A dual-temperature and dual-pressure based leakage fault diagnosis method for heating secondary pipe networks

Abstract

A dual-temperature, dual-pressure, dual-heating secondary pipe network leakage fault diagnosis method, belonging to the field of smart heating technology, comprises the following steps: constructing a hydraulic and thermal simulation model of the secondary heating pipe network based on actual secondary heating pipe network parameters; coupling the hydraulic simulation model with the thermal simulation model to construct an actual heating system simulation model; constructing hydraulic and thermal data sets of the heating pipe network under different fault conditions based on the heating system simulation model; training a BP neural network leakage fault diagnosis model based on dual-pressure data and an LSTM neural network leakage fault diagnosis model based on dual-temperature time series data based on the data sets; and calling the trained leakage fault diagnosis model to perform fault diagnosis on leakage in the secondary heating pipe network. The present invention can quickly and accurately locate the pipe section where the leakage occurs, allowing for timely repair, thereby ensuring the safe operation of the heating system.

Description

Technical Field

[0001] The present invention relates to a pipeline leakage diagnosis method, and in particular to a dual-temperature and dual-pressure based leakage fault diagnosis method for a heating secondary pipe network, belonging to the field of smart heating technology. Background Art

[0002] The rapid development of smart heating technology has enriched the number of sensors in secondary heating pipe networks. Some highly informatized secondary heating pipe networks are equipped with dual-temperature and dual-pressure sensors at the heat inlet of heat users, providing the foundation for data-driven diagnosis of pipe network faults. However, related research still has certain shortcomings:

[0003] (1) Research on fault diagnosis of heating pipe networks mainly focuses on macroscopic heating systems or primary pipe network failures, and lacks research on secondary pipe networks.

[0004] (2) Some studies did not conduct hydraulic and thermal modeling of the pipeline network to be diagnosed, but only identified anomalies based on individual sensor data, lacking an overall consideration of the dual temperature and dual pressure changes at multiple nodes under pipeline network fault conditions.

[0005] In view of the above, it is still impossible to accurately diagnose pipeline leakage in order to adapt to timely maintenance and ensure the safe operation of the heating system. Summary of the Invention

[0006] In order to overcome the shortcomings of the prior art, the present invention provides a dual-temperature and dual-pressure based leakage fault diagnosis method for a heating secondary pipe network. This application constructs a hydraulic and thermal operating condition model and a coupled simulation model of a heating secondary pipe network, which includes heat source, pipe section, heat user and node models, based on the actual parameters of the heating secondary pipe network. Based on the coupled simulation model, the normal operating conditions and fault conditions of the secondary heating network in different heating stages are simulated to construct a multi-operating condition data set. Based on the data set, a pressure-based BP neural network model and an LSTM neural network model based on temperature time series data are trained respectively. Finally, a fault diagnosis model is obtained that can identify the location of pipe network leakage through the dual-temperature and dual-pressure data of the pipe network, thereby improving the accuracy of the model.

[0007] A dual-temperature and dual-pressure based leakage fault diagnosis method for a heating secondary pipe network comprises the following steps:

[0008] S1. Construct a hydraulic and thermal simulation model of the secondary heating pipe network based on the actual parameters of the secondary heating pipe network;

[0009] S2. Couple the hydraulic simulation model with the thermal simulation model to construct a simulation model of the actual heating system;

[0010] S3. Constructing hydraulic and thermal data sets of the heating network when different fault conditions occur in the heating network based on the heating system simulation model;

[0011] S4, training a BP neural network leakage fault diagnosis model based on dual-pressure data and a LSTM neural network leakage fault diagnosis model based on dual-temperature time series data according to the data set;

[0012] S5. Obtain a section of time series data on water replenishment to detect whether there is a leak in the pipe network. When the water replenishment volume of the heating pipe network does not exceed the threshold, there is no leak. When the water replenishment volume of the heating pipe network continuously exceeds the threshold, a pipe network leakage alarm signal is sent. At this time, the data of the dual-temperature and dual-pressure sensors of the heating system are collected and input into the trained dual-temperature and dual-pressure data leakage fault diagnosis model to obtain the faulty pipe section where the leak occurs, and complete the leakage fault diagnosis of the secondary heating pipe network.

[0013] Compared with the prior art, the present invention has the following advantages:

[0014] 1. This application, based on a data-driven approach, fully utilizes the abundant dual-temperature and dual-pressure data from the secondary heating pipe network to construct a pipe network leak diagnosis model. This model can quickly and accurately locate the leaking pipe section and conduct timely repairs. This ensures the safe and economical operation of the heating system and creates positive social and economic benefits.

[0015] 2. The hydraulic and thermal coupled simulation model proposed in this application can fully and accurately reflect the dynamic thermal and hydraulic characteristics of the secondary heating network. It can be used to conduct targeted operating simulation and fault diagnosis model construction for different secondary heating networks, improving the accuracy of the model.

