTM model is trained using the downhole environment state training set until the LSTM model converges, including:

[0030] S2-1, taking the downhole environment state vector as the characteristic variable, the risk category and its risk level as the dual objective variables, and outputting the risk category prediction and risk level prediction;

[0031] The expression of the downhole environment state vector is: ;

[0032] Characterize the Mth pre-processed downhole state parameter, A downhole environment state vector constructed by M preprocessed downhole state parameters representing the t-th time node;

[0033] The expression for outputting risk category prediction and risk level prediction is:

[0034] ;

[0035] in, represents the risk category prediction, represents the risk level prediction, represents the hidden state of the node at the previous time. Indicates the memory state of the previous time node;

[0036] S2-2, use the cross entropy loss function to calculate the category prediction loss between the risk category prediction and the actual risk category in the training set; use the mean square error loss function to calculate the level prediction loss between the risk level prediction and the actual risk level in the training set;

[0037] The cross entropy loss function is: ;

[0038] in, represents the category prediction loss, C represents the number of risk categories, represents the true label of the i-th category at the t-th time node, Represents the predicted probability of the i-th category at the t-th time node;

[0039] The mean square error loss function is:

[0040] ;

[0041] in, represents the level prediction loss, N represents the number of time nodes, represents the actual risk level of the ith risk category at the tth time node, Risk level prediction for the i-th risk category at the t-th time node

[0042] S3-3, according to the category predicted loss and the level predicted loss, the total risk predicted loss is obtained;

[0043] The expression of the total loss of risk prediction is:

[0044] ;

[0045] in, represents the total loss of risk prediction, and They represent the weights of category prediction loss and level prediction loss respectively, characterizing the impact of category prediction loss and level prediction loss on the total risk prediction loss;

[0046] S3-4, use the gradient descent algorithm to update the model parameters;

[0047] The expression for updating the model parameters is:

[0048] ;

[0049] in, Indicates The model parameters of each time node, represents the model parameters at the t-th time node, represents the learning rate, Represents model parameters The gradient of the loss function;

[0050] S3-5, in the training, iteratively updating the model parameters to minimize the total loss of risk prediction; when the LSTM model reaches the convergence condition, stopping the training, and taking the model parameters at this time as the optimal parameters; wherein the convergence condition is that the loss change is less than a preset threshold or reaches the maximum number of iterations;

[0051] The expression of the optimal parameter is:

[0052] ;

[0053] in, represents the change in loss, represents the preset threshold of loss change, Indicates the current iteration number, represents the maximum number of iterations, represents the optimal parameters.

[0054] Compared with the prior art, the multi-parameter sensing urban underground pipeline monitoring method of the present invention ensures that the model has higher accuracy when simultaneously predicting risk categories and levels through a dual-objective loss function, avoids the prediction bias problem caused by single-objective optimization, and enables the model to synchronously predict risk categories and risk levels.

[0055] In a second aspect, the present invention provides a multi-parameter sensing urban underground pipe network monitoring system, comprising: a manhole cover state acquisition unit, for acquiring the manhole cover state at a current time node, wherein the manhole cover state is characterized as a mechanical micro-displacement switch being in a closed state or in an open state;

[0056] The multi-parameter monitoring instruction application unit is used to apply the multi-parameter monitoring instruction in real time if the mechanical micro-displacement switch is in an off state, otherwise, the multi-parameter monitoring instruction is applied according to a preset time frequency;

[0057] A parameter sensing unit is used to execute a multi-parameter monitoring instruction to obtain M downhole state parameters; the downhole state parameters include: the opening and closing state of the well cover, oxygen content, carbon monoxide content, hydrogen sulfide content, methane content, temperature, and liquid level;

[0058] A vector construction unit, used for constructing a downhole environment state vector using the preprocessed M downhole state parameters;

[0059] The monitoring and prediction unit is used to input the underground environment state vector of the current time node into the underground pipeline network monitoring model trained by the LSTM general model to obtain the potential risk category prediction and risk level prediction of the underground pipeline network at future time nodes;

[0060] Among them, the training samples of the underground pipe network monitoring model are downhole environmental state vectors at historical time nodes, and the downhole environmental state vectors are arranged in time series to form an downhole environmental state training set.

