Community central heating system and pipe network leakage fault intelligent grading detection method

Abstract

The invention provides a community central heating system and pipe network leakage fault intelligent grading detection method, the heating system comprises a heat source, a heating pipe network and a user, and further comprises a pipe network on-line monitoring system, and the monitoring system reads the pressure of the heat source and the user and the instantaneous flow of a make-up pump at the heat source according to a set fixed time interval; and storing in a database in real time. According to the intelligent grading detection method for the pipe network leakage faults, an improved accumulation and CUSUM method in a point change theory in the field of statistics and a BP neural network method in a machine learning algorithm are fused, the two methods are organically combined based on a heat supply pipe network hydraulic simulation model, and whether leakage occurs or not, the leakage degree and the specific positions of leakage points are diagnosed in a grading mode. And optimizing the simulation model through real-time operation data by using a genetic algorithm. According to the system and the method, the leakage fault detection precision is improved, and timely and accurate diagnosis and positioning of the leakage fault of the community central heating system are realized.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of central heating, and particularly relates to a community central heating system and an intelligent hierarchical detection method for pipe network leakage faults. BACKGROUND

[0002] Central heating pipe network is an indispensable part of modern urban infrastructure. It centralizes heat sources and uses a pipeline network to deliver heat energy to users, meeting the heating and hot water needs of residential, commercial and industrial users. However, during the long-term operation of the central heating system, pipe network leakage problems occur from time to time, which not only affects the heating efficiency, but also may cause safety hazards. Therefore, the central heating pipe network leakage detection technology is of great significance for the operation, maintenance and safety of the heating system.

[0003] For the leakage problem of central heating pipe network, there are currently a variety of detection technologies applied. Traditional detection methods include on-site manual inspection and pipe pressure testing. These methods, although low in cost, are not efficient and difficult to find hidden leakage problems in a timely manner. With the development of technology, advanced detection methods such as direct-buried warning line method, distributed optical fiber temperature measurement method and infrared imaging detection method have been gradually applied to the monitoring of heating pipe network. However, these methods have defects such as complex installation process, large-scale construction, high installation cost and susceptibility to external environment.

[0004] In addition, with the development of Internet of Things and big data technology, intelligent monitoring systems have also been introduced into the management of central heating pipe network. Mainly including model method, statistical detection method, etc., by installing sensors at key nodes of the pipeline, the temperature, pressure and other parameters of the pipe network are monitored in real time. The model method is to establish a steady-state or transient model of the heating pipe network, compare the simulated data of the pipe network with the actual operation data (such as flow or pressure), and analyze whether there is a leakage. The accuracy of this method mainly depends on the accuracy of the pipe network model. The statistical detection method is based on statistical theory, analyzes the operation data under the leakage condition, and estimates the leakage amount and leakage position by establishing a functional relationship with the normal working condition data. This method does not need to establish a complex physical model, only needs to perform a small amount of probability calculation of pressure and flow, and has strong adaptability, but requires high accuracy of measuring instruments.

[0005] The Chinese patent application with publication number CN110245411A proposes a city centralized heat supply system and a leakage point detection method. Based on the improved cumulative sum (CUSUM) method in the change point theory in the field of statistics, the occurrence time and size of the leakage fault are diagnosed. In combination with a heat supply pipe network leakage simulation model, the leakage position is determined by comparing the actual data of the heat supply system pressure monitoring point with the simulation data generated by the simulation model. This method completely depends on the accuracy of the heat supply system simulation model, and the accuracy of the heat supply system simulation model still needs to be further improved. Moreover, for the community heat supply system which generally lacks detection instruments and instrument installation positions, it is difficult to ensure the accuracy of the simulation model, thereby seriously affecting the improvement of the leakage position detection efficiency. SUMMARY

[0006] The present application aims at the deficiencies in the prior art, and provides a community centralized heat supply system and a pipe network leakage fault intelligent grading detection method for real-time detection of community pipe network leakage to solve the technical problem of real-time detection of community heat supply pipe network leakage.

[0007] To achieve the above-mentioned purpose, the present application provides a community centralized heat supply system, which comprises a heat source, a heat supply pipe network and heat users, the heat supply pipe network comprises a water supply pipe network and a return water pipe network, and further comprises a pipe network online monitoring system. The pipe network online monitoring system reads the pressure at the heat source and the heat users and the instantaneous flow data of the water replenishing pump at the heat source according to a set fixed time interval, and stores them in a database in real time, and calculates the cumulative sum of the water replenishing flow. The heat users are provided with a number of sampling pressure gauges, which accounts for at least 1 / 6 of the number of user units, and are distributed according to the principle of near, middle and far from the heat source, and the ratio of the water supply side to the return water side is 2:1.