[0016] 3. This application performs leakage diagnosis based on dual-pressure and dual-temperature time series data respectively. The two complement each other to improve the accuracy and applicability of diagnosis. Among them, the leakage diagnosis model based on dual-pressure data responds quickly, and the diagnostic accuracy is limited by the accuracy of the dual-pressure sensor. The leakage diagnosis model based on temperature time series data requires a certain response time and has a higher diagnostic accuracy.

[0017] The present application will be further described below with reference to the accompanying drawings and implementation methods: BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 This is a flow chart of the dual-temperature and dual-pressure heating secondary pipe network leakage fault diagnosis method of this application;

[0019] Figure 2 This is a flow chart of constructing a heating system simulation model in step S2;

[0020] Figure 3 is the directed connectivity graph of the heating system pipe network;

[0021] Figure 4 Schematic diagram of simulated pipeline leakage when building a multi-case dataset;

[0022] Figure 5 It is the BP neural network diagram in the leakage fault diagnosis model;

[0023] Figure 6 Schematic diagram of the LSTM unit structure in the leakage fault diagnosis model;

[0024] Figure 7 This is a schematic diagram of a heating network in the embodiment;

[0025] Figure 8 This is a graph showing the dynamic temperature change of a heat user in the embodiment;

[0026] Figure 9 BP neural network and LSTM neural network fault diagnosis model training process diagram in the embodiment;

[0027] Figure 10 The confusion matrix diagram of the BP fault diagnosis model on the test set in the embodiment;

[0028] Figure 11 This is a confusion matrix diagram of the LSTM fault diagnosis model on the test set in the embodiment;

[0029] Figure 12 This is a flowchart of the fault diagnosis model application in the embodiment. DETAILED DESCRIPTION

[0030] The following is a detailed description of the embodiments of the technical solution of the present application in conjunction with the accompanying drawings. Unless otherwise specified, the technical terms or scientific terms used in this application are generally understood by those skilled in the art.

[0031] This example constructs a coupled simulation model of the hydraulic and thermal conditions of a secondary heating network based on actual secondary heating network parameters, including models of heat sources, pipe segments, heat users, and nodes. Based on this coupled simulation model, the network's normal and faulty operating conditions during different heating phases are simulated to construct a multi-condition dataset. Based on these datasets, a pressure-based BP neural network model and an LSTM neural network model based on temperature time series data are trained, resulting in a fault diagnosis model capable of identifying leak locations using dual-temperature and dual-pressure data.

[0032] Specifically, a dual-temperature and dual-pressure based leakage fault diagnosis method for a heating secondary pipe network is provided, which comprises the following steps:

[0033] S1. Construct a hydraulic and thermal simulation model of the secondary heating pipe network based on the actual parameters of the secondary heating pipe network;

[0034] When constructing a hydraulic simulation model, the following assumptions must be made when performing hydraulic calculations:

[0035] ① The flow velocity, pressure and other parameters in any cross section of the pipeline are evenly distributed, that is, the flow of fluid in the pipe is simplified to a one-dimensional process.

[0036] ②. The fluid in the tube is incompressible.

[0037] Based on the node flow balance equation, loop pressure balance equation, pipeline resistance loss equation, and pump head characteristic equation, a steady-state hydraulic simulation model equation group is constructed. The steady-state hydraulic model is solved according to the basic loop method to obtain the flow, pressure, pipe section pressure drop, and flow of each node.

[0038]

[0039] Where: B represents the number of branches in the heating network;

[0040] N represents the number of heating network nodes;

[0041] A represents the incidence matrix, N-1×B dimensions;

[0042] B f represents the basic loop matrix, B-N+1×B dimension;

[0043] G represents the pipeline flow column vector, B dimension, kg / s;

[0044] Q represents the column vector of the node's net inflow, that is, the flow exchanged between the node and the outside world of the system, with inflow being positive and outflow being negative, (N-1) dimension, kg / s;

[0045] ΔH represents the pipeline pressure drop column vector, B dimension, Pa;

[0046] Z represents the pipeline potential energy difference column vector, B dimension, Pa;

[0047] S represents the pipeline impedance matrix, B×B dimensions, Pa·s 2 / kg 2 ;

[0048] ΔH pump represents the pump head column vector, B dimension, Pa;

[0049] C0, C1, and C2 represent the diagonal matrices of water pump characteristic parameters, with B×B dimensions.

[0050] Constructing a thermal simulation model: When a heating system is adjusted, the response time of unsteady hydraulic processes is much shorter than the response time of thermal processes. Therefore, the hydraulic changes in the pipe network can be viewed as a series of quasi-steady-state processes. Approximate instantaneous flow and pressure information of the pipe network can be obtained through a steady-state hydraulic simulation model, and this information can be used to analyze the dynamic thermal conditions of the pipe network. The node method requires the following assumptions when performing thermal calculations:

[0051] ① The temperature distribution in any cross section of the tube is uniform, which means that there is no radial heat transfer inside the fluid.