[0061] In a third aspect, the present invention provides an electronic device, comprising a memory and a processor, wherein the memory stores at least one computer executable instruction, and the processor is configured to run the computer executable instruction. When the computer executable instruction is executed by the processor, the multi-parameter sensing urban underground pipeline monitoring method described in the first aspect is implemented.

[0062] In a fourth aspect, the present invention provides a computer-readable storage medium having a computer program stored thereon, and when the computer program is executed by a processor, the urban underground pipe network monitoring method based on multi-parameter sensing described in the first aspect is implemented.

[0063] Compared with the prior art, the beneficial effects of the multi-parameter sensing urban underground pipe network monitoring system, electronic equipment and storage medium of the present invention are the same as the beneficial effects of the multi-parameter sensing urban underground pipe network monitoring method mentioned above, so they will not be repeated here. BRIEF DESCRIPTION OF THE DRAWINGS

[0064] Figure 1 It is a schematic diagram of the process of the urban underground pipe network monitoring method using multi-parameter sensing according to the present invention;

[0065] Figure 2 A schematic diagram of the steps for constructing an underground pipe network monitoring model of the present invention;

[0066] Figure 3 It is a structural block diagram of a multi-parameter sensing urban underground pipe network monitoring system of the present invention;

[0067] Figure 4 The figure is a schematic diagram of an electronic device of the present invention. DETAILED DESCRIPTION

[0068] In order to more clearly understand the purpose, technical solutions and advantages of the present application, the present application is described and illustrated below in conjunction with the accompanying drawings and embodiments.

[0069] Unless otherwise defined, the technical terms or scientific terms involved in this application shall have the general meaning understood by people with general skills in the technical field to which this application belongs. The words "one", "a", "the", "these" and the like in this application do not indicate a quantitative limitation, and they may be singular or plural. The terms "include", "comprise", "have" and any variants thereof involved in this application are intended to cover non-exclusive inclusions; for example, a process, method and system, product or device comprising a series of steps or modules (units) is not limited to the listed steps or modules (units), but may include unlisted steps or modules (units), or may include other steps or modules (units) inherent to these processes, methods, products or devices. The words "connect", "connected", "coupled" and the like involved in this application are not limited to physical or mechanical connections, but may include electrical connections, whether direct or indirect. The "multiple" involved in this application refers to two or more. "And / or" describes the association relationship of associated objects, indicating that there may be three relationships, for example, "A and / or B" may mean: A exists alone, A and B exist at the same time, and B exists alone. Generally, the character " / " indicates that the objects associated with each other are in an "or" relationship. The terms "first", "second", "third", etc. in this application are only used to distinguish similar objects and do not represent a specific ordering of the objects.

[0070] refer to Figures 1 to 4 In an embodiment of the present invention, a method for monitoring an urban underground pipe network using multi-parameter sensing is provided.

[0071] Figure 1 1 is a flow chart of the multi-parameter sensing urban underground pipe network monitoring method of the present invention, which includes:

[0072] Obtaining the state of the manhole cover at the current time node, wherein the state of the manhole cover is characterized by whether the mechanical micro-displacement switch is in a closed state or in an open state;

[0073] If the mechanical micro-displacement switch is in the disconnected state, the multi-parameter monitoring instruction is applied in real time, otherwise, the multi-parameter monitoring instruction is applied according to the preset time frequency;

[0074] Execute the multi-parameter monitoring instruction to obtain M downhole status parameters; the downhole status parameters include: the opening and closing status of the well cover, oxygen content, carbon monoxide content, hydrogen sulfide content, methane content, temperature, and liquid level;

[0075] Use the preprocessed M downhole state parameters to construct a downhole environment state vector;

[0076] The downhole environmental state vector of the current time node is input into the underground pipeline network monitoring model trained by the LSTM general model to obtain the potential risk category prediction and risk level prediction of the underground pipeline network at future time nodes; wherein, the training samples of the underground pipeline network monitoring model are the downhole environmental state vectors at historical time nodes, and the downhole environmental state vectors are arranged in time series to construct an downhole environmental state training set.

