Power plant equipment fault diagnosis method and system based on Internet of Things

Abstract

The invention relates to the technical field of industrial Internet of Things equipment fault diagnosis, and discloses a power plant equipment fault diagnosis method and system based on the Internet of Things, and the method comprises the steps: obtaining an equipment working condition monitoring data set, executing coupling analysis and abnormal triggering collection, and obtaining a multi-equipment association state data set; performing node analysis and path distribution based on the data set to form a fault diffusion path scheme; carrying out risk calculation and trend reasoning to obtain a real-time fault risk assessment result; according to an evaluation result, searching an optimal parameter combination from a preset parameter library and generating a parameter instruction set; after the instruction set is executed, initial operation state data are collected, delay evaluation and parameter correction are carried out, and final operation state data are obtained; and performing difference analysis and risk scoring according to the state data, adjusting a path weight, and updating a fault diffusion path scheme. The method can solve the problem of fault diagnosis lag.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of industrial internet of things equipment fault diagnosis, and in particular to a power plant equipment fault diagnosis method and system based on the internet of things. BACKGROUND

[0002] With the in-depth application of industrial internet of things (IIoT) in the energy field, large industrial systems such as power plants begin to widely deploy sensor networks and data acquisition systems for real-time monitoring of equipment operating status, assessment of potential risks and assistance in decision optimization. The core goal of such technology is to improve the reliability, safety and intelligent maintenance capability of equipment operation, which is an important foundation for realizing intelligent power plant operation and maintenance.

[0003] Existing fault diagnosis methods generally rely on single-device monitoring and state evaluation in the internet of things environment, that is, by collecting equipment operating data (such as temperature, pressure, vibration, etc.) and comparing with historical abnormal patterns to judge the equipment health status. However, in actual power plant operation, there is a high coupling relationship between devices, and the abnormality of one device may affect multiple other devices through pipe networks, loads or control systems, forming a fault diffusion chain. The existing method lacks recognition and modeling of such "interaction between devices", which easily causes incomplete risk identification. Therefore, the existing method has the problem of fault diagnosis lag in the face of dynamic changes in operating conditions, and it is difficult to adjust key operating parameters in time to avoid risks. SUMMARY

[0004] The present application provides a power plant equipment fault diagnosis method and system based on the internet of things to solve the problem of fault diagnosis lag.

[0005] In a first aspect, to solve the above technical problems, the present application provides a power plant equipment fault diagnosis method based on the internet of things, comprising:

[0006] Obtaining each device working condition monitoring data set;

[0007] According to the each device working condition monitoring data set, coupling analysis and abnormal trigger collection are performed to obtain a multi-device associated state data set;

[0008] According to the multi-device associated state data set, node analysis and path allocation are performed to obtain a fault diffusion path scheme;

[0009] According to the fault diffusion path scheme, risk calculation and trend reasoning are performed to obtain a real-time fault risk evaluation result;

[0010] Based on the real-time fault risk evaluation result, an optimal parameter combination is searched in a preset device parameter library, and an instruction set is generated for the optimal parameter combination, thereby obtaining a parameter instruction set for high-risk equipment.

[0011] Executing the parameter instruction set, collecting initial operating status data, and performing delay evaluation and parameter correction on the initial operating status data to obtain final operating status data after equipment adjustment;

[0012] According to the final operating state data, a state difference analysis and a risk score calculation are performed, and based on the risk score, the path weights in the fault diffusion path scheme are numerically updated to obtain a final fault diffusion path scheme.