[0052] ②. Ignore the axial heat conduction effect of the fluid.

[0053] ③. For the supply and return pipe sections, the thermal resistance between the fluid in the pipe, the pipe wall, the soil and the atmospheric environment is evenly distributed in the flow direction of the pipe.

[0054] 1) Build a pipe network model

[0055] According to the law of conservation of energy, the temperature change of the fluid flowing in the pipeline can be expressed by formula (2). Based on (2), the time term is taken as a backward difference, and the space term is taken as a first-order upwind difference to obtain formula (3);

[0056]

[0057] Where: ρ represents the density of the fluid in the tube, kg / m 3 ;

[0058] T represents the temperature of the fluid in the tube, °C;

[0059] T env Indicates outdoor ambient temperature, ℃;

[0060] u represents the flow velocity in the pipe, m / s;

[0061] c p represents the specific heat capacity of the fluid in the tube, J / (kg·℃);

[0062] t, Δt represents time, unit time length, s;

[0063] x, Δx represents the pipe length, unit pipe length, m;

[0064] A c represents the cross-sectional area of ​​the pipe, m 2 ;

[0065] d represents the inner diameter of the pipe, m;

[0066] λ represents the thermal conductivity of the fluid in the tube, W / (m·℃);

[0067] R represents the equivalent heat transfer resistance per unit length between the fluid and the pipe wall-soil-atmosphere environment, m·℃ / W;

[0068] u (n) represents the velocity of the flow in the pipe at time n, m / s;

[0069] represents the outdoor ambient temperature at time n, ℃;

[0070] represents the temperature of the ith discrete node at time n after the network is discretized, ℃;

[0071] 2) Build a hot user model

[0072] The heating building is simplified as a mass point with lumped parameters. The indoor temperature of each building is assumed to be an average temperature. The building exchanges heat with the radiator and the outside. The heat capacity of the equivalent heating building is concentrated at the mass point, and the formula (4) is listed. The backward difference method is used for the time term to obtain the indoor temperature of the building, as shown in formula (5). The heat dissipation of the heat dissipation equipment is shown in formula (6), where the inlet and outlet water temperatures of the heat dissipation equipment are equivalent to the supply and return water temperatures of the secondary pipe network. Combining formulas (4) to (6), the indoor temperature changes and supply and return water temperatures of the heat users can be calculated.

[0073]

[0074] Where: C b It represents the equivalent total heat capacity of the building, J / ℃;

[0075] T n Indicates the average indoor temperature of the building, °C;

[0076] represents the average indoor temperature of the building at time n, ℃;

[0077] Q rad , G rad Indicates the heat dissipation and flow rate of the radiator, W;

[0078] They represent the heat dissipation and flow rate of the radiator at time n, W;

[0079] K b Indicates the comprehensive heat transfer coefficient of the building envelope, W / ℃;

[0080] K rad Indicates the equivalent heat transfer coefficient of the radiator, W / ℃;

[0081] η rad Indicates the flow time of the fluid in the radiator, which can be taken as a unit time step, η rad =1;

[0082] T su , T re Indicates the inlet and outlet water temperatures of the heat dissipation equipment, ℃;

[0083] Respectively represent heat dissipation devices n-η rad Inlet water temperature at time n, outlet water temperature at time n (equivalent to the supply and return water temperature of the node), °C;

[0084] 3) Build a heat source model

[0085] The heat source is the starting point for the dynamic calculation of the heating network, and its thermal characteristics are the boundary conditions of the dynamic simulation model. They can be obtained in the following way: given the heat source at each moment, that is, the outlet water temperature of the secondary pipe network side of the heat exchange station The outlet water temperature is controlled to reach the set value through the automatic control system of the heat exchange station;

[0086] 4) Build pipeline node model

[0087] According to the energy conservation equation, the node temperature calculation formula (7) is obtained

[0088]

[0089] Where: T i represents the temperature of node i, °C;

[0090] T c,j represents the terminal temperature of pipe j flowing into node i, °C;

[0091] T s，k represents the starting temperature of pipe k flowing out of node i, °C;

[0092] Φ start,i represents the set of all pipes flowing out of node i;

[0093] Φ end，i represents the set of all pipes flowing into node i.