[0077] In this embodiment, when it is detected that the manhole cover is in a disconnected state (which may mean that the manhole cover is opened or an abnormal situation occurs), a multi-parameter monitoring instruction is applied in real time to immediately obtain the state of the underground environment, thereby preventing the expansion of potential risks. Conversely, under normal circumstances (the manhole cover is closed), the system monitors at a preset time frequency to avoid unnecessary frequent detection.

[0078] In addition, this method uses the LSTM model to predict future risk categories and risk levels, and achieve preventive maintenance and early warning. The LSTM model is good at processing time series data, and can remember and analyze the state changes of underground state parameters and risks at past time nodes, and then predict the risk categories and risk levels that may occur in the future. Therefore, this solution can effectively use past data history to make more accurate risk predictions.

[0079] Figure 2 The step of constructing the underground pipe network monitoring model in the multi-parameter sensing urban underground pipe network monitoring method of the present invention includes:

[0080] Step S1, obtaining a downhole environment state training set.

[0081] Step S1 specifically includes:

[0082] S1-1, obtaining N training samples along continuous historical time nodes in a time series; wherein the input feature of the training sample is the downhole environment state vector, and the target label of the training sample is the risk category and risk level corresponding to the downhole environment state vector;

[0083] S1-2, collecting training samples of N consecutive time nodes along the time series to construct the downhole environment state training set.

[0084] Furthermore, step S1-1 further includes:

[0085] S1-1-1, collecting M downhole state parameters at each historical time node;

[0086] S1-1-2, substituting the M underground state parameters into the risk category and level determination formula to determine the risk category and risk level of the historical time node;

[0087] The risk category level determination formula is:

[0088] ;

[0089] in, Represents the risk level score, Indicates the oxygen content, Indicates the carbon monoxide content, Indicates the hydrogen sulfide content, Indicates the methane content, Indicates temperature, Indicates the liquid level, Indicates the state of the manhole cover, 0 is closed and 1 is open; Represents the weighting of each downhole state parameter.

[0090] Furthermore, the formula also includes a risk scoring function for each downhole state parameter, specifically:

[0091] The risk score function representing the oxygen content is:

[0092] ;

[0093] The risk score function representing the carbon monoxide content is:

[0094] ;

[0095] The risk score function representing the hydrogen sulfide content is:

[0096] ;

[0097] The risk score function representing the methane content is:

[0098] ;

[0099] The risk score function representing temperature is:

[0100] ;

[0101] The risk scoring function representing the liquid level is:

[0102] ;

[0103] The risk scoring function representing the state of the manhole cover is:

[0104] ;

[0105] In addition, the risk level scoring rules are:

[0106] : Low risk (normal);

[0107] : Medium risk;

[0108] : High risk;

[0109] : Extremely high risk.

[0110] S1-1-3, preprocessing M downhole state parameters;

[0111] S1-1-4, constructing the preprocessed M downhole state parameters into the downhole environment state vector;

[0112] S1-1-5, the risk category and its risk level are used as dual target variables, and the downhole environment state vector is used as a feature variable to construct the training sample.

[0113] Furthermore, step S1-1-3 may also include:

[0114] Normalizing the original gas parameters including oxygen content, carbon monoxide content, hydrogen sulfide content, and methane content to obtain pre-processed gas parameters;

[0115] After the original liquid parameters including the liquid level are dimensionless, they are normalized again to obtain the pre-processed liquid parameters;

[0116] Standardize the temperature to obtain the pre-processed temperature parameters;

[0117] The pre-processed gas parameters, pre-processed liquid parameters and pre-processed temperature parameters are uniformly normalized to obtain parameter features with equal data scales.

[0118] Specifically, the purpose of normalizing the original gas parameters is to convert them into a fixed numerical range, usually [0,1], thereby eliminating the dimension differences between different parameters. For example, the unit of oxygen may be percentage (%), while the unit of carbon monoxide or hydrogen sulfide is ppm (parts per million). Normalization can put them in the same range for easy comparison and analysis.

