A heat supply network anomaly diagnosis method based on a machine learning algorithm

Abstract

The application provides a heat supply pipe network anomaly diagnosis method based on a machine learning algorithm, and belongs to the technical field of intelligent heat supply, and specifically comprises the following steps: identifying return water temperature by using an anomaly identification model, and when there is an abnormal return water temperature in the return water temperature, obtaining the temperature sensor corresponding to the abnormal return water temperature, and based on historical temperature data of the temperature sensor, determining that the temperature sensor does not belong to bad data, obtaining return water pipe pressure, return water pipe flow of a return water pipe corresponding to the abnormal return water temperature, and water supply pipe pressure, water supply pipe flow and water supply temperature of a water supply pipe connected with the return water pipe, and identifying bad data in combination with historical data, and when all of them do not belong to bad data, constructing an input set by using return water pipe pressure, return water pipe flow, water supply pipe pressure, water supply pipe flow and water supply temperature, and obtaining a fault diagnosis result based on a PSO-GA-SVM algorithm fault diagnosis model, so that the reliability and accuracy of the fault diagnosis result are ensured.

Description

Technical Field

[0001] This invention belongs to the field of smart heating technology, and in particular relates to a method for diagnosing anomalies in heating pipe networks based on machine learning algorithms. Background Technology

[0002] In actual engineering projects, the timing and patterns of heating network failures are often unpredictable, especially when it comes to the location of blockages or leaks. This causes great trouble for the operation and maintenance management of the heating system, seriously affects the economy and safety of the heating system, and leads to the waste of heat and water resources, resulting in increased operating costs for heating companies. Therefore, how to achieve abnormal diagnosis and location of heating network has become an urgent technical problem to be solved.

[0003] To achieve anomaly diagnosis and location in heating pipe networks, the paper "Research on Fault Diagnosis of Heating Pipe Networks Based on Simulation Technology" established a fault diagnosis model for heating pipe networks using the principle of BP neural networks. Partial operating parameters of the heating pipe network under fault conditions obtained from the simulation system were used as learning and prediction samples. Pressure and flow rate at the main pipe section were used as the node input layer, and the faulty pipe section location number was used as the node output layer to locate two types of faults. However, the following technical problems exist:

[0004] 1. The failure to screen operating parameter data means that in actual fault diagnosis of heating pipe networks, the sensors at the measuring points of the heating pipe network will inevitably fail after a long period of operation due to the harsh operating environment. Therefore, if the operating parameter data of the above-mentioned faults are not screened, the final fault diagnosis and location results will not be accurate and reliable.

[0005] 2. Failure to consider screening for abnormal return water temperatures in fault diagnosis. For heating networks, especially large heating networks, the number of nodes and measuring points that need to be observed is enormous. Therefore, if the screening for abnormal return water temperatures is not considered when determining whether to conduct fault diagnosis, the efficiency of fault diagnosis may be affected.

[0006] To address the aforementioned technical problems, this invention provides a method for diagnosing anomalies in heating pipe networks based on machine learning algorithms. Summary of the Invention

[0007] To achieve the objectives of this invention, the following technical solution is adopted:

[0008] According to one aspect of the present invention, a method for diagnosing anomalies in heating pipe networks based on machine learning algorithms is provided.

[0009] A method for diagnosing anomalies in heating pipe networks based on machine learning algorithms, characterized in that it specifically includes:

[0010] S11. The return water temperature of the heating network is acquired in real time, and an anomaly identification model based on Kmeans clustering algorithm is used to identify the return water temperature. When there is an abnormal return water temperature, the process proceeds to step S12.

[0011] S12 acquires the temperature sensor corresponding to the abnormal return water temperature, and determines whether the abnormal return water temperature is bad data based on the historical temperature data of the temperature sensor. If yes, output that the temperature sensor is abnormal; otherwise, proceed to step S13.

[0012] S13 obtains the return water pipe pressure and flow rate of the return water pipe corresponding to the abnormal return water temperature, as well as the supply water pipe pressure, flow rate, and temperature of the supply water pipe connected to the return water pipe. Combined with historical data, it identifies the return water pipe pressure, flow rate, supply water pipe pressure, flow rate, and temperature as bad data. When the return water pipe pressure, flow rate, supply water pipe pressure, flow rate, and temperature are not bad data, proceed to step S14.