[0094] S2. Couple the hydraulic simulation model with the thermal simulation model to construct a simulation model of the actual heating system;

[0095] Reference Figure 2-Figure 4 , based on hydraulic simulation and thermal model theory, an integrated method is proposed, Figure 2 The simulation process of the actual heating system modeling is demonstrated. The hydraulic model is coupled with the thermal model. The thermal models of the heat source, pipe section, and heat user are connected in a certain order. The temperature of each node is calculated using the node model to build a complete simulation model of the actual heating system, that is, the hydraulic and thermal coupling simulation model. Specifically:

[0096] S21. Based on the design information of the heating pipe network, the node information and pipe segment information of the heating pipe network and the connection relationship are input into the computer. Based on the graph theory, a directed connectivity graph of the heating pipe network is generated. Figure 3As shown in the figure, based on graph theory, the heating system is abstracted into a directed connection graph consisting of two types of elements: pipes and nodes. The spatial network topology of the secondary heating network is established. The heating network is a closed-loop flow system. Each branch represents each section of the heating pipe. The node is the connection point of the heating pipe. Each node is connected to at least two pipes, and there is at most one branch between two nodes. The nodes of the secondary heating network are represented by the set N consisting of {n1, n2, n3, ...}. There are N heating network nodes in total. The branch set is B = {n1-n2, n2-n3, n3-n4, ...}. There are B heating network branches in total.

[0097] S22, setting the time step, space step and simulation period of the discrete solution, and setting the initial value of the iteration according to the measured data or the design value;

[0098] S23, using a breadth-first traversal algorithm to determine the calculation order of the pipeline;

[0099] S24. Based on the hydraulic simulation model of step S1, iteratively calculate the hydraulic parameters of the heating pipe network at time t, and provide flow parameters of each node and pipe section for thermal calculation;

[0100] S25. Based on the thermal simulation model of step S2, the system is divided into three parts: heat source, heat user, and heating pipe network. First, the heat source model is called to determine the water supply temperature of the heat source at the current moment. According to the pipe network calculation sequence, the pipe network model is called to perform dynamic thermal calculations on each water supply pipe segment in sequence starting from the heat source to obtain the outlet temperature of the pipe, and the node temperature is calculated according to formula (7). Then, the heat user model is called to obtain the temperature and flow of the return water node. According to the pipe network calculation sequence, the pipe network model is called to calculate the outlet temperature of each return water pipe segment.

[0101] S3. Constructing hydraulic and thermal data sets of the heating network when different fault conditions occur in the heating network based on the heating system simulation model;

[0102] The simulation model constructed based on the above method can accurately reflect the operating status of the heating system. At this time, the hydraulic and thermal characteristics dataset of the heating network under different fault conditions is constructed through the hydraulic and thermal coupling simulation model.

[0103] Further use machine learning methods to train the leakage diagnosis model based on the dataset:

[0104] First, when constructing the leakage condition dataset, it is necessary to add a new node nleak to the normal pipe network topology structure to represent the specific location where the leakage occurs. Figure 4 As shown in the figure, the original pipeline n2-n4 assumed to be leaking is split into two new pipelines n2-nleak and nleak-n4, and the other pipelines remain unchanged.

[0105] The main parameters of the heating pipeline are pipe length l, pipe diameter d, impedance S and heat transfer coefficient per unit pipe length K p , before and after the leaking pipe is split into two new pipe sections, its pipe diameter d and unit pipe length heat transfer coefficient K p The pipe length l and the impedance S remain unchanged, while the pipe length l and the impedance S are redistributed to two new pipe sections according to the location of the leak point, and then:

[0106] l2=l0-l1 (8)

[0107] l1=μl0 (9)

[0108] S1=μS0 (10)

[0109] S2=S0-S1 (11)

[0110] Where l0 represents the length of the original pipe section, m;

[0111] l1, l2 represent the length of each pipe segment after splitting, m;

[0112] μ represents the ratio of the leakage point to the total length of the original pipe section, which is dimensionless;

[0113] S0 represents the impedance of the original pipe section, Pa·s 2 / kg 2 ;

[0114] S1 and S2 represent the impedance of each pipe segment after splitting, Pa·s 2 / kg 2 .

[0115] In view of the above, the process of constructing the hydraulic and thermal data sets of the heating pipeline network when different fault conditions occur is as follows: first, the pipeline segments are selected according to the actual situation of the heating pipeline network. Multiple different leakage locations μ = {μ1, μ2, μ3, μ4, ...} are set for each pipeline segment. Multiple leakage amounts are set for each leakage location for simulation calculation to obtain node pressure data and temperature time series data, where each data is a vector containing the values ​​of multiple pressure or temperature measurement points. The faulty pipeline segment is encoded with a unique hot label to complete the construction of the fault diagnosis dataset.