[0119] The original liquid parameter (i.e. liquid level) is first dimensionally processed. Dimensionality conversion is the conversion of a physical quantity with a unit into a unitless quantity, usually by dividing it with a certain benchmark value or a specific reference standard. For example, the liquid level can be divided by a certain maximum liquid level reference value to form a unitless proportional quantity. Secondly, the dimensionally processed liquid parameters are normalized again (also compressing the data into the [0,1] interval). This step allows the liquid level parameters to be processed uniformly with other types of parameters (such as gas parameters).

[0120] The temperature is "standardized", which is different from normalization. Standardization refers to converting the parameter to a form with a mean of 0 and a standard deviation of 1. This processing method can adjust the temperature parameter to a zero-mean distribution, so that temperatures of different ranges can easily participate in subsequent model operations.

[0121] The last step is to normalize all preprocessed parameters, including gas parameters, liquid parameters, and temperature parameters, to ensure that all parameters have the same numerical scale in the final feature vector, which is convenient for subsequent model processing or training.

[0122] The process also includes:

[0123] Step S2: Use the underground environment state training set to train the LSTM model until the LSTM model converges. LSTM (Long Short-Term Memory Network) is a special type of recurrent neural network (RNN) designed specifically for processing and predicting time series-based data. Compared with traditional RNN, LSTM has significant advantages in dealing with long-term dependency problems (i.e., how to maintain and utilize earlier time series information). The core parts are: memory cells, hidden states, forget gates, input gates, and output gates.

[0124] Step S2 specifically includes:

[0125] S2-1, taking the downhole environment state vector as the characteristic variable, the risk category and its risk level as the dual objective variables, and outputting the risk category prediction and risk level prediction;

[0126] The expression of the downhole environment state vector is: , representing the M pre-processed downhole state parameters at the t-th time node;

[0127] The expression for outputting risk category prediction and risk level prediction is:

[0128] ;

[0129] in, represents the risk category prediction, represents the risk level prediction, represents the hidden state of the node at the previous time. Indicates the memory state of the previous time node;

[0130] Specifically, the hidden state and memory status It is the core of the LSTM model, which generates output through the following process:

[0131] Memory state update: The LSTM unit has a "memory cell" (i.e., memory state), which uses the "forget gate" to decide how much memory of the previous time node to retain, and uses the "input gate" to choose whether to add information of the current time node.

[0132] Hidden state update: The hidden state is used to generate the output of the model, which is updated by the current input.

[0133] Downhole environment state vector , and the updated memory state Calculate the new hidden state , the hidden state is combined with the adjustment of the "output gate" to determine the output of each time node. Combined with the new hidden state and the input of the current time node , the model generates risk category predictions and risk level prediction .

[0134] S2-2, use the cross entropy loss function to calculate the category prediction loss between the risk category prediction and the actual risk category in the training set; use the mean square error loss function to calculate the level prediction loss between the risk level prediction and the actual risk level in the training set;

[0135] The cross entropy loss function is: ;

[0136] in, represents the category prediction loss, C represents the number of risk categories, represents the true label of the i-th category at the t-th time node, Represents the predicted probability of the i-th category at the t-th time node;

[0137] The mean square error loss function is:

[0138] ;

[0139] in, represents the level prediction loss, N represents the number of time nodes, represents the actual risk level of the ith risk category at the tth time node, Risk level prediction for the ith risk category at the tth time node.

[0140] S2-3, according to the category predicted loss and the level predicted loss, the total risk predicted loss is obtained;

[0141] The expression of the total loss of risk prediction is:

[0142] ;

[0143] in, represents the total loss of risk prediction, and They represent the weights of category prediction loss and level prediction loss respectively, characterizing the impact of category prediction loss and level prediction loss on the total risk prediction loss.