[0008] A pipe network leakage fault intelligent grading detection method for a community centralized heat supply system, comprising the following steps: Step S1: establishing a heat supply pipe network hydraulic simulation model; Step S2: optimizing the heat supply pipe network hydraulic simulation model; Step S3: real-time monitoring of the water replenishing flow data of the heat supply pipe network to determine whether a leakage occurs in the heat supply pipe network; Step S4: analyzing the water replenishing flow monitoring data to determine the leakage time; Step S5: establishing a heat supply pipe network leakage position diagnosis model to determine the specific position of the leakage point.

[0009] Further, the step S1 specifically comprises the following steps: Step S11: establishing a basic incidence matrix and a basic loop matrix; specifically, the pipe sections and nodes of the heat supply pipe network are numbered, and the basic incidence matrix and the basic loop matrix are obtained by analyzing the overall topological structure of the pipe network; Step S12: establishing a resistance characteristic coefficient diagonal matrix; specifically, according to the inner diameter and length parameters of each pipeline, the resistance characteristic coefficients of each pipe section are calculated, and then the resistance characteristic coefficient diagonal matrix of the heat supply pipe network is obtained; Step S13: flow iteration; specifically, a Maxwell matrix is constructed to calculate the pressure drop balance difference of the heat supply pipe network, and an improved square root method is used for numerical solution; in this process, the precision value is set, and the calculation result is adjusted constantly through iteration operation until the precision requirement is met, and then the next step is entered; if the iteration number exceeds the set upper limit and the result still does not converge, an unconverged signal is output; Step S14: calculation of pipe network node pressure and flow distribution; specifically, according to the actual distribution of the heat supply pipe network, the node pressure and flow of each pipe section are further calculated, and the calculation results are displayed in the form of a matrix.

[0010] Further, the step S2 specifically comprises the following steps: Step S21: encoding; specifically, a binary encoding strategy is used to simulate the pipe network resistance characteristic coefficient solution set into a chromosome form, and convert it into a corresponding digital string to represent the trial solution in the feasible region of the optimization problem; Step S22: initialization; specifically, the size of the population is set, and the digital string individuals meeting the requirements are selected by a random selection function to obtain an initialized population; Step S23: selection and replication; specifically, individuals meeting the requirements are selected from the initialized population for replication to obtain a population; Step S24: crossover; specifically, a pair of individuals are randomly selected from the population, and a position is randomly selected to exchange part of the genes between them to generate two new individuals, and the operation is repeated until 65% of the individuals in the population are crossed to update the entire population; Step S25: mutation; specifically, a mutation probability of 0.005 is selected for an individual in the population, and a mutation operation is randomly performed on a position in the individual to effectively prevent the genetic algorithm from maturing too early; Step S26: establishing an adaptive function; specifically, the adaptive function of the heat supply pipe network is established as: In the formula, F(S) represents the resistance characteristic function; n represents the number of users of the pipe network; n represents the number of users of the pipe network; represents the measured value of the user inlet pressure, unit: kPa; represents the calculated value of the user inlet pressure, unit: kPa; Puser represents the measured value of the user outlet pressure, unit: kPa; Puser represents the calculated value of the user outlet pressure, unit: kPa; The constraint condition is: In the formula: Qpipe represents the pipe section flow, unit: m³ / h; Qsup represents the water supply flow, unit: m³ / h; Ppipe represents the pipe section resistance pressure drop, unit: kPa; S represents the pipe section resistance characteristic coefficient, unit: kPa·h 2 / m 6 ; DH represents the water pump pressure, unit: kPa; Smin represents the minimum pipe section resistance characteristic coefficient, unit: kPa·h 2 / m 6 ; Smax represents the maximum pipe section resistance characteristic coefficient, unit: kPa·h 2 / m 6 ; Step S27: evolution iteration; specifically, through continuous loop iteration, the fitness of the population is continuously improved, new individuals are generated, and an optimized pipe network resistance characteristic coefficient matrix is obtained; Step S28: based on the optimized pipe network resistance characteristic coefficient matrix, a heat supply pipe network optimization hydraulic simulation model is established, and then simulation and calculation of the normal working condition and the leakage working condition of the pipe network are carried out, to generate simulation label data under various working conditions.

[0011] Further, the step S3 specifically comprises: based on the change point theory in the field of statistics, using the improved CUSUM method, the pipe network water supply flow monitoring data is tracked and analyzed in real time, when the cumulative sum of the water supply flow is found to continuously increase and exceed the set threshold, an alarm is issued, and it is judged that the region has a leakage.