[0013] Preferably, the coupling analysis and abnormal triggering acquisition are performed based on the data sets of the working condition monitoring of each device to obtain a data set of multi-device associated status, including:

[0014] The equipment operating condition monitoring data sets include: generator vibration frequency, boiler temperature gradient change and cooling system pressure fluctuation;

[0015] Performing time synchronization processing on the equipment operating condition monitoring data set to obtain a time-synchronized operating condition data set;

[0016] Performing coupling degree calculation and matrix construction on the time-synchronized working condition data set to obtain an inter-device coupling degree matrix;

[0017] If the generator vibration frequency of the time-synchronized operating condition dataset exceeds a preset frequency threshold, the device corresponding to the generator vibration frequency is marked as an abnormal device, and the boiler temperature gradient change and cooling system pressure fluctuation in the current time window of the abnormal device are additionally collected as extended data;

[0018] The extended data and the inter-device coupling matrix are subjected to dimensionality reduction processing to extract the main correlation features between multiple devices and obtain a multi-device correlation status data set.

[0019] Preferably, performing node analysis and path allocation based on the multi-device association status data set to obtain a fault diffusion path solution includes:

[0020] Obtain historical failure mode library;

[0021] Sorting the importance of the multi-device association status data set to obtain an importance ranking table of the device nodes;

[0022] Based on the importance ranking table and in combination with the historical fault mode library, similarity calculation is performed one by one. If the similarity exceeds a preset similarity threshold, parameter fluctuation data of the device is collected;

[0023] Performing time series analysis on the parameter fluctuation data, determining a fault propagation time window, and combining the inter-device coupling degree matrix to construct a path structure and calculate path weights, obtaining a fault diffusion path scheme.

[0024] Preferably, the risk calculation and trend inference according to the fault diffusion path scheme obtain real-time fault risk assessment results, including:

[0025] Mapping the fault diffusion path scheme to the corresponding device nodes and performing risk probability calculation to obtain the risk probability of each device node;

[0026] If the risk probability is greater than a preset probability threshold, the device node corresponding to the risk probability is marked as an abnormal node, and the structural position of the abnormal node in the fault diffusion path scheme is retrieved, while the risk diffusion trend is inferred to obtain real-time fault risk assessment results.

[0027] Preferably, based on the real-time fault risk assessment results, the optimal parameter combination is searched in the preset device parameter library, and the instruction set is generated for the optimal parameter combination, thereby obtaining the parameter instruction set for high-risk devices, including:

[0028] Based on the real-time fault risk assessment results, select a parameter combination in the preset device parameter library, perform a running state prediction to obtain a corresponding state prediction result, and determine the optimal parameter combination according to the state prediction result;

[0029] Based on the optimal parameter combination, perform instruction generation and step adjustment to obtain the parameter instruction set for high-risk devices.

[0030] Preferably, the parameter instruction set is executed to collect initial running state data, and the initial running state data is subjected to delay evaluation and parameter correction to obtain the final running state data of the device after adjustment, including:

[0031] The parameter instruction set is executed, and initial running state data is collected;

[0032] The initial running state data is subjected to delay calculation, and if the delay is greater than a preset delay threshold, the initial running state data is subjected to parameter correction and adjustment to obtain the final running state data of the device after adjustment.

[0033] Preferably, the final running state data is subjected to state difference analysis and risk score calculation, and the numerical value of the path weight in the fault diffusion path scheme is updated based on the risk score, thereby obtaining a final fault diffusion path scheme, including:

[0034] Obtain a historical benchmark data set;

[0035] Based on the historical benchmark data set, calculating the state difference value of the final operating state data to obtain a state difference value;

[0036] Calculating a risk score based on the state difference value to obtain a risk score;

[0037] If the risk score is greater than a preset risk threshold, the path weights in the fault diffusion path solution are adjusted to obtain a final fault diffusion path solution.