[0013] S14 constructs an input set using the return water pipe pressure, return water pipe flow rate, supply water pipe pressure, supply water pipe flow rate, and supply water temperature, and transmits the input set to the fault diagnosis model based on the PSO-GA-SVM algorithm to obtain the fault diagnosis result.

[0014] By employing an anomaly identification model based on the K-means clustering algorithm to identify the return water temperature, a preliminary screening of the return water temperature is achieved. This avoids the technical problem of low diagnostic efficiency caused by real-time fault diagnosis of all return water temperatures, further ensuring overall reliability and diagnostic efficiency, and avoiding unnecessary diagnostics and waste of heating resources.

[0015] By identifying anomalies in the temperature sensor, fault diagnosis results caused by temperature sensor malfunctions are avoided, further ensuring the accuracy and reliability of the final fault diagnosis results, while also preventing unnecessary fault diagnosis.

[0016] By employing a fault diagnosis model based on the PSO-GA-SVM algorithm to determine the fault diagnosis results, the efficiency and accuracy of fault diagnosis are further improved, and the comprehensiveness of the fault diagnosis results is also guaranteed.

[0017] A further technical solution involves the following specific steps for confirming the abnormal return water temperature:

[0018] S21 acquires the return water temperature of the heating network in real time and sends the return water temperature into the anomaly identification model based on the Kmeans clustering algorithm;

[0019] S22 Based on the anomaly identification model, determine whether there are outliers in the return water temperature; if so, proceed to step S23.

[0020] S23 defines the outlier point as the abnormal return water temperature.

[0021] A further technical solution involves determining the frequency of obtaining the return water temperature of the heating network before acquiring the return water temperature based on the operating time of different return water pipes in the heating network, the complaint rate of the heating area corresponding to the return water pipe, and the number of failures of the return water pipe. The longer the operating time of the return water pipe, the higher the complaint rate of the heating area corresponding to the return water pipe, and the more failures of the return water pipe, the higher the frequency of acquiring the return water temperature of the return water pipe of the heating network.

[0022] A further technical solution is that, when the number of failures in the return water pipe exceeds a first failure frequency threshold, the formula for calculating the frequency of obtaining the return water temperature is:

[0023]

[0024] Where P1 is the base frequency, T S T represents the complaint rate of the heating area corresponding to the return water pipe. slimit is the complaint rate threshold, and [] is the rounding function.

[0025] When the number of failures in the return water pipe is less than or equal to the first failure frequency threshold, the formula for calculating the frequency of obtaining the return water temperature is:

[0026]

[0027] By updating the calculation formula for the frequency of return water temperature acquisition by combining multiple factors, the frequency of return water temperature acquisition for different return water pipes can be determined from multiple perspectives. This makes the fault diagnosis of return water pipes more targeted and ensures the stability and reliability of heating pipe operation.

[0028] A further technical solution involves the following specific steps for confirming the abnormal return water temperature data:

[0029] S31 acquires the historical temperature data corresponding to the temperature sensor, and uses the average value of the previous day's historical temperature data corresponding to the temperature sensor as the average temperature value, and determines whether the difference between the average temperature value and the abnormal return water temperature is greater than the first threshold. If yes, proceed to step S32; otherwise, determine that the abnormal return water temperature is not bad data.

[0030] S32 Based on the historical temperature data, determine whether the maximum value of the rate of change of the historical temperature data of the temperature sensor in the past day is the first rate of change threshold. If yes, proceed to step S33. If no, determine that the abnormal return water temperature is not bad data.

[0031] S33 treats the abnormal return water temperature as bad data.

[0032] By combining the rate of change and the average temperature to identify defective data, defective data can be identified from multiple perspectives, ensuring the accuracy of defective data identification and further avoiding the technical problem of poor accuracy of the final fault result due to defective data.

[0033] A further technical solution is that the first threshold is determined based on the historical temperature data of the temperature sensor and the external ambient temperature, wherein the higher the historical temperature data of the temperature sensor and the lower the external ambient temperature, the larger the first threshold is.