[0012] Further, the step S4 specifically comprises: when the detection system issues a leakage alarm, the community pipe network water supply flow monitoring data is automatically analyzed, and according to the cumulative sum and change curve of the water supply flow, the change point occurrence position is backtracked, so as to determine the specific time when the leakage starts and the leakage reaches the maximum, and calculate the leakage amount at the corresponding time.

[0013] Further, the step S5 specifically comprises: generating leakage condition label data through a pipe network hydraulic simulation model, establishing a hybrid data set of actual running pressure and simulation pressure, training by using a machine learning BP neural network, and finally obtaining a heat supply pipe network leakage position model to determine the specific position of the leakage point; specifically comprising the following steps: Step S51: data processing; specifically, re-examining and verifying the heating system pressure data in the database, using the sample mean to replace the missing values by using the estimation method for the missing data, deleting the invalid data from the database, checking the inconsistent data with the help of the integrity constraint mechanism of the database management system, and then correcting the inconsistent data by referring to the corresponding data values in the database; ensuring the correctness and logical consistency of the data; Step S52: supplementing data; specifically, using the heating pipe network hydraulic simulation optimization model to add "virtual users" to correspondingly increase "virtual nodes" on the supply and return water pipeline, realizing pipe node leakage simulation, and supplementing leakage working condition label data; Step S53: generating a data set; specifically, performing data normalization processing on the prepared data, reading the pressure data values of the corresponding working conditions from the database according to the different operating conditions of the heating pipe network, and generating training set data and labels and test set data and labels under various working condition states; dividing into training set / training set labels and test set / test set labels; Step S54: network training; specifically, inputting the training set data into the BP neural network, calculating the network output and the training set label to obtain the network error, using the error back propagation algorithm to constantly correct the network weight and bias, making the network error meet the set accuracy requirement, completing the network training, and generating a trained BP neural network model; Step S55: network testing; specifically, inputting the test set data into the trained BP neural network model, outputting the label, and thus outputting the fault detection result to determine the specific location of the leakage point.

[0014] Further, the step S54 specifically comprises: Step S541: initializing the weight matrix and bias of the BP neural network; Step S542: inputting a group of training set data and labels into the BP neural network; Step S543: calculating the network output, subtracting the network output from the training set label to obtain the network error, and determining whether the error meets the requirement; Step S544: if the error meets the accuracy requirement, the training is completed, and the network parameters are saved; if the error does not meet the requirement, the derivative of the network error is calculated, the derivative is back propagated, and the weight matrix and bias of the BP network are sequentially corrected; Step S545: repeating the above process until the network error meets the accuracy requirement, the network training process is completed, and a trained BP neural network model is generated.

[0015] Compared with the prior art, the present application adopts the BP neural network in the machine learning theory to establish a diagnosis model of the leakage fault position of the heating system, replaces the comparison of the actual data of the pressure monitoring point of the heating system with the simulation data generated by the simulation model in the prior art to determine the leakage position, thereby effectively reducing the dependence of the leakage diagnosis efficiency on the accuracy of the simulation model of the heating system and improving the diagnosis efficiency of the leakage position; the genetic optimization algorithm is adopted, the resistance characteristic coefficient of the heating pipe network is optimized according to the actual operation data, and the accuracy of the simulation model of the heating system is further improved, so as to provide more accurate normal working condition and leakage working condition label data for the training of the BP neural network model.

[0016] The present application has the following advantages: (1) The improved cumulative sum (CUSUM) method in the change point theory in the field of statistics and the BP neural network method in the machine learning algorithm are combined organically based on the heating pipe network hydraulic simulation model, the two methods are used to diagnose whether the leakage occurs, the leakage degree and the specific position of the leakage point in stages, the genetic algorithm is used to optimize the simulation model through real-time operation data, the online staged detection system and method for the heating pipe network leakage based on the change point theory and machine learning are proposed, the leakage detection efficiency of the heating pipe network is improved, the false positive rate and the false negative rate are reduced, and the safe operation of the heating pipe network is ensured; (2) The method comprises two-stage intelligent detection of the leakage fault, the first-stage detection system based on the improved CUSUM method is used to diagnose whether the leakage occurs and the leakage degree by monitoring the change of the water supply flow of the community; the second-stage detection system based on the BP neural network and the pipe network simulation model is used to diagnose the specific leakage position by monitoring the change of the sampling pressure gauge and the pressure data of the simulation model under the leakage working condition, and the first-stage and second-stage leakage diagnosis results are comprehensively processed and alarmed; (3) A new idea of comprehensively monitoring the leakage occurrence in stages by detecting the water supply flow of the community pipe network and the change of the sampling pressure gauge is proposed, the online monitoring data of the heating pipe network are fully utilized, and the simulation data of the pipe network are organically combined, so that the detection speed is fast and the cost is low; (4) The data processing technology, the change point theory, the machine learning and the computer simulation technology are organically combined, the leakage detection efficiency of the community heating pipe network is improved, the leakage position is determined, and the safe operation of the heating pipe network and the heat source unit is ensured. The system and method improve the leakage fault detection precision, realize the timely and accurate diagnosis and positioning of the leakage fault of the community central heating system. BRIEF DESCRIPTION OF DRAWINGS