[0038] In a second aspect, the present invention provides a power plant equipment fault diagnosis system based on the Internet of Things, comprising:

[0039] Data acquisition module, used to obtain data sets for monitoring the working conditions of each device;

[0040] A multi-device module is used to perform coupling analysis and abnormal triggering collection based on the working condition monitoring data sets of each device to obtain a multi-device associated status data set;

[0041] A path plan module, configured to perform node analysis and path allocation based on the multi-device association status data set to obtain a fault diffusion path plan;

[0042] A risk assessment module is used to perform risk calculation and trend reasoning based on the fault diffusion path scheme to obtain real-time fault risk assessment results;

[0043] An instruction parameter set module is used to search for an optimal parameter combination in a preset device parameter library based on the real-time fault risk assessment result, and generate an instruction set for the optimal parameter combination, thereby obtaining a parameter instruction set for high-risk equipment;

[0044] An operating status module is used to execute the parameter instruction set, collect initial operating status data, and perform delay evaluation and parameter correction on the initial operating status data to obtain final operating status data after equipment adjustment;

[0045] The final path solution module is used to perform state difference analysis and risk score calculation according to the final operating state data, and based on the risk score, numerically update the path weights in the fault diffusion path solution to obtain a final fault diffusion path solution.

[0046] In a third aspect, the present invention also provides an electronic device comprising a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein when the processor executes the computer program, it implements any one of the above-mentioned methods for diagnosing power plant equipment faults based on the Internet of Things.

[0047] In a fourth aspect, the present invention also provides a computer-readable storage medium, which includes a stored computer program, wherein when the computer program is running, the device where the computer-readable storage medium is located is controlled to execute any one of the above-mentioned power plant equipment fault diagnosis methods based on the Internet of Things.

[0048] Compared with the prior art, the present invention has the following beneficial effects:

[0049] (1) This method acquires the vibration frequency, temperature, and pressure data of each device and performs time synchronization and coupling analysis to extract the main correlation characteristics between devices. Compared with traditional independent analysis methods, this method can better reflect the characteristics of device linkage operation and effectively identify abnormal linkage situations, thereby improving the accuracy and coverage of fault location.

[0050] (2) Based on the identification of the association status of multiple devices, the present invention combines the historical fault pattern library and the device importance ranking to dynamically allocate and weight paths. This method can accurately deduce the path of fault propagation and the order of impact, avoiding misdiagnosis or missed diagnosis due to path judgment errors.

[0051] (3) After the parameters are issued, the present invention collects the initial response data of the device and calculates whether the response delay exceeds the preset range. If the delay is abnormal, the parameter combination is re-optimized based on the feedback data and a correction instruction is issued. This operation process ensures that the parameter adjustment has dynamic response capability and adaptability, significantly reducing the risk of secondary failure caused by adjustment lag.

[0052] (4) This method performs a state difference analysis on the final operating state data, calculates a fault risk score, and updates the key weights in the path plan based on the score results. Compared with static rule paths, this method can automatically correct fault propagation trends based on actual operation, enhancing the diagnostic system's adaptability to continuously changing operating conditions. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] Figure 1 This is a flow chart of a method for diagnosing power plant equipment faults based on the Internet of Things provided by the first embodiment of the present invention;

[0054] Figure 2 This is a structural diagram of a power plant equipment fault diagnosis system based on the Internet of Things provided by the second embodiment of the present invention. DETAILED DESCRIPTION

[0055] With reference to the drawings of the embodiments of the present application, the technical solutions in the embodiments of the present application will be clearly and completely described, obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of protection of the present application.

[0056] With reference to Figure 1 The first embodiment of the present application provides a power plant equipment fault diagnosis method based on Internet of Things, comprising the following steps:

[0057] S11, obtaining each equipment working condition monitoring data set;

[0058] S12, according to the each equipment working condition monitoring data set, coupling analysis and abnormal triggering collection are carried out, and a multi-equipment correlation state data set is obtained;

[0059] S13, according to the multi-equipment correlation state data set, node analysis and path distribution are carried out, and a fault diffusion path scheme is obtained;

[0060] S14, according to the fault diffusion path scheme, risk calculation and trend reasoning are carried out, and a real-time fault risk assessment result is obtained;

[0061] S15, based on the real-time fault risk assessment result, searching for an optimal parameter combination in a preset equipment parameter library, and generating an instruction set for the optimal parameter combination, so as to obtain a parameter instruction set for high-risk equipment;

[0062] S16, executing the parameter instruction set, collecting initial running state data, and performing delay evaluation and parameter correction on the initial running state data, so as to obtain the final running state data of the equipment after adjustment;

[0063] S17, according to the final running state data, performing state difference analysis and risk score calculation, and based on the risk score, updating the path weight in the fault diffusion path scheme, so as to obtain a final fault diffusion path scheme.