[0116] S4, training a BP neural network leakage fault diagnosis model based on dual-pressure data and a LSTM neural network leakage fault diagnosis model based on dual-temperature time series data according to the data set;

[0117] In this step, the pipeline network leakage diagnosis model training uses machine learning methods to mine the potential fault characteristics contained in multi-node dual-temperature and dual-pressure, and trains and obtains pipeline network leakage diagnosis models based on pressure data and temperature time series data respectively;

[0118] Among them, the pipeline network fault diagnosis model based on the BP neural network of dual pressure data is trained. The BP neural network structure is as follows: Figure 5 As shown, the BP neural network leakage fault diagnosis model is a multi-layer neural network structure, where the number of neurons in the input layer is x n Equal to the number of thermal users equipped with dual pressure sensors for thermal inlet, the number of neurons in the output layer y n The model takes the dual-pressure data of heat users as the input value of the neural network and the probability of failure of each pipeline section to be diagnosed as the output value.

[0119] Among them, the LSTM neural network leakage diagnosis model based on dual-temperature time series data is trained. The LSTM unit structure is as follows Figure 6 As shown, the LSTM neural network leakage fault diagnosis model is a multi-layer neural network structure, where the number of neurons in the input layer is x n Equal to the number of thermal users equipped with dual temperature sensors at the thermal inlet, the number of neurons in the output layer y n The model takes the dual-temperature time series data of heat users as input variables and the probability of failure of each pipe section to be diagnosed as the output value.

[0120] S5. Obtain a section of time series data on water replenishment to detect whether there is a leak in the pipe network. When the water replenishment volume of the heating pipe network does not exceed the threshold, there is no leak. When the water replenishment volume of the heating pipe network continuously exceeds the threshold, a pipe network leakage alarm signal is sent. At this time, the data of the dual-temperature and dual-pressure sensors of the heating system are collected and input into the trained dual-temperature and dual-pressure data leakage fault diagnosis model to obtain the faulty pipe section where the leak occurs, and complete the leakage fault diagnosis of the secondary heating pipe network.

[0121] In this step, for the actual secondary heating network, after the network leakage diagnosis model is constructed according to the above method, the model can be used to perform fault diagnosis based on dual-temperature and dual-pressure data. The network is detected to see if there is a leak by monitoring the water replenishment volume of the heating system. The leakage threshold can be set as 2% of the circulating water volume of the secondary heating network. When the water replenishment volume of the heating network exceeds the threshold continuously, a network leakage alarm signal is sent. At this time, the dual-temperature and dual-pressure sensor data of the heating system are collected and input into the corresponding trained fault diagnosis model to obtain the faulty pipe section where the leak occurs. The overall diagnosis process is as follows: Figure 1 shown.

[0122] The present application will be further described below with reference to the following embodiments:

[0123] Taking an actual secondary heating network in Beijing as an example, the specific implementation of this method is as follows:

[0124] The specific implementation process can be divided into four steps: constructing hydraulic and thermal simulation models of the secondary heating pipeline network and a coupling-based simulation model of the actual heating system, constructing a multi-condition data set for the pipeline network, constructing a pipeline network leakage diagnosis model, and applying the fault diagnosis model.

[0125] Step 1: Build hydraulic, thermal and coupled simulation models for the secondary heating network

[0126] A typical secondary heating system in Beijing was selected as the research object, and its design parameters are shown in Table 1;

[0127] Table 1 Heating network design parameters

[0128]

[0129] like Figure 7 As shown in Figure 1, the district heating network consists of 23 supply pipes, 23 return pipes, and 48 nodes. The largest pipe diameter is DN400 and the smallest pipe diameter is DN100. The heat source is equivalent to a pipe segment in the heating network, numbered n1'-n1, as shown in Figure 1. Figure 4 As shown, the system is equipped with a water replenishment constant pressure device at node n1' to maintain a constant pressure of 0.2 MPa under design conditions.

[0130] Based on the hydraulic and thermal simulation model of the secondary heating pipe network, a hydraulic and thermal coupling simulation model is constructed to obtain a simulation model of the actual heating system, which accurately reflects the dynamic thermal and hydraulic characteristics of the heating pipe network.

[0131] Figure 8 This simulation shows the dynamic temperature of the heating system for seven days, from January 1st to January 7th, during the mid-term heating period. Heat user N11 is used as an example. It can be seen that the heat exchange station adjusts the water supply temperature in response to changes in ambient temperature, resulting in a temperature close to that of the heating station. The indoor temperature fluctuates around 20°C, with a smaller fluctuation than the ambient temperature, and its trend lags slightly behind the ambient temperature.

[0132] Step 2: Constructing a multi-condition dataset for the pipe network

[0133] A total of 23 main pipes n1-n24 and their adjacent branch pipes were selected, as shown in Table 2. Nine different leakage locations μ = {0.1, 0.2, 0.3, ..., 0.9} were set for each pipe segment, and 10 leakage rates were simulated and calculated for each leakage location.

[0134] The time step Δt is set to 10 minutes and the spatial step Δx is set to 10 meters. A large amount of steady-state pressure data and temperature time series data are obtained according to the simulation results. Each piece of data is a vector containing the pressure or temperature values ​​of multiple measuring points.