[0034] A further technical solution involves the following specific steps in constructing the fault diagnosis model:

[0035] S41 sets the maximum number of iterations K max =100;

[0036] S42 initializes the population parameters: population size N = 20, two learning factors c1 and c2 for the particle swarm optimization algorithm, the range of particle velocity V, the number of generations T in the particle swarm optimization algorithm, and the crossover probability P in the GA algorithm. c =0.8 and the probability of mutation P m =0.2;

[0037] S43 calculates the fitness value of the particles, and selects the accuracy under 5-fold cross-validation in SVM as the fitness value;

[0038] S44 updates the individual and population extreme values ​​of the particles based on the fitness value calculated in the previous step. If the fitness value of the particle in the k-th iteration is greater than the fitness value calculated in the previous iteration, then the individual position of the particle in this iteration is taken as the individual's optimal value; otherwise, the individual position of the particle in the previous iteration is taken as the individual's optimal value. If the fitness value of the particle in the k-th iteration is greater than the global optimal value, then the population extreme value of the previous iteration is taken as the global optimal value; otherwise, the global optimal value remains unchanged.

[0039] S45 determines whether the algorithm has fallen into a local optimum. The number of iterations is pre-set to c = 60. If the global optimum remains unchanged after exceeding this number of iterations, the algorithm is considered to have fallen into a local optimum. At this point, the genetic algorithm is executed to perform crossover and mutation operations on the population.

[0040] S46 determines whether the termination condition is met. The termination condition is selected as the preset error or number of iterations. If the termination condition is not met, the velocity and position of the particles are updated according to the formula and the process returns to step S43. If the termination condition is met, the calculation stops, and the algorithm outputs the optimal SVM parameters.

[0041] By further combining the GA algorithm to perform crossover and mutation operations on the PSO algorithm after it gets stuck in a local optimum, the efficiency and accuracy of the PSO algorithm in finding the optimum are further improved, and the accuracy of the SVM parameters is also guaranteed.

[0042] A further technical solution is that the optimal SVM parameters are the values ​​of the penalty factor C of the SVM algorithm and the parameter g of the kernel function.

[0043] A further technical solution is that the formula for calculating the inertia weight of the PSO algorithm is:

[0044]

[0045] Where w max w is the maximum value of the inertia weight. min T represents the minimum value of the inertia weight, where t is the current iteration number. max The maximum number of iterations is given by `min()`, which is the function to find the minimum value, and `rand(0, 1)` is a random number that takes values ​​between 0 and 1.

[0046] The formula for calculating the learning factor of the PSO algorithm is as follows:

[0047]

[0048]

[0049] c1 and c2 are learning factors.

[0050] By optimizing the inertia weight and learning factor, the efficiency and accuracy of the PSO algorithm for optimization are further improved, and the technical problem of low optimization efficiency and accuracy caused by getting stuck in local optima is further reduced.

[0051] On the other hand, this application provides a computer-readable storage medium storing a computer program thereon, which, when executed in a computer, causes the computer to perform the above-mentioned method for diagnosing abnormalities in a heating pipe network based on a machine learning algorithm.

[0052] On the other hand, this application provides a computer program product, characterized in that the computer program product stores instructions, which, when executed by a computer, cause the computer to implement the above-mentioned method for diagnosing abnormalities in heating pipe networks based on machine learning algorithms. Attached Figure Description

[0053] The above and other features and advantages of the present invention will become more apparent from a detailed description of exemplary embodiments thereof with reference to the accompanying drawings.

[0054] Figure 1 This is a flowchart of a heating network anomaly diagnosis method based on machine learning algorithm according to Embodiment 1.

[0055] Figure 2 This is a framework diagram of a computer-readable storage medium according to Embodiment 2. Detailed Implementation

[0056] Exemplary embodiments will now be described more fully with reference to the accompanying drawings. However, these exemplary embodiments can be implemented in many ways and should not be construed as limited to the embodiments set forth herein; rather, they are provided so that the invention will be thorough and complete, and the concept of the exemplary embodiments will be fully conveyed to those skilled in the art. The same reference numerals in the drawings denote the same or similar structures, and therefore their detailed description will be omitted.