[0064] In step S11, each equipment working condition monitoring data set is obtained.

[0065] It is worth noting that each equipment working condition monitoring data set includes generator vibration frequency, boiler temperature gradient change and cooling system pressure fluctuation.

[0066] For example, the generator's vibration frequency is acquired using an accelerometer mounted on the generator's bearing housing. This sensor can sense the mechanical vibrations generated during operation in real time. The sensor's sampling frequency is set to 1kHz, capturing subtle fluctuations in frequency between 0 and 50Hz. For example, if the generator is unbalanced or misaligned, the vibration frequency will rapidly change. The accelerometer immediately records this value and transmits it to an edge acquisition node via RS485 or CAN bus.

[0067] To detect changes in boiler temperature gradients, multiple thermocouple sensors need to be arranged and placed at different heights in the boiler or at the inlet and outlet of the thermal channel to form temperature difference data pairs. These thermocouples monitor the temperature values ​​of the corresponding points in real time, and by setting a 10s interval to take the difference, the temperature gradient can be obtained. The 10s time interval difference is based on the fact that boiler temperature changes relatively slowly. The 10-second interval is sufficient to capture effective gradient changes while avoiding data redundancy and system burden caused by frequent sampling. Taking the K-type thermocouple commonly used in industrial applications as an example, its measurement accuracy can reach ±1.5°C, and the response time does not exceed 1 second. The data acquisition unit will continuously receive temperature values ​​from the upper and lower measuring points.

[0068] Monitoring cooling system pressure fluctuations relies on pressure transmitters installed in the cooling water piping. These transmitters are typically installed at the water inlet and outlet to measure pressure changes over time. A typical configuration uses a capacitive pressure sensor with a 0–5 MPa range, with a sampling period set to once per second. Data is uploaded via a 4–20 mA analog signal or Modbus RTU protocol and stored in a local cache. If the cooling system experiences blockages or pump flow changes, the pressure transmitter will reflect these transient fluctuations in real time.

[0069] In step S12, coupling analysis and abnormal triggering collection are performed based on the equipment working condition monitoring data set to obtain a multi-equipment associated state data set, including:

[0070] The equipment operating condition monitoring data sets include: generator vibration frequency, boiler temperature gradient change and cooling system pressure fluctuation;

[0071] Performing time synchronization processing on the equipment operating condition monitoring data set to obtain a time-synchronized operating condition data set;

[0072] Performing coupling degree calculation and matrix construction on the time-synchronized working condition data set to obtain an inter-device coupling degree matrix;

[0073] If the generator vibration frequency of the time synchronization working condition data set exceeds the preset frequency threshold, the device corresponding to the generator vibration frequency is marked as an abnormal device, and the boiler temperature gradient change and the cooling system pressure fluctuation of the current time window of the abnormal device are additionally collected as extended data.

[0074] The extended data is subjected to dimension reduction processing with the coupling degree matrix between devices to extract main correlation characteristics between multiple devices, and a multi-device correlation state data set is obtained.

[0075] It is worth noting that in actual engineering implementation, the NTP network time protocol is used to realize time synchronization. Specifically, all sensor collection terminals arranged on the power plant site are internally provided with independent local clock modules, and these clocks are calibrated by a central time server to form a unified time reference. The time server is connected to an external GPS satellite or a national time center, and broadcasts time signals through an in-plant Ethernet or 5G industrial communication network. After receiving the time calibration signal, the sensor compares and adjusts its own clock with the reference time to control the error within ±50 milliseconds. The collected data are attached to each data record in the form of Unix timestamp, which is the UTC time accurate to milliseconds, for example, "2025-08-18T14:23:15.357". In this way, during subsequent data fusion, the time stamps can be compared to realize accurate point-to-point alignment, ensuring that different types of data can be placed in the same time window for analysis and processing. For example, the vibration frequency of the generator and the pressure data of the cooling system differ by no more than 100 milliseconds in time, and can be regarded as the joint state in the same time period.