[0135] Table 2 Leakage setting of simulation data set

[0136]

[0137] Step 3: Pipeline network leakage diagnosis model construction

[0138] (1) BP neural network leakage diagnosis model based on dual pressure data

[0139] An LSTM pipeline fault diagnosis model based on dual-temperature time series data was built and trained. This model has a multi-layer neural network structure. The dual-temperature time series of heat users after a leak was used as the neural network input, resulting in 12 neurons in the input layer. The number of pipeline sections to be diagnosed, i.e., the number of neurons in the output layer, was 23. The hidden layer had 2 layers, with 128 neurons in the first layer, followed by a Sigmoid activation function; the second layer had 64 neurons, followed by a ReLU activation function. The cross-entropy loss was selected as the loss function, and the Adam method was selected as the optimization function. Finally, the Softmax function was used to convert the final output into the probability of pipeline section failure.

[0140] The pressure dataset is randomly divided into training set, validation set and test set, which account for 60%, 20% and 20% of the dataset respectively. When the pressure accuracy is 0.001MPa, the training process is as follows Figure 9 The model with the highest accuracy on the validation set is used as the final model obtained through training. At this point, the accuracy of the model in the test set is 86.23%, and the confusion matrix of its diagnosis results is as follows: Figure 10 shown.

[0141] (2) LSTM neural network leakage diagnosis model based on dual-temperature time series data

[0142] An LSTM pipeline network fault diagnosis model based on dual-temperature time series data was built and trained. This model has a multi-layer neural network structure. The dual-temperature time series of heat users after a leak is used as the neural network input. The number of neurons in the input layer is 12. The number of pipeline sections to be diagnosed, i.e., the number of neurons in the output layer, is 23. The number of hidden layers is 2, with 256 neurons in each layer. A Softmax function is applied to the model output to convert it into a fault probability. The loss function and optimization function are the same as those used in a BP neural network.

[0143] The temperature time series dataset is randomly divided into training set, validation set and test set, accounting for 60%, 20% and 20% of the dataset respectively. Finally, the LSTM model is trained for 5000 generations. The training process is as follows Figure 9 As shown in the figure on the right. Figure 11The performance of the optimal model on the test set is shown. As can be seen from the figure, the model accurately diagnoses the faulty pipe section under most fault conditions. While the diagnostic accuracy decreases for individual pipe sections, such as n6 and n10, it still maintains a certain degree of effectiveness. Overall, the diagnostic accuracy reaches 90.62%, meeting practical engineering requirements.

[0144] The proposed method for diagnosing faults in a secondary heating pipe network based on pressure data has a diagnostic accuracy of 86.23%. However, reduced pressure sensor measurement accuracy can cause this model to lose accuracy or even become ineffective. The proposed dual-temperature-based fault diagnosis model achieves a diagnostic accuracy of 90.62%, meeting practical engineering requirements.

[0145] Step 4: Fault diagnosis model application

[0146] After the model training is completed, the model can be used to diagnose the faults of the actual secondary heating pipe network. The specific application process is as follows:

[0147] Figure 12 Determine whether the heating system is leaking based on the water replenishment volume monitored by the system. If a leak occurs, perform the following steps:

[0148] 1) Based on the dual temperature and dual pressure sensors of the secondary pipe network, data is collected for fault diagnosis.

[0149] 2) Combined with the actual engineering conditions of the secondary pipe network, the data collection frequency is determined based on the performance of the sensor and data collection centralized system, the number of historical failures, the complaint rate and other data.

[0150] 3) Complete data cleaning and feature engineering at the smart heating system data layer.

[0151] 4) Input the processed dual-temperature and dual-pressure data into the fault diagnosis model to perform fault diagnosis and obtain the faulty pipe section.

[0152] The fault diagnosis models based on dual temperature and dual pressure data can complement each other and provide a complete and feasible method to improve the accuracy and reliability of fault diagnosis in the secondary heating pipe network.

[0153] This embodiment makes full use of the dual-temperature and dual-pressure data of the heating secondary pipe network to construct a pipe network leakage diagnosis model, which improves the accuracy of the model, can quickly and accurately locate the pipe section where the leakage occurs, and carry out timely maintenance to ensure the safety and economy of the heating system operation. For different heating secondary pipe networks, it is possible to carry out targeted working condition simulation and fault diagnosis model construction. During diagnosis, leakage diagnosis is performed based on dual-pressure and dual-temperature time series data respectively. The two complement each other to improve the accuracy and applicability of diagnosis. Among them, the leakage diagnosis model based on dual-pressure data responds quickly, and the diagnostic accuracy is limited by the accuracy of the dual-pressure sensor; the leakage diagnosis model based on temperature time series data requires a certain response time and has a higher diagnostic accuracy.