[0144] S2-4, use the gradient descent algorithm to update the model parameters;

[0145] The expression for updating the model parameters is:

[0146] ;

[0147] in, Indicates The model parameters of each time node, represents the model parameters at the tth time node, represents the learning rate, Represents model parameters The gradient of the loss function;

[0148] Specifically, in this embodiment, the model parameters refer to the input gate parameters, forget gate parameters, output gate parameters, memory cell state update parameters and hidden state update parameters of the LSTM model; more specifically, the input gate, forget gate, output gate, and the corresponding weight matrix and bias term in the memory cell state; based on the LSTM model, the final hidden state is output; the final hidden state is mapped to the output space through the fully connected layer, and then converted into a probability distribution through the activation function, which is the final required risk category prediction and risk level prediction in this embodiment.

[0149] S2-5, in the training, iteratively updating the model parameters to minimize the total loss of risk prediction; when the LSTM model reaches the convergence condition, stopping the training, and taking the model parameters at this time as the optimal parameters; wherein the convergence condition is that the loss change is less than a preset threshold or reaches the maximum number of iterations;

[0150] The expression of the optimal parameter is:

[0151] ;

[0152] in, represents the change in loss, represents the preset threshold of loss change, Indicates the current iteration number, represents the maximum number of iterations, represents the optimal parameters.

[0153] The process also includes:

[0154] Step S3: The model parameters reaching the convergence condition are taken as the optimal parameters and determined as the underground pipe network monitoring model. When the model reaches the convergence condition during the training process, the acquired parameters are considered to be optimal, and then these parameters are used as the optimal parameters of the underground pipe network monitoring model for actual underground pipe network monitoring model prediction.

[0155] This embodiment uses "dual-objective prediction" to not only predict the risk category at future time nodes, but also accurately estimate the level of the risk, thereby providing more valuable prediction results for the management and preventive maintenance of underground pipe networks. Among them, dual-objective prediction expands risk assessment from a single dimension to two dimensions: risk category (such as gas leakage, liquid level abnormality, etc.) and risk level (such as low, medium, and high). It can not only determine the type of risk occurring at a specific time point, but also quantitatively assess the severity of the risk. This can provide more accurate risk information, prioritize the most urgent risk events according to risk levels, avoid resource waste, and improve the accuracy of decision-making.

[0156] In addition, this embodiment uses the cross entropy loss function to calculate the prediction error of the risk category, and uses the mean square error loss function to calculate the prediction error of the risk level; by weighted combination of these two parts of the loss, a dual-objective loss function is formed, and the gradient descent algorithm is used to iteratively optimize the model parameters, thereby improving the accuracy of the model for dual-objective predictions. The dual-objective loss function ensures that the model has higher accuracy when predicting risk categories and levels at the same time, avoids the prediction bias problem caused by single-objective optimization, and enables the model to synchronously predict risk categories and risk levels. In summary, this embodiment can perform multi-dimensional monitoring of multi-parameters of underground pipelines, and can also perform synchronous real-time monitoring and synchronous early warning of risk categories and risk levels.

[0157] The embodiment of the present invention also provides a multi-parameter sensing urban underground pipe network monitoring system, which is used to implement the above method embodiment, which has been described and will not be repeated. The terms "module", "unit", "subunit", etc. used below can implement a combination of software and / or hardware for predetermined functions. Although the system described in the following embodiments is preferably implemented in software, the implementation of hardware, or a combination of software and hardware, is also possible and conceivable.

[0158] like Figure 3As shown, Figure 3 This is a structural block diagram of a multi-parameter sensing urban underground pipe network monitoring system of the present invention, the system includes:

[0159] A manhole cover status acquisition unit, used to acquire the manhole cover status at a current time node, wherein the manhole cover status is characterized by whether the mechanical micro-displacement switch is in a closed state or in an open state;

[0160] The multi-parameter monitoring instruction application unit is used to apply the multi-parameter monitoring instruction in real time if the mechanical micro-displacement switch is in an off state, otherwise, the multi-parameter monitoring instruction is applied at a preset time frequency;

[0161] A parameter sensing unit is used to execute a multi-parameter monitoring instruction to obtain M downhole state parameters; the downhole state parameters include: the opening and closing state of the well cover, oxygen content, carbon monoxide content, hydrogen sulfide content, methane content, temperature, and liquid level;