[0017] Figure 1 A community central heating system schematic diagram is shown; Figure 2 A flow chart of an optimized heating pipe network hydraulic simulation model is shown; Figure 3The flowchart of the BP neural network training algorithm for the secondary leak location detection method in pipeline network is shown; Figure 4 The flowchart of the overall algorithm for the graded detection method of leakage in heating pipe networks is shown. Detailed Implementation

[0018] To clearly illustrate the technical features of this solution, specific implementation methods are described below. The accompanying drawings are simplified schematic diagrams, used only to illustrate the technical solution of this invention.

[0019] Figure 1 The diagram illustrates a centralized heating system for a residential community. This system includes a heat source (a secondary heat exchange station or boiler room), a heating network, and heat users. The heating network comprises a supply water network and a return water network, arranged symmetrically. It also includes an online monitoring system that reads pressure data at the heat source and user locations, as well as the instantaneous flow rate of the makeup water pump at the heat source, at fixed time intervals. This data is stored in a database in real time, and the cumulative makeup water flow rate is calculated. The flow rate at each user location is simultaneously the outlet flow rate of the supply water network and the inlet flow rate of the return water network. A leak is detected based on a given threshold, and an alarm is triggered. At least one-sixth of the heat user units must have sampling pressure gauges, distributed according to their distance from the heat source (near, medium, far), with a 2:1 ratio between the supply and return water sides. Instantaneous flow meters must be installed on the makeup water pumps.

[0020] A smart hierarchical detection method for pipeline leakage faults in a community's centralized heating system is proposed. This method monitors changes in the pressure at the heat source and user locations, as well as the water supply flow rate at the heat source, through an online monitoring system. First, based on changes in the cumulative deviation of the water supply flow rate, the method determines whether a leak has occurred and the extent of the leak. Then, based on changes in the pressure gauges on the pipeline sampling point, leak label data is supplemented using a pipeline simulation model. Features of the pipeline pressure data are extracted using a BP neural network to diagnose the leak location. Finally, the results of these two levels of leak diagnosis are comprehensively analyzed to determine whether a leak has occurred, the extent of the leak, and the location of the leak point. Specifically, the alarm method uses the cumulative deviation of the community's pipeline water supply flow rate to determine whether a leak has occurred and the extent of the leak, and uses the characteristics of pipeline pressure changes to determine the leak location, notifying management personnel.

[0021] A method for intelligent hierarchical detection of pipeline leakage faults in a centralized heating system for residential communities includes the following steps: Step S1: Establish a hydraulic simulation model of the heating network; Detailed process of hydraulic simulation of the heating network: Step S11: Establish the basic correlation matrix and basic loop matrix; number the pipe segments and nodes of the pipeline network, and obtain its basic correlation matrix and basic loop matrix by analyzing the overall topology of the pipeline network. Step S12: Establish a diagonal matrix of resistance characteristic coefficients; calculate the resistance characteristic coefficients of each pipe segment based on parameters such as the inner diameter and length of each pipe, and then obtain the diagonal matrix of resistance characteristic coefficients of the pipe network. Step S13: Flow Iteration; Construct the Maxwell M matrix to calculate the pressure drop balance difference of the pipeline network, and use the improved square root method for numerical solution; In this process, it is necessary to set the accuracy value, and continuously adjust the calculation results through iterative calculation until the accuracy requirements are met before proceeding to the next step; If the number of iterations exceeds the set upper limit and the result still does not converge, a non-converged signal will be output. Step S14: Calculation of pressure and flow distribution at pipeline nodes; This module is used to express the solution results of the flow iteration module; Based on the actual distribution of the pipeline network, the node pressure and flow of each pipe segment are further calculated, and the calculation results are displayed in the form of a corresponding matrix.