[0154] This application has been disclosed as above with preferred implementation cases, but it is not intended to limit this application. Any technician familiar with this profession can make slight changes or modifications to the above-disclosed structures and technical contents without departing from the scope of the technical solution of this application, and the equivalent implementation cases with equivalent changes are still within the scope of the technical solution of this application.

Claims

1. A dual-temperature and dual-pressure heating secondary pipe network leakage fault diagnosis method, characterized by: The following steps are involved: S1. Construct a hydraulic and thermal simulation model of the secondary heating pipe network based on the actual parameters of the secondary heating pipe network; S2. Couple the hydraulic simulation model with the thermal simulation model to construct a simulation model of the actual heating system; The process of constructing the heating system simulation model in step S2 is as follows: S21. Based on the heating pipe network design information, input the node information and pipe segment information and connection relationships of the heating pipe network into a computer, and generate a directed connectivity graph of the heating pipe network based on graph theory; S22, setting the time step, space step and simulation period of the discrete solution, and setting the initial value of the iteration according to the measured data or the design value; S23, using a breadth-first traversal algorithm to determine the calculation order of the pipeline; S24. Based on the hydraulic simulation model of step S1, iteratively calculate the hydraulic parameters of the heating pipe network at time t, and provide flow parameters of each node and pipe section for thermal calculation; S25. Based on the thermal simulation model of step S1, the system is divided into three parts: heat source, heat user, and heating pipe network; first, the heat source model is called to determine the water supply temperature of the heat source at the current moment; According to the pipe network calculation sequence, the pipe network model is called to perform dynamic thermal calculations on each water supply pipe segment starting from the heat source to obtain the outlet temperature of the pipe and calculate the node temperature according to formula (7); then the heat user model is called to obtain the temperature and flow of the return water node; according to the pipe network calculation sequence, the pipe network model is called to calculate the outlet temperature of each return water pipe segment; Where: T i represents the temperature of node i, °C; T c,j represents the terminal temperature of pipe j flowing into node i, °C; T s,k represents the starting temperature of pipe k flowing out of node i, °C; Φ start，i represents the set of all pipes flowing out of node i; Φ end，i represents the set of all pipes flowing into node i; S3. Constructing hydraulic and thermal data sets of the heating network when different fault conditions occur in the heating network based on the heating system simulation model; S4, training a BP neural network leakage fault diagnosis model based on dual-pressure data and a LSTM neural network leakage fault diagnosis model based on dual-temperature time series data according to the data set; S5. Obtain a section of time series data on water replenishment to detect whether there is a leak in the pipe network. When the water replenishment volume of the heating pipe network does not exceed the threshold, there is no leak. When the water replenishment volume of the heating pipe network continuously exceeds the threshold, a pipe network leakage alarm signal is sent. At this time, the data of the dual-temperature and dual-pressure sensors of the heating system are collected and input into the trained dual-temperature and dual-pressure data leakage fault diagnosis model to obtain the faulty pipe section where the leak occurs, and complete the leakage fault diagnosis of the secondary heating pipe network.

2. The dual-temperature and dual-pressure heating secondary pipe network leakage fault diagnosis method according to claim 1, characterized in that: The construction process of the hydraulic simulation model in step S1 is: Assume that the flow velocity and pressure parameters of any cross section of the pipe network are uniformly distributed. Assume that the fluid in the pipes of the pipe network is incompressible; Then, the steady-state hydraulic simulation model equations are constructed as follows: According to the basic loop method, the steady-state hydraulic model is solved to obtain the flow rate, pressure, pipe section pressure drop, and flow rate of each node; Where: B represents the number of branches in the heating network; N represents the number of heating network nodes; A represents the incidence matrix, N-1×B dimensions; B f represents the basic loop matrix, B-N+1×B dimension; G represents the pipeline flow column vector, B dimension, kg / s; Q represents the column vector of the node's net inflow, that is, the flow exchanged between the node and the outside world of the system, with inflow being positive and outflow being negative, (N-1) dimension, kg / s; ΔH represents the pipeline pressure drop column vector, B dimension, Pa; Z represents the pipeline potential energy difference column vector, B dimension, Pa; S represents the pipeline impedance matrix, B×B dimensions, Pa·s 2 / kg 2 ; ΔH pump represents the pump head column vector, B dimension, Pa; C0, C1, and C2 represent the diagonal matrices of water pump characteristic parameters, with B×B dimensions.