[0057] The terms “a,” “one,” “the,” and “the” are used to indicate the existence of one or more elements / components / etc.; the terms “including” and “having” are used to indicate an open-ended meaning of inclusion and that other elements / components / etc. may exist in addition to the listed elements / components / etc.

[0058] When scheduling different energy storage methods in the past, it was often impossible to optimize the scheduling between different energy storage methods according to their characteristics and load regulation needs. As a result, different energy storage methods could not be fully utilized, and the flexibility of the unit and the safety of the equipment were also affected.

[0059] Example 1

[0060] To solve the above problems, according to one aspect of the present invention, such as Figure 1 As shown, a method for diagnosing anomalies in heating pipe networks based on machine learning algorithms is provided, characterized by specifically including:

[0061] S11. The return water temperature of the heating network is acquired in real time, and an anomaly identification model based on Kmeans clustering algorithm is used to identify the return water temperature. When there is an abnormal return water temperature, the process proceeds to step S12.

[0062] S12 acquires the temperature sensor corresponding to the abnormal return water temperature, and determines whether the abnormal return water temperature is bad data based on the historical temperature data of the temperature sensor. If yes, output that the temperature sensor is abnormal; otherwise, proceed to step S13.

[0063] S13 obtains the return water pipe pressure and flow rate of the return water pipe corresponding to the abnormal return water temperature, as well as the supply water pipe pressure, flow rate, and temperature of the supply water pipe connected to the return water pipe. Combined with historical data, it identifies the return water pipe pressure, flow rate, supply water pipe pressure, flow rate, and temperature as bad data. When the return water pipe pressure, flow rate, supply water pipe pressure, flow rate, and temperature are not bad data, proceed to step S14.

[0064] S14 constructs an input set using the return water pipe pressure, return water pipe flow rate, supply water pipe pressure, supply water pipe flow rate, and supply water temperature, and transmits the input set to the fault diagnosis model based on the PSO-GA-SVM algorithm to obtain the fault diagnosis result.

[0065] By employing an anomaly identification model based on the K-means clustering algorithm to identify the return water temperature, a preliminary screening of the return water temperature is achieved. This avoids the technical problem of low diagnostic efficiency caused by real-time fault diagnosis of all return water temperatures, further ensuring overall reliability and diagnostic efficiency, and avoiding unnecessary diagnostics and waste of heating resources.

[0066] By identifying anomalies in the temperature sensor, fault diagnosis results caused by temperature sensor malfunctions are avoided, further ensuring the accuracy and reliability of the final fault diagnosis results, while also preventing unnecessary fault diagnosis.

[0067] By employing a fault diagnosis model based on the PSO-GA-SVM algorithm to determine the fault diagnosis results, the efficiency and accuracy of fault diagnosis are further improved, and the comprehensiveness of the fault diagnosis results is also guaranteed.

[0068] For example, the specific steps for confirming the abnormal return water temperature are as follows:

[0069] S21 acquires the return water temperature of the heating network in real time and sends the return water temperature into the anomaly identification model based on the Kmeans clustering algorithm;

[0070] S22 Based on the anomaly identification model, determine whether there are outliers in the return water temperature; if so, proceed to step S23.

[0071] S23 defines the outlier point as the abnormal return water temperature.

[0072] In another possible embodiment, for the return water temperatures of the heating grid being 50 degrees, 25 degrees, 51 degrees, and 61 degrees respectively, an anomaly identification model based on the K-means clustering algorithm can be used to identify outliers with an abnormal return water temperature of 25 degrees, and these outliers can be used as abnormal return water temperatures.

[0073] For example, before obtaining the return water temperature of the heating network, it is necessary to determine the frequency of obtaining the return water temperature of the heating network based on the operating time of different return water pipes of the heating network, the complaint rate of the heating area corresponding to the return water pipe, and the number of failures of the return water pipe. The longer the operating time of the return water pipe, the higher the complaint rate of the heating area corresponding to the return water pipe, and the more failures of the return water pipe, the higher the frequency of obtaining the return water temperature of the return water pipe of the heating network.