[0022] Based on the above steps, here's an example of the modeling process: Let the number of pipe segments in the heating network be m, and the number of nodes be n+1. Then, using network graph theory and Kirchhoff's laws, the equations for the hydraulic simulation model of this heating network are as follows: In the formula, A is the basic correlation matrix of the pipeline network, of order n×m; B f G is the basic loop matrix of the pipeline network, of order (mn) × m; G is the flow vector of the pipe segment, G = (G1, G2, ..., G...). m ); ΔH is the pressure drop due to pipe section resistance, ΔH = (ΔH1, ΔH2, ..., ΔH) m S is the pipe segment resistance characteristic coefficient matrix (m-order diagonal matrix), S=diag{S1,S2,…,S} m}; |G| is the absolute value of the flow rate G in the pipe segment, an m-order diagonal matrix, |G|=diag{|G1|,|G2|,…,|G m |};DH is the pump vector of the pipe section, DH=(DH1,DH2,...,DH m ) T When the pipe section does not contain a pump, DH=0; when a pump is present, DH is the pump head. Z is the potential energy difference vector between two nodes in the pipe section branch, Z=(Z1,Z2,...,Z...). m ) T Q represents the node leakage, Q = (Q1, Q2, ..., Q...). n Under normal operating conditions, Q=0; In the formula, K is the equivalent absolute roughness of the pipe wall, in meters (m). For heating pipes, K is generally 0.0005 m; d is the inner diameter of the pipe, in meters; l, l dThese represent the calculated length of the pipe segment and the equivalent length of local resistance in the pipeline network, respectively, in meters (m); ρ is the average density of the fluid medium inside the pipe, in kg / m³. 3 .

[0023] make , Then the above hydraulic simulation model equations become In the formula, M is the Max Well matrix, which is a matrix with respect to B. f M is a symmetric positive definite matrix of order (mn) × (mn) based on the tree; the M matrix corresponds to a certain tree, and different trees correspond to different M matrices; Δh k —Algebraic sum of voltage drops in the basic loop sections, when When it is a solution to the system of equations, its value is 0; k is the number of iterations.

[0024] Using the improved square root method, the M matrix is ​​uniquely decomposed into... In this matrix, L is a unit lower triangular matrix, D is a diagonal matrix, and since the M matrix is ​​a symmetric positive definite matrix, all diagonal elements of D are positive numbers.

[0025] Therefore, the equations of the hydraulic simulation model become By matrix multiplication, and note , ,have to To avoid double counting, introduce Then the formula for calculating the L and T elements row by row is: make y only requires solving two simple systems of triangle equations: ① Solution Substitute to find , ② Solution Find the answer by back substitution. .

[0026] Taking the root of equation ① yields the following relation: Then, the following relationship can be derived from equation ②; Finally, the flow rate G of each pipe segment in the pipeline network and the pressure drop ΔH of each pipe segment are calculated. Then, based on the pressure of the set reference node (usually the fixed pressure point of the pipeline network), the pressure P value of each node in the pipeline network is calculated.

[0027] Step S2: Optimize the hydraulic simulation model of the heating network, the process is as follows: Figure 2 As shown, addressing the lack of operational data, particularly user flow data, for community heating systems, this paper proposes utilizing actual operating pressure data of the heating network and employing a genetic algorithm to optimize the resistance characteristic coefficients in the hydraulic simulation model of the network. A selection strategy is used to ensure population balance during the genetic algorithm search. The specific steps include: Step S21: Encoding; Using a binary encoding strategy, the solution set of pipeline resistance characteristic coefficients is simulated into a chromosome form and converted into corresponding number strings. A large number of number strings are used to represent the trial solutions within the feasible region of the optimization problem. Step S22: Initialization; Set the population size by randomly selecting individuals from the generated string of numbers that meet the requirements to obtain the initial population; Step S23: Select for replication; Select individuals that meet the requirements from the initial population for replication to obtain the population; Step S24: Crossover; Randomly select one pair of individuals from the population, then randomly select one position, and exchange some of their genes to generate two new individuals. Repeat the operation until more than 65% of the individuals in the population have been crossovered, thereby updating the entire population. Step S25: Mutation; Select an individual in the population with a very small probability (preferably a mutation probability of 0.005), and then randomly perform a mutation operation at a certain position in that individual to effectively prevent the genetic algorithm from maturing too early; Step S26: Establish the fitness function; the fitness function for this centralized heating network is established as follows: In the formula: F(S) represents the resistance characteristic function; n represents the number of users in the pipeline network; This indicates the measured value of the user's inlet pressure, in kPa. This represents the calculated value of the user's inlet pressure, in kPa. This indicates the measured value of the user's outlet pressure, in kPa. This represents the calculated value of the user's outlet pressure, in kPa. The constraints are: In the formula: This indicates the flow rate of the pipe section, in m³ / h. This indicates the water replenishment flow rate, in m³ / h. S represents the pressure drop due to pipe segment resistance, in kPa; S represents the pipe segment resistance characteristic coefficient, in kPa·h. 2 / m 6 DH represents the pump pressure, in kPa. This represents the minimum resistance characteristic coefficient of the pipe section, in kPa·h. 2 / m 6 ; This represents the maximum resistance characteristic coefficient of the pipe section, in kPa·h. 2 / m 6 ; Step S27: Evolutionary Iteration; Through continuous iterative cycles, the fitness of the population is continuously improved, new individuals are generated, and an optimized pipeline resistance characteristic coefficient matrix is ​​obtained. Step S28: Based on the optimized pipeline resistance characteristic coefficient matrix, establish an optimized hydraulic simulation model of the heating pipeline network, and then perform simulation calculations for normal and leakage conditions of the pipeline network to generate simulation label data under various conditions.