3. The dual-temperature and dual-pressure heating secondary pipe network leakage fault diagnosis method according to claim 2, characterized in that: The construction process of the thermal simulation model in step S1 is as follows: Assumptions: The temperature distribution in any cross section of the pipe network is uniform; Ignore the axial heat conduction effect of the fluid; For the supply and return pipe sections, the thermal resistance between the fluid in the pipe, the pipe wall, the soil, and the atmosphere is evenly distributed in the pipe flow direction; 1) Build a pipe network model According to the law of conservation of energy, the temperature change of the fluid flowing in the pipeline can be expressed by formula (2). Based on (2), the time term is taken as a backward difference, and the space term is taken as a first-order upwind difference to obtain formula (3); Where: ρ represents the density of the fluid in the tube, kg / m 3 ; T represents the temperature of the fluid in the tube, °C; T env Indicates outdoor ambient temperature, ℃; u represents the flow velocity in the pipe, m / s; c p represents the specific heat capacity of the fluid in the tube, J / (kg·℃); t, Δt represents time, unit time length, s; x, Δx represents the pipe length, unit pipe length, m; A c represents the cross-sectional area of ​​the pipe, m 2 ; d represents the inner diameter of the pipe, m; λ represents the thermal conductivity of the fluid in the tube, W / (m·℃); R represents the equivalent heat transfer resistance per unit length between the fluid and the pipe wall-soil-atmosphere environment, m·℃ / W; u (n) represents the flow velocity in the pipe at time n, m / s; represents the outdoor ambient temperature at time n, ℃; T i (n) represents the temperature of the ith discrete node at time n after the network is discretized, ℃; 2) Build a hot user model The heating building is simplified as a mass point with lumped parameters, and the indoor temperature of each building is assumed to be an average temperature. The building exchanges heat with the radiator and the outside. The heat capacity of the equivalent heating building is concentrated at the mass point, and the formula (4) is listed. The backward difference method is used for the time term to obtain the indoor temperature of the building, as shown in formula (5). The heat dissipation of the heat dissipation equipment is shown in formula (6), where the inlet and outlet water temperatures of the heat dissipation equipment are equivalent to the supply and return water temperatures of the secondary pipe network. Combining formulas (4)-(6), the indoor temperature changes and supply and return water temperatures of the heat users can be calculated. Where: C b It represents the equivalent total heat capacity of the building, J / ℃; T n Indicates the average indoor temperature of the building, °C; represents the average indoor temperature of the building at time n, ℃; Q rad , G rad Indicates the heat dissipation and flow rate of the radiator, W; They represent the heat dissipation and flow rate of the radiator at time n, W; K b Indicates the comprehensive heat transfer coefficient of the building envelope, W / ℃; K rad Indicates the equivalent heat transfer coefficient of the radiator, W / ℃; η rad Indicates the flow time of the fluid in the radiator, which can be taken as a unit time step, η rad =1; T su , T re Indicates the inlet and outlet water temperatures of the heat dissipation equipment, ℃; Respectively represent the heat dissipation devices n-η rad Inlet water temperature at time n, outlet water temperature at time n (equivalent to the supply and return water temperature of the node), °C; 3) Build a heat source model The heat source is the starting point for the dynamic calculation of the heating network, and its thermal characteristics are the boundary conditions of the dynamic simulation model. They can be obtained in the following way: given the heat source at each moment, that is, the outlet water temperature of the secondary pipe network side of the heat exchange station The outlet water temperature is controlled to reach the set value through the automatic control system of the heat exchange station; 4) Build pipeline node model According to the energy conservation equation, the node temperature calculation formula (7) is obtained.

4. The dual-temperature and dual-pressure heating secondary pipe network leakage fault diagnosis method according to claim 3, characterized in that: The process of constructing the hydraulic and thermal data sets of the heating network when different fault conditions occur in the heating network in step S3 is as follows: First, pipe sections are selected based on the actual situation of the heating pipe network. Multiple different leakage locations μ = {μ1, μ2, μ3, μ4, …} are set for each pipe section. Multiple leakage amounts are set for each leakage location for simulation calculation to obtain node pressure data and temperature time series data. Each data is a vector containing the values ​​of multiple pressure or temperature measurement points. The faulty pipe sections are uniquely labeled and the fault diagnosis dataset is constructed.

5. A dual-temperature and dual-pressure heating secondary pipe network leakage fault diagnosis method according to claim 1 or 4, characterized in that: The BP neural network leakage fault diagnosis model in step S4 is a multi-layer neural network structure, wherein the number of neurons in the input layer is x n Equal to the number of thermal users equipped with dual pressure sensors for thermal inlet, the number of neurons in the output layer y n The model takes the dual-pressure data of heat users as the input value of the neural network and the probability of failure of each pipeline section to be diagnosed as the output value.

6. A dual-temperature and dual-pressure heating secondary pipe network leakage fault diagnosis method according to claim 1 or 4, characterized in that: The LSTM neural network leakage fault diagnosis model in step S4 is a multi-layer neural network structure, in which the number of neurons in the input layer is x n Equal to the number of thermal users equipped with dual temperature sensors at the thermal inlet, the number of neurons in the output layer y n The model takes the dual-temperature time series data of heat users as input variables and the probability of failure of each pipe section to be diagnosed as the output value.

Images