[0074] For a specific example, when the number of failures in the return water pipe exceeds a first failure threshold, the formula for calculating the frequency of obtaining the return water temperature is:

[0075]

[0076] Where P1 is the base frequency, T S T represents the complaint rate of the heating area corresponding to the return water pipe. slimit is the complaint rate threshold, and [] is the rounding function.

[0077] When the number of failures in the return water pipe is less than or equal to the first failure frequency threshold, the formula for calculating the frequency of obtaining the return water temperature is:

[0078]

[0079] By updating the calculation formula for the frequency of return water temperature acquisition by combining multiple factors, the frequency of return water temperature acquisition for different return water pipes can be determined from multiple perspectives. This makes the fault diagnosis of return water pipes more targeted and ensures the stability and reliability of heating pipe operation.

[0080] For example, the specific steps for confirming abnormal return water temperature data are as follows:

[0081] S31 acquires the historical temperature data corresponding to the temperature sensor, and uses the average value of the previous day's historical temperature data corresponding to the temperature sensor as the average temperature value, and determines whether the difference between the average temperature value and the abnormal return water temperature is greater than the first threshold. If yes, proceed to step S32; otherwise, determine that the abnormal return water temperature is not bad data.

[0082] S32 Based on the historical temperature data, determine whether the maximum value of the rate of change of the historical temperature data of the temperature sensor in the past day is the first rate of change threshold. If yes, proceed to step S33. If no, determine that the abnormal return water temperature is not bad data.

[0083] S33 treats the abnormal return water temperature as bad data.

[0084] By combining the rate of change and the average temperature to identify defective data, defective data can be identified from multiple perspectives, ensuring the accuracy of defective data identification and further avoiding the technical problem of poor accuracy of the final fault result due to defective data.

[0085] For example, the first threshold is determined based on the historical temperature data of the temperature sensor and the external ambient temperature. The higher the historical temperature data of the temperature sensor and the lower the external ambient temperature, the larger the first threshold.

[0086] For example, the specific steps for constructing the fault diagnosis model are as follows:

[0087] S41 sets the maximum number of iterations K max =100;

[0088] S42 initializes the population parameters: population size N = 20, two learning factors c1 and c2 for the particle swarm optimization algorithm, the range of particle velocity V, the number of generations T in the particle swarm optimization algorithm, and the crossover probability P in the GA algorithm. c =0.8 and the probability of mutation P m =0.2;

[0089] S43 calculates the fitness value of the particles, and selects the accuracy under 5-fold cross-validation in SVM as the fitness value;

[0090] S44 updates the individual and population extreme values ​​of the particles based on the fitness value calculated in the previous step. If the fitness value of the particle in the k-th iteration is greater than the fitness value calculated in the previous iteration, then the individual position of the particle in this iteration is taken as the individual's optimal value; otherwise, the individual position of the particle in the previous iteration is taken as the individual's optimal value. If the fitness value of the particle in the k-th iteration is greater than the global optimal value, then the population extreme value of the previous iteration is taken as the global optimal value; otherwise, the global optimal value remains unchanged.

[0091] S45 determines whether the algorithm has fallen into a local optimum. The number of iterations is pre-set to c = 60. If the global optimum remains unchanged after exceeding this number of iterations, the algorithm is considered to have fallen into a local optimum. At this point, the genetic algorithm is executed to perform crossover and mutation operations on the population.

[0092] S46 determines whether the termination condition is met. The termination condition is selected as the preset error or number of iterations. If the termination condition is not met, the velocity and position of the particles are updated according to the formula and the process returns to step S43. If the termination condition is met, the calculation stops, and the algorithm outputs the optimal SVM parameters.

[0093] By further combining the GA algorithm to perform crossover and mutation operations on the PSO algorithm after it gets stuck in a local optimum, the efficiency and accuracy of the PSO algorithm in finding the optimum are further improved, and the accuracy of the SVM parameters is also guaranteed.

[0094] For example, the optimal SVM parameters are the values ​​of the penalty factor C of the SVM algorithm and the parameter g of the kernel function.

[0095] For example, the formula for calculating the inertia weight in the PSO algorithm is as follows:

[0096]

[0097] Where w max w is the maximum value of the inertia weight. min T represents the minimum value of the inertia weight, where t is the current iteration number. max The maximum number of iterations is given by `min()`, which is the function to find the minimum value, and `rand(0, 1)` is a random number that takes values ​​between 0 and 1.