[0028] Step S3: Monitor the water supply flow data of the heating network in real time to determine whether there is a leak in the heating network; Based on the change point theory in statistics, an improved CUSUM method is used to track and analyze the water supply flow monitoring data of the community pipe network in real time. Once the accumulation and continuous increase of the water supply flow is detected and exceeds the set threshold, an alarm is issued to determine that a leak has occurred.

[0029] Step S4: Analyze the water replenishment flow monitoring data to determine the time of leakage; Once the system issues a leak alarm for the community's pipe network, the improved CUSUM method will continue to calculate the cumulative sum until the cumulative sum changes from increasing to decreasing, at which point the calculation will stop. Based on the cumulative sum change curve of the community's pipe network water supply flow, the system will calculate the time when the leak occurred and the time when the leak reached its maximum value, and calculate the maximum leakage amount.

[0030] The specific scheme for determining whether a leak has occurred in the pipeline network and the extent of the leak based on the cumulative change of the pipeline network water supply flow offset has been disclosed in Chinese patent application CN110245411A, and will not be described in detail here.

[0031] Step S5: Establish a leak location diagnosis model for the heating network to determine the specific location of the leak. Specifically, this involves generating leak condition label data through a network hydraulic simulation model, establishing a hybrid dataset of actual operating pressure and simulated pressure, training it using a backpropagation (BP) neural network, and finally obtaining a leak location model for the heating network to determine the specific location of the leak. This includes the following steps: Step S51: Data processing; re-examine and verify the heating system pressure data in the database, correct missing, invalid, and inconsistent data, and ensure the correctness and logical consistency of the data. 1) Handling missing data: Due to network transmission failures, missing values ​​may appear in the database. For missing data values, an estimation method is used, and the sample mean is used to replace the missing values. 2) Handling invalid data: If invalid values ​​appear in the pressure data in the database due to sensor malfunction, they will be deleted from the database. 3) Handling inconsistent data: Use the integrity constraint mechanism of the database management system to check for inconsistent data, and then correct it by referring to the corresponding data values ​​in the database; Step S52: Supplement data; Using the hydraulic simulation optimization model of the heating network, by adding "virtual users", "virtual nodes" are added to the supply and return water pipelines to realize pipeline node leakage simulation and supplement various leakage condition label data; Step S53: Generate dataset; Normalize the prepared data, read the corresponding pressure data values ​​from the database according to the different operating conditions of the heating network, and generate training set data and labels, test set data and labels under various operating conditions; Divide into training set / training set labels and test set / test set labels; Step S54: Network training; Input the training set data into the BP neural network, calculate the network error by comparing the network output with the training set labels, and continuously adjust the network weights and biases using the backpropagation algorithm to ensure that the error meets the set accuracy requirements. The network training is then complete; A trained BP neural network model is generated. Figure 3 The flowchart of the BP neural network training algorithm is shown below: Step S541: Initialize the weight matrix and biases of the BP neural network; Step S542: Feed a set of training data and labels into the BP neural network; Step S543: Calculate the network output, subtract the network output from the training set labels to obtain the network error, and determine whether the error meets the requirements; Step S544: If the error meets the accuracy requirements, training ends and the network parameters are saved; if the error does not meet the requirements, the error gradient is calculated, that is, the network error is differentiated, and the derivative is used for backpropagation to correct the weight matrix and bias of the BP network in turn. Step S545: Repeat the above process until the network error meets the accuracy requirements, the network training process is completed, and a trained BP neural network model is generated. Step S55: Network testing; Input the test set data into the pre-trained BP neural network model, output labels, and thus output the fault detection results to determine the specific location of the leak point.