[0162] A vector construction unit is used to construct a downhole environment state vector using the preprocessed M downhole state parameters

[0163] The monitoring and prediction unit is used to input the underground environmental state vector of the current time node into the underground pipeline network monitoring model trained by the LSTM general model to obtain the potential risk category prediction and risk level prediction of the underground pipeline network at future time nodes; wherein, the training samples of the underground pipeline network monitoring model are the underground environmental state vectors at historical time nodes, and the underground environmental state vectors are arranged in time series to construct an underground environmental state training set.

[0164] In the above system, the manhole cover status at the current time node is obtained through the manhole cover status acquisition unit; the multi-parameter monitoring instruction is applied through the multi-parameter monitoring instruction application unit; M downhole status parameters are obtained through the parameter sensing unit; the downhole environment state vector is constructed through the vector construction unit; the potential risk category prediction and risk level prediction of the underground pipeline network at future time nodes are obtained through the monitoring and prediction unit; thereby solving the problem that the prior art lacks the ability to simultaneously perform intelligent prediction of risk categories and risk levels.

[0165] The embodiment of the present invention provides related equipment for the urban underground pipe network monitoring method using multi-parameter sensing. Figure 4 As shown, an embodiment of the present invention provides an electronic device, which includes a memory 230 and a processor 210. The memory 230 stores at least one computer executable instruction. The processor 210 is configured to run the computer executable instruction. When the computer executable instruction is run by the processor 210, the above-mentioned multi-parameter sensing urban underground pipeline monitoring method is implemented.

[0166] The electronic device may include a processor 210, a communication interface 220, a memory 230, and a communication bus 240, wherein the processor 210, the communication interface 220, and the memory 230 communicate with each other through the communication bus 240. The processor 210 may call the logic instructions in the memory 230 to execute the multi-parameter sensing urban underground pipe network monitoring method disclosed in this embodiment.

[0167] In addition, the logic instructions in the above-mentioned memory 230 can be implemented in the form of software functional units and can be stored in a computer-readable storage medium when sold or used as an independent product. Based on such an understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product, which is stored in a storage medium and includes several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the methods of each embodiment of the present invention. The aforementioned storage medium includes: various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk.

[0168] The system embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, i.e., they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Those of ordinary skill in the art may understand and implement it without creative effort.