[0076] The degree of coupling is calculated for each synchronized device data. This measure measures the strength of mutual influence between devices and is calculated using the Pearson correlation coefficient. First, the average values ​​of each parameter sequence are calculated. For example, the average values ​​of the vibration frequency sequence and the boiler temperature gradient sequence collected over a certain period are taken. Next, the difference between each data point and its average value is calculated, multiplied, and the corresponding products at all moments are summed to determine the degree of covariance between the two parameters. Next, the fluctuation range of each parameter is calculated. This is done by squaring the difference between each value and the average value, summing the squares, and taking the mean of the sum of squares. Finally, the degree of covariance between the two parameters is compared with their own fluctuations to derive a correlation coefficient reflecting the trend of their synchronization. After performing the Pearson correlation coefficient calculation for all devices, a complete coupling matrix is ​​constructed. This matrix is ​​a two-dimensional table with devices as rows and columns, and the horizontal and vertical axes list the names or numbers of all monitored devices. For example, if the system includes five key devices—a generator, a boiler, a cooling water pump, and a transformer—then the matrix would be 5 rows by 5 columns. Each cell in the matrix corresponds to a correlation value between two devices and is populated with the correlation coefficient between their corresponding parameters.

[0077] If the generator vibration frequency in the equipment condition monitoring data exceeds a preset frequency threshold within a certain period, the device is marked as abnormal. This preset frequency threshold is set at 12Hz, based on the equipment's factory standard. For example, under normal operation, the vibration frequency of a certain generator model remains stable between 8 and 11Hz. If it exceeds 12Hz, it may indicate a potential fault hazard, such as looseness or imbalance in the mechanical structure. Therefore, setting 12Hz as the fault warning threshold can both detect fault signs in advance and avoid false alarms.

[0078] When a device is marked as abnormal, the system immediately performs extended data collection. Specifically, if the generator vibration frequency exceeds the limit, the system automatically searches for other devices with a high degree of coupling with the generator, such as boilers and cooling systems, and collects boiler temperature gradient changes and cooling system pressure fluctuations from these devices in the current time window. The collection operation is completed through IoT sensors, and the device collection frequency is 1Hz, that is, it updates once per second. The extended data collection time window is set to within 10 seconds after the abnormality occurs to ensure that the acquired extended data can effectively reflect the immediate dynamic response behavior between devices.

[0079] All the expansion data and coupling data are normalized according to a unified standard, and the Z-score standardization method is adopted, that is, each data is subtracted by the sample mean and then divided by the standard deviation, so that the data of different dimensions have a unified scale, which is convenient for subsequent covariance operation. After normalization, the system calculates the covariance matrix based on the standardized data. Each element of the covariance matrix represents the cooperative change trend between any two groups of parameters, that is, if one parameter changes and another parameter also increases, the covariance is positive, and if not, the covariance is negative.

[0080] Next, the system solves the eigenvalues and eigenvectors according to the covariance matrix. Operationally, the system first constructs the characteristic equation corresponding to the covariance matrix, that is, subtracts the unit matrix multiplied by λ from the covariance matrix, and sets the determinant to zero, and then solves all the eigenvalues λ. Subsequently, for each eigenvalue, substitute it into the covariance matrix to solve its corresponding non-zero solution vector, which is the eigenvector of the eigenvalue. The eigenvalue represents the size of the variance of the original data explained by the corresponding eigenvector, and the larger the value, the more information carried in the direction of the vector.