[0098] The formula for calculating the learning factor of the PSO algorithm is as follows:

[0099]

[0100]

[0101] c1 and c2 are learning factors.

[0102] By optimizing the inertia weight and learning factor, the efficiency and accuracy of the PSO algorithm for optimization are further improved, and the technical problem of low optimization efficiency and accuracy caused by getting stuck in local optima is further reduced.

[0103] Example 2

[0104] like Figure 2 As shown, this application provides a computer-readable storage medium storing a computer program thereon. When the computer program is executed in a computer, it causes the computer to perform the above-described method for diagnosing abnormalities in a heating network based on a machine learning algorithm.

[0105] Example 3

[0106] This application provides a computer program product, characterized in that the computer program product stores instructions, which, when executed by a computer, cause the computer to implement the above-described method for diagnosing abnormalities in a heating pipe network based on a machine learning algorithm.

[0107] In this embodiment of the invention, the term "multiple" refers to two or more, unless otherwise explicitly defined. The terms "install," "connect," and "fix" should be interpreted broadly. For example, "connect" can mean a fixed connection, a detachable connection, or an integral connection. Those skilled in the art can understand the specific meaning of the above terms in this embodiment of the invention based on the specific circumstances.

[0108] In the description of the embodiments of the present invention, it should be understood that the terms "upper" and "lower" indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are only for the convenience of describing the embodiments of the present invention and simplifying the description, and do not indicate or imply that the device or unit referred to must have a specific orientation or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on the embodiments of the present invention.

[0109] In the description of this specification, the terms "an embodiment," "a preferred embodiment," etc., refer to a specific feature, structure, material, or characteristic described in connection with that embodiment or example, which is included in at least one embodiment or example of the present invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0110] The above are merely preferred embodiments of the present invention and are not intended to limit the scope of the present invention. For those skilled in the art, the embodiments of the present invention can have various modifications and variations. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the embodiments of the present invention should be included within the protection scope of the embodiments of the present invention.

Claims

1. A method for diagnosing anomalies in heating pipe networks based on machine learning algorithms, characterized in that, Specifically, it includes: S11. The return water temperature of the heating network is acquired in real time, and an anomaly identification model based on Kmeans clustering algorithm is used to identify the return water temperature. When there is an abnormal return water temperature, the process proceeds to step S12. S12 acquires the temperature sensor corresponding to the abnormal return water temperature, and determines whether the abnormal return water temperature is bad data based on the historical temperature data of the temperature sensor. If yes, output that the temperature sensor is abnormal; otherwise, proceed to step S13. S13 obtains the return water pipe pressure and flow rate of the return water pipe corresponding to the abnormal return water temperature, as well as the supply water pipe pressure, flow rate, and temperature of the supply water pipe connected to the return water pipe. Combined with historical data, it identifies the return water pipe pressure, flow rate, supply water pipe pressure, flow rate, and temperature as bad data. When the return water pipe pressure, flow rate, supply water pipe pressure, flow rate, and temperature are not bad data, proceed to step S14. S14 constructs an input set using the return water pipe pressure, return water pipe flow rate, supply water pipe pressure, supply water pipe flow rate, and supply water temperature, and transmits the input set to the fault diagnosis model based on the PSO-GA-SVM algorithm to obtain the fault diagnosis result; Before obtaining the return water temperature of the heating network, it is necessary to determine the frequency of obtaining the return water temperature of the return water pipe based on the operating time of different return water pipes of the heating network, the complaint rate of the heating area corresponding to the return water pipe, and the number of failures of the return water pipe. The longer the operating time of the return water pipe, the higher the complaint rate of the heating area corresponding to the return water pipe, and the more failures of the return water pipe, the higher the frequency of obtaining the return water temperature of the return water pipe of the heating network. When the number of failures in the return water pipe exceeds a first failure frequency threshold, the formula for calculating the frequency of obtaining the return water temperature is: Where P1 is the base acquisition frequency, TS is the complaint rate of the heating area corresponding to the return water pipe, Tslimit is the complaint rate threshold, and [] is the rounding function; when the number of failures of the return water pipe is less than or equal to the first failure number threshold, the calculation formula for the acquisition frequency of the return water temperature is: The specific steps for confirming abnormal return water temperature data are as follows: Obtain the historical temperature data corresponding to the temperature sensor, and take the average value of the previous day's historical temperature data corresponding to the temperature sensor as the average temperature value. Determine whether the difference between the average temperature value and the abnormal return water temperature is greater than a first threshold. If yes, proceed to the next step; otherwise, determine that the abnormal return water temperature is not bad data. Based on the historical temperature data, determine whether the maximum value of the rate of change of the historical temperature data of the temperature sensor in the past day is the first rate of change threshold. If yes, proceed to the next step; otherwise, determine that the abnormal return water temperature is not bad data. The abnormal return water temperature is considered as bad data.