[0169] Through the description of the above implementation methods, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a necessary general hardware platform, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solution is essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0170] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A multi-parameter sensing method for monitoring urban underground pipe networks, applied to the server side, characterized in that: Obtaining the state of the manhole cover at the current time node, wherein the state of the manhole cover is characterized by whether the mechanical micro-displacement switch is in a closed state or in an open state; If the mechanical micro-displacement switch is in the disconnected state, the multi-parameter monitoring instruction is applied in real time, otherwise, the multi-parameter monitoring instruction is applied according to the preset time frequency; Execute multi-parameter monitoring instructions to obtain M downhole status parameters; Use the preprocessed M downhole state parameters to construct a downhole environment state vector; The underground environment state vector at the current time node is input into the underground pipeline network monitoring model trained by the LSTM general model to obtain the potential risk category prediction and risk level prediction of the underground pipeline network at future time nodes; The training samples of the underground pipe network monitoring model are underground environment state vectors at historical time nodes, and the underground environment state vectors are arranged in time series to form an underground environment state training set; Wherein, the method for acquiring the underground environment state training set includes: S1-1, obtaining N training samples along continuous historical time nodes in a time series; wherein the input feature of the training sample is the downhole environment state vector, and the target label of the training sample is the risk category and risk level corresponding to the downhole environment state vector; The risk category and its risk level are used as dual target variables, and the downhole environment state vector is used as a characteristic variable to construct the training sample; S1-2, collecting the training samples of N consecutive time nodes along the time series to construct the downhole environment state training set; The method for constructing the underground pipe network monitoring model also includes: S2, train the LSTM model using the downhole environment state training set until the LSTM model converges; S3, taking the model parameters when the convergence condition is reached as the optimal parameters and determining them as the underground pipe network monitoring model; The application of the underground environment state training set to train the LSTM model includes: S2-1, taking the downhole environment state vector as the characteristic variable, the risk category and its risk level as the dual objective variables, and outputting the risk category prediction and risk level prediction; The expression of the downhole environment state vector is: ; Characterize the Mth pre-processed downhole state parameter, A downhole environment state vector constructed by M preprocessed downhole state parameters representing the t-th time node; The expression for outputting risk category prediction and risk level prediction is: ； in, represents the risk category prediction, represents the risk level prediction, represents the hidden state of the node at the previous time. Indicates the memory state of the previous time node; S2-2, use the cross entropy loss function to calculate the category prediction loss between the risk category prediction and the actual risk category in the training set; use the mean square error loss function to calculate the level prediction loss between the risk level prediction and the actual risk level in the training set; The cross entropy loss function is: ; in, represents the category prediction loss, C represents the number of risk categories, represents the true label of the i-th category at the t-th time node, Represents the predicted probability of the i-th category at the t-th time node; The mean square error loss function is: ； in, represents the level prediction loss, N represents the number of time nodes, represents the actual risk level of the ith risk category at the tth time node, Risk level prediction for the i-th risk category at the t-th time node S3-3, according to the category predicted loss and the level predicted loss, the total risk predicted loss is obtained; The expression of the total loss of risk prediction is: ; in, represents the total loss of risk prediction, and They represent the weights of category prediction loss and level prediction loss respectively, characterizing the impact of category prediction loss and level prediction loss on the total risk prediction loss; S3-4, use the gradient descent algorithm to update the model parameters; The expression for updating the model parameters is: ； in, Indicates The model parameters of each time node, represents the model parameters at the tth time node, represents the learning rate, Represents model parameters The gradient of the loss function; S3-5, in the training, iteratively updating the model parameters to minimize the total loss of risk prediction; when the LSTM model reaches the convergence condition, stopping the training, and taking the model parameters at this time as the optimal parameters; wherein the convergence condition is that the loss change is less than a preset threshold or reaches the maximum number of iterations; The expression of the optimal parameter is: ； in, represents the change in loss, represents the preset threshold of loss change, Indicates the current iteration number, represents the maximum number of iterations, represents the optimal parameters.

2. The multi-parameter sensing urban underground pipe network monitoring method according to claim 1 is characterized in that: The step of acquiring N training samples along continuous historical time nodes in the time series includes: S1-1-1, collecting M downhole state parameters at each historical time node; S1-1-2, substituting the M underground state parameters into the risk category and level determination formula to determine the risk category and risk level of the historical time node; S1-1-3, preprocessing M downhole state parameters; S1-1-4, construct the M pre-processed downhole state parameters into the downhole environment state vector.

3. The multi-parameter sensing urban underground pipe network monitoring method according to claim 2 is characterized in that: The preprocessing of the M downhole state parameters includes: Normalizing the original gas parameters including oxygen content, carbon monoxide content, hydrogen sulfide content, and methane content to obtain pre-processed gas parameters; After the original liquid parameters including the liquid level are dimensionless, they are normalized again to obtain the pre-processed liquid parameters; Standardize the temperature to obtain the pre-processed temperature parameters; The pre-processed gas parameters, pre-processed liquid parameters and pre-processed temperature parameters are uniformly normalized to obtain parameter features with equal data scales.