[0081] The system sorts the eigenvalues in descending order according to their sizes, and selects the eigenvectors with cumulative contribution rate of more than 80% as the principal components output in turn, to complete the dimensionality reduction operation. Using these principal components to construct a matrix, the multi-device correlation state data set can be obtained, and the most core change trend in the original expansion data and coupling information is retained, such as reflecting the linkage behavior of temperature and pressure when the vibration frequency changes sharply, thereby improving the efficiency and operability of fault correlation analysis.

[0082] In step S13, according to the multi-device correlation state data set, node analysis and path allocation are performed to obtain a fault diffusion path scheme, including:

[0083] Obtaining a historical fault mode library;

[0084] According to the multi-device correlation state data set, importance ranking is performed to obtain an importance ranking table of device nodes;

[0085] Based on the importance ranking table, combined with the historical fault mode library, similarity calculation is performed one by one, and if the similarity exceeds a preset similarity threshold, parameter fluctuation data of the device is collected;

[0086] Performing time series analysis on the parameter fluctuation data to determine a fault propagation time window, and combining the coupling degree matrix between devices, path structure construction and path weight calculation are performed to obtain a fault diffusion path scheme.

[0087] It is worth noting that a historical fault pattern library is obtained for fault analysis. In specific implementation, the historical fault pattern library consists of fault records, alarm logs, and manual inspection data in the operation and maintenance system, which are processed in a unified format to form a structured database. Each fault record contains at least: the time of fault occurrence, the identification of the equipment involved, the abnormal range of the main parameters (such as vibration frequency, temperature, pressure, etc.), the duration of the fault, and a confirmation label. Taking the industrial boiler system as an example, a historical record shows that the vibration sensor frequency exceeded 12Hz for 5 minutes, accompanied by a temperature increase of more than 310°C, and eventually caused the combustion system to shut down. This record will be numbered and stored in the library.

[0088] The coupling values ​​between devices are extracted from the multi-device association status data set obtained in the previous step S12 for use in ranking the importance of the devices. The ranking method is degree centrality analysis. This method extracts the coupling values ​​between the devices from the multi-device association status data set, and the coupling values ​​reflect the correlation between the operating states of the devices. For a certain target device, the system accumulates the coupling values ​​between it and all other devices to obtain the original connection strength of the device. For example, the coupling degrees of device A with devices B, C, and D are 0.75, 0.62, and 0.58, respectively, then the connection strength of device A is 0.75+0.62+0.58=1.95. This accumulation operation is performed on all devices in the system to form a set of connection strength value vectors, which represent the overall degree of association of each device with other devices in the system.

[0089] The system then performs linear normalization on this connection strength value vector to unify its numerical range for easier comparison and sorting. The normalization method is as follows: the maximum connection strength value of all devices is mapped to 1, the minimum connection strength value is mapped to 0, and the normalized values ​​of the remaining devices are proportionally distributed between 0 and 1. Still taking devices A, B, and C as an example, assuming their connection strengths are 1.95, 1.68, and 1.44, respectively, with the maximum being 1.95 and the minimum being 1.44, the normalization calculation is as follows: Device A is normalized to (1.95–1.44) / (1.95–1.44) = 1.00; Device B is normalized to (1.68–1.44) / (1.95–1.44) ≈ 0.47; and Device C is normalized to (1.44–1.44) / (1.95–1.44) = 0.00. Finally, the system generates an importance ranking table of device nodes based on the normalized connection strength values.

[0090] Based on this, the current operating status parameters of each device are collected one by one according to the device importance ranking table. Similarity calculations are performed with each record in the historical fault pattern library to identify potential fault pattern matches. This similarity calculation is performed using cosine similarity. Specifically, the current operating status parameters (e.g., vibration frequency 13 Hz, temperature 312°C, pressure 2.7 MPa) are combined with the corresponding parameters of a historical pattern (e.g., vibration frequency 12 Hz, temperature 310°C, pressure 2.5 MPa) to form two vectors, and the cosine of the angle between them is calculated. If the similarity is greater than 0.85, indicating a high degree of similarity between the current status and the historical fault, a potential fault risk is determined.