[0032] Figure 4 The overall algorithm flowchart of the graded detection method for leaks in heating pipe networks is shown below: 1) Read the pipeline water supply flow monitoring data sequence according to the set time interval (e.g., 2 minutes), perform cumulative calculation, and then determine whether the cumulative sum exceeds the set threshold. If it does not exceed the threshold, continue the cumulative calculation; if it does, trigger a leak alarm. 2) Continue calculating the cumulative sum, and at the same time start calculating the rate of change of the cumulative sum until it is less than or equal to zero. Calculate the rate of change of the cumulative sum when the leakage begins and when the leakage reaches its maximum, respectively, to obtain the value at the time when the leakage begins and the value at the time when the leakage reaches its maximum. Then obtain the water replenishment flow rate at the corresponding time, and finally calculate the maximum leakage amount. 3) Load the pre-trained BP neural network diagnostic model, read the pipeline network detection pressure data, and output the specific diagnostic location of the leak point.

[0033] The technical features of this invention not described can be implemented by or using existing technology, and will not be repeated here. Of course, the above description is not a limitation of this invention, and this invention is not limited to the examples above. Any changes, modifications, additions or substitutions made by those skilled in the art within the scope of this invention should also be within the protection scope of this invention.

Claims

1. A centralized heating system for a residential community, comprising a heat source, a heating network, and heat users, wherein the heating network includes a water supply network and a return network, characterized in that, It also includes an online pipeline monitoring system, which reads the pressure at the heat source and heat user and the instantaneous flow rate of the water supply pump at the heat source at a set fixed time interval, stores it in the database in real time, and calculates the cumulative sum of the water supply flow. For heat users, the number of sampling pressure gauges should account for at least 1 / 6 of the number of user units, and they should be distributed according to the principle of being close to, medium to far from the heat source, with the ratio of supply water side to return water side being 2:

1.

2. A method for intelligent hierarchical detection of pipeline leakage faults in a centralized heating system for residential communities, characterized in that, Includes the following steps: Step S1: Establish a hydraulic simulation model of the heating network; Step S2: Optimize the hydraulic simulation model of the heating network; Step S3: Monitor the water supply flow data of the heating network in real time to determine whether there is a leak in the heating network; Step S4: Analyze the water replenishment flow monitoring data to determine the time of leakage; Step S5: Establish a diagnostic model for the location of leaks in the heating network to determine the specific location of the leak.

3. The intelligent hierarchical detection method for pipeline leakage faults in a centralized heating system for residential communities according to claim 2, characterized in that, Step S1 specifically includes the following steps: Step S11: Establish the basic correlation matrix and basic loop matrix; specifically, number the pipe segments and nodes of the heating network, and obtain its basic correlation matrix and basic loop matrix by analyzing the overall topology of the network. Step S12: Establish the diagonal matrix of resistance characteristic coefficients; specifically, calculate the resistance characteristic coefficients of each pipe segment based on the inner diameter and length parameters of each pipe, and then obtain the diagonal matrix of resistance characteristic coefficients of the heating network. Step S13: Flow rate iteration; specifically, construct the Maxwell matrix to calculate the pressure drop balance difference of the heating network, and use the improved square root method for numerical solution; during this process, set the accuracy value, and continuously adjust the calculation results through iterative calculation until the accuracy requirements are met before proceeding to the next step; if the number of iterations exceeds the set upper limit and the result still does not converge, output a non-converged signal. Step S14: Calculation of pressure and flow distribution at pipe network nodes; specifically, based on the actual distribution of the heating pipe network, the node pressure and flow of each pipe segment are further calculated, and the calculation results are displayed in the form of a matrix.

4. The intelligent hierarchical detection method for pipeline leakage faults in a centralized heating system of a residential community according to claim 2, characterized in that, Step S2 specifically involves using actual operating pressure data of the heating network to optimize the resistance characteristic coefficients in the hydraulic simulation model of the heating network using a genetic algorithm, and employing a selection strategy to ensure the balance of the population search in the genetic algorithm. This includes the following steps: Step S21: Encoding; Specifically, a binary encoding strategy is adopted to simulate the solution set of pipeline resistance characteristic coefficients into chromosome form and convert it into corresponding number strings to represent the trial solutions within the feasible region of the optimization problem. Step S22: Initialization; specifically, setting the population size, selecting individual digit strings that meet the requirements through a random selection function to obtain the initialized population; Step S23: Select for replication; specifically, select individuals that meet the requirements from the initial population for replication to obtain the population; Step S24: Crossover; Specifically, randomly select one pair of individuals from the population, then randomly select one position, exchange some of their genes to generate two new individuals, repeat the operation until 65% of the individuals in the population have been crossed, thereby updating the entire population. Step S25: Mutation; Specifically, select an individual in the population with a mutation probability of 0.005, and then randomly perform a mutation operation at a certain position in that individual to effectively prevent the genetic algorithm from maturing too early; Step S26: Establish the fitness function; specifically, establish the fitness function for the heating network as follows: In the formula: F(S) represents the resistance characteristic function; n represents the number of users in the pipeline network; This indicates the measured value of the user's inlet pressure, in kPa. This represents the calculated value of the user's inlet pressure, in kPa. This indicates the measured value of the user's outlet pressure, in kPa. This represents the calculated value of the user's outlet pressure, in kPa. The constraints are: In the formula: This indicates the flow rate of the pipe section, in m³ / h. This indicates the water replenishment flow rate, in m³ / h. S represents the pressure drop due to pipe segment resistance, in kPa; S represents the pipe segment resistance characteristic coefficient, in kPa·h. 2 / m 6 DH represents the pump pressure, in kPa. This represents the minimum resistance characteristic coefficient of the pipe section, in kPa·h. 2 / m 6 ; This represents the maximum resistance characteristic coefficient of the pipe section, in kPa·h. 2 / m 6 ; Step S27: Evolutionary iteration; specifically, through continuous iterative cycles, the fitness of the population is continuously improved, new individuals are generated, and an optimized pipeline resistance characteristic coefficient matrix is ​​obtained. Step S28: Based on the optimized pipeline resistance characteristic coefficient matrix, establish an optimized hydraulic simulation model of the heating pipeline network, and then perform simulation calculations for normal and leakage conditions of the pipeline network to generate simulation label data under various conditions.