4. A multi-parameter sensing urban underground pipe network monitoring system using the multi-parameter sensing urban underground pipe network monitoring method according to claim 1, characterized in that: include: A manhole cover status acquisition unit, used to acquire the manhole cover status at a current time node, wherein the manhole cover status is characterized by whether the mechanical micro-displacement switch is in a closed state or in an open state; A multi-parameter monitoring instruction application unit is used to apply the multi-parameter monitoring instruction in real time if the mechanical micro-displacement switch is in an off state, otherwise, the multi-parameter monitoring instruction is applied at a preset time frequency; A parameter sensing unit, used to execute multi-parameter monitoring instructions and obtain M downhole state parameters; A vector construction unit, used for constructing a downhole environment state vector using the preprocessed M downhole state parameters; The monitoring and prediction unit is used to input the underground environment state vector of the current time node into the underground pipeline network monitoring model trained by the LSTM general model to obtain the potential risk category prediction and risk level prediction of the underground pipeline network at future time nodes; The training samples of the underground pipe network monitoring model are underground environment state vectors at historical time nodes, and the underground environment state vectors are arranged in time series to form an underground environment state training set; Wherein, the method for acquiring the underground environment state training set includes: S1-1, obtaining N training samples along continuous historical time nodes in a time series; wherein the input feature of the training sample is the downhole environment state vector, and the target label of the training sample is the risk category and risk level corresponding to the downhole environment state vector; The risk category and its risk level are used as dual target variables, and the downhole environment state vector is used as a characteristic variable to construct the training sample; S1-2, collecting the training samples of N consecutive time nodes along the time series to construct the downhole environment state training set; The method for constructing the underground pipe network monitoring model also includes: S2, train the LSTM model using the downhole environment state training set until the LSTM model converges; S3, taking the model parameters when the convergence condition is reached as the optimal parameters and determining them as the underground pipe network monitoring model; The application of the underground environment state training set to train the LSTM model includes: S2-1, taking the downhole environment state vector as the characteristic variable, the risk category and its risk level as the dual objective variables, and outputting the risk category prediction and risk level prediction; The expression of the downhole environment state vector is: ; Characterize the Mth pre-processed downhole state parameter, A downhole environment state vector constructed by M preprocessed downhole state parameters representing the t-th time node; The expression for outputting risk category prediction and risk level prediction is: ； in, represents the risk category prediction, represents the risk level prediction, represents the hidden state of the node at the previous time. Indicates the memory state of the previous time node; S2-2, use the cross entropy loss function to calculate the category prediction loss between the risk category prediction and the actual risk category in the training set; use the mean square error loss function to calculate the level prediction loss between the risk level prediction and the actual risk level in the training set; The cross entropy loss function is: ; in, represents the category prediction loss, C represents the number of risk categories, represents the true label of the i-th category at the t-th time node, Represents the predicted probability of the i-th category at the t-th time node; The mean square error loss function is: ； in, represents the level prediction loss, N represents the number of time nodes, represents the actual risk level of the ith risk category at the tth time node, Risk level prediction for the i-th risk category at the t-th time node S3-3, according to the category predicted loss and the level predicted loss, the total risk predicted loss is obtained; The expression of the total loss of risk prediction is: ; in, represents the total loss of risk prediction, and They represent the weights of category prediction loss and level prediction loss respectively, characterizing the impact of category prediction loss and level prediction loss on the total risk prediction loss; S3-4, use the gradient descent algorithm to update the model parameters; The expression for updating the model parameters is: ； in, Indicates The model parameters of each time node, represents the model parameters at the tth time node, represents the learning rate, Represents model parameters The gradient of the loss function; S3-5, in the training, iteratively updating the model parameters to minimize the total loss of risk prediction; when the LSTM model reaches the convergence condition, stopping the training, and taking the model parameters at this time as the optimal parameters; wherein the convergence condition is that the loss change is less than a preset threshold or reaches the maximum number of iterations; The expression of the optimal parameter is: ； in, represents the change in loss, represents the preset threshold of loss change, Indicates the current iteration number, represents the maximum number of iterations, represents the optimal parameters.

5. An electronic device, characterized in that: It includes a memory and a processor, the memory stores at least one computer executable instruction, and the processor is configured to run the computer executable instruction. When the computer executable instruction is run by the processor, the multi-parameter sensing urban underground pipe network monitoring method according to any one of claims 1 to 3 is implemented.

6. A computer-readable storage medium, characterized in that: The storage medium stores a computer program, and when the computer program is executed by the processor, the multi-parameter sensing urban underground pipe network monitoring method according to any one of claims 1 to 3 is implemented.

Images