2. The method for diagnosing heating network anomalies based on machine learning algorithms as described in claim 1, characterized in that, The specific steps for confirming the abnormal return water temperature are as follows: The return water temperature of the heating network is acquired in real time, and the return water temperature is sent to the anomaly identification model based on the Kmeans clustering algorithm. Based on the anomaly identification model, determine whether there are outliers in the return water temperature; if so, proceed to the next step. The outlier points are considered as abnormal return water temperatures.

3. The method for diagnosing heating network anomalies based on machine learning algorithms as described in claim 1, characterized in that, The first threshold is determined based on the historical temperature data of the temperature sensor and the external ambient temperature. The higher the historical temperature data of the temperature sensor and the lower the external ambient temperature, the larger the first threshold.

4. The method for diagnosing heating network anomalies based on machine learning algorithms as described in claim 1, characterized in that, The specific steps for constructing the fault diagnosis model are as follows: S41 sets the maximum number of iterations Kmax = 100; S42 initializes the population parameters: population size N = 20, two learning factors c1 and c2 for the particle swarm optimization algorithm, the range of particle flight velocity V, the number of generations T in the particle swarm optimization algorithm, and the crossover probability Pc = 0.8 and mutation probability Pm = 0.2 in the GA algorithm. S43 calculates the fitness value of the particles, and selects the accuracy under 5-fold cross-validation in SVM as the fitness value; S44 updates the individual and population extreme values ​​of the particles based on the fitness value calculated in the previous step. If the fitness value of the particle in the k-th iteration is greater than the fitness value calculated in the previous iteration, then the individual position of the particle in this iteration is taken as the individual's optimal value; otherwise, the individual position of the particle in the previous iteration is taken as the individual's optimal value. If the fitness value of the particle in the k-th iteration is greater than the global optimal value, then the population extreme value of the previous iteration is taken as the global optimal value; otherwise, the global optimal value remains unchanged. S45 determines whether the algorithm has fallen into a local optimum. The number of iterations is pre-set to c = 60. If the global optimum remains unchanged after exceeding this number of iterations, the algorithm is considered to have fallen into a local optimum. At this point, the genetic algorithm is executed to perform crossover and mutation operations on the population. S46 determines whether the termination condition is met. The termination condition is selected as the preset error or number of iterations. If the termination condition is not met, the velocity and position of the particles are updated according to the formula and the process returns to step S43. If the termination condition is met, the calculation stops, and the algorithm outputs the optimal SVM parameters.

5. The method for diagnosing heating network anomalies based on machine learning algorithms as described in claim 4, characterized in that, The optimal SVM parameters are the values ​​of the penalty factor C of the SVM algorithm and the parameter g of the kernel function.

6. In the heating network anomaly diagnosis method based on machine learning algorithm as described in claim 4, the calculation formula for the inertia weight of the PSO algorithm is as follows: Where wmax is the maximum value of the inertia weight, wmin represents the minimum value of the inertia weight, t is the current iteration number, Tmax is the maximum iteration number, min() is the minimum value function, and rand(0,1) is a random number between 0 and 1; the formula for calculating the learning factor of the PSO algorithm is: c1 and c2 are learning factors.

7. A computer-readable storage medium having a computer program stored thereon, which, when executed in a computer, causes the computer to perform a method for diagnosing abnormalities in a heating network based on a machine learning algorithm as described in any one of claims 1-6.

Images