5. The intelligent hierarchical detection method for pipeline leakage faults in a centralized heating system for residential communities according to claim 2, characterized in that, Specifically, step S3 involves using an improved CUSUM method based on the change point theory in statistics to track and analyze the pipeline water supply flow monitoring data in real time. When the accumulation and continuous increase of the water supply flow is found to exceed the set threshold, an alarm is issued to determine that a leak has occurred in the area.

6. The intelligent hierarchical detection method for pipeline leakage faults in a centralized heating system of a residential community according to claim 2, characterized in that, Specifically, step S4 involves automatically analyzing the water supply flow monitoring data of the community pipe network when the detection system issues a leak alarm, and deducing the location of the change point based on the cumulative and changing curve of the water supply flow, thereby determining the specific time when the leak started and the leak reached its maximum, and calculating the leak amount at the corresponding time.

7. The intelligent hierarchical detection method for pipeline leakage faults in a centralized heating system for residential communities according to claim 2, characterized in that, Step S5 specifically involves generating leakage condition label data through a pipeline hydraulic simulation model, establishing a hybrid dataset of actual operating pressure and simulated pressure, training it using a backpropagation (BP) neural network, and finally obtaining a leakage location model of the heating pipeline network to determine the specific location of the leakage point; this includes the following steps: Step S51: Data processing; specifically, the heating system pressure data in the database is re-examined and verified. For missing data, an estimation method is used to replace the missing values ​​with the sample mean. For invalid data, it is deleted from the database. For inconsistent data, the integrity constraint mechanism of the database management system is used to check for inconsistent data, and then the corresponding data values ​​in the database are referenced for correction. This ensures the correctness and logical consistency of the data. Step S52: Supplement data; Specifically, using the hydraulic simulation optimization model of the heating network, by adding "virtual users", "virtual nodes" are added to the supply and return water pipelines to realize pipeline node leakage simulation and supplement leakage condition label data; Step S53: Generate dataset; specifically, normalize the prepared data, read the pressure data values ​​corresponding to the operating conditions of the heating network from the database, and generate training set data and labels, test set data and labels for various operating conditions; divide them into training set / training set labels and test set / test set labels. Step S54: Network training; Specifically, the training set data is input into the BP neural network, the network error is obtained by calculating the network output and the training set labels, and the network weights and biases are continuously corrected using the backpropagation algorithm to make the network error meet the set accuracy requirements. The network training is completed, and a trained BP neural network model is generated. Step S55: Network testing; specifically, the test set data is input into the pre-trained BP neural network model, and the labels are output to determine the fault detection results and the specific location of the leak point.

8. The intelligent hierarchical detection method for pipeline leakage faults in a centralized heating system for residential communities according to claim 7, characterized in that, Step S54 specifically involves: Step S541: Initialize the weight matrix and biases of the BP neural network; Step S542: Feed a set of training data and labels into the BP neural network; Step S543: Calculate the network output, subtract the network output from the training set labels to obtain the network error, and determine whether the error meets the requirements; Step S544: If the error meets the accuracy requirements, training ends and the network parameters are saved; If the error does not meet the requirements, the network error is differentiated, and the derivative is used for backpropagation to correct the weight matrix and bias of the BP network in turn. Step S545: Repeat the above process until the network error meets the accuracy requirements, the network training process is completed, and a trained BP neural network model is generated.

Images