[0091] Once a match is confirmed, the device's recent parameter fluctuation data is immediately extracted from the IoT sensor system. For example, the device's vibration frequency and temperature change curves over the past 10 minutes are obtained. A sliding time window method is used to segment the continuous data. Each time segment is preset to 1 minute in length, dividing the 10-minute data into 10 time segments. The starting and ending points of any data increases or mutations are analyzed segment by segment. Specifically, the historical mean μ and standard deviation σ of the previous 5 minutes are first calculated. If the mean of three consecutive time segments in the subsequent data is greater than μ + 2σ, a significant increase or mutation is considered to have occurred. For example, if the device's vibration frequency has a mean of 11Hz and a standard deviation of 0.5Hz from the 1st to the 5th minute, if the consecutive mean values ​​from the 6th to the 8th minute are 12.3Hz, 12.8Hz, and 13.1Hz, respectively, all exceeding 11 + 1.0 = 12Hz, a mutation is considered to have occurred, and the fault propagation time window is 6 to 8 minutes.

[0092] Finally, based on the time differences in the fault propagation time windows between different devices and the coupling matrix constructed in the previous step, the path weights between the devices are calculated. The calculation of path weights takes into account two factors: the strength of the coupling between the devices and the closeness of the propagation time. Specifically, if the coupling between devices A and B is 0.80 and the propagation time difference is only 2 minutes, the path weight can be assigned a value of 0.80 × (1–2 / maximum allowable time difference). The maximum allowable time difference of 10 minutes is derived from an analysis of the distribution of propagation lag times across multiple historical operating conditions and strikes a balance between error tolerance and the accuracy of causal association identification. In this case, the path weight is 0.80 × 0.8 = 0.64.

[0093] In step S14, risk calculation and trend reasoning are performed according to the fault diffusion path scheme to obtain a real-time fault risk assessment result, including:

[0094] Mapping the fault diffusion path scheme to the corresponding device nodes and performing risk probability calculation to obtain the risk probability of each device node;

[0095] If the risk probability is greater than the preset probability threshold, the device node corresponding to the risk probability is marked as an abnormal node, and the structural position of the abnormal node in the fault diffusion path scheme is retrieved. At the same time, risk diffusion trend reasoning is performed to obtain a real-time fault risk assessment result.

[0096] It's worth noting that the path information in the fault diffusion path scheme is mapped to the corresponding set of device nodes in the form of a directed graph, forming a topological structure consisting of multiple nodes and directed edges. The direction of each edge is determined by the order in which the fault propagates. This order is determined by identifying the starting time of parameter mutations from the sensor data of each device, forming the "fluctuation start timestamp" for each device. The system then sorts the timestamps of all relevant nodes in ascending order, and determines the directionality of directed edges by assigning the device with the first fluctuation as the upstream node of the propagation path and the device with the later fluctuation as the downstream node. The edge weight is derived from the cumulative impact strength calculated along the propagation path. This topological structure serves as the spatial basis for risk calculation.

[0097] Next, for each node in the above topology, the risk probability is calculated point by point, combining the upstream node’s own failure probability and connection weight. First, the device node’s own failure probability value (hereinafter referred to as ) is derived from the statistical analysis of historical operating data. Specifically, the operation records of the equipment in the past year are divided into time segments, and the number of failures of the equipment in each time segment is counted and divided by the total number of operating cycles to obtain the probability value in the form of frequency. For example, a cooling pump has 18 abnormal start and stop events in 365 days. If the equipment runs a full cycle every day, its probability can be calculated. In order to reflect a wider range of equipment status fluctuations, it is also possible to appropriately introduce multiple types of fault indicators (such as high vibration, temperature limit, current mutation, etc.) and calculate the average value after statistics.

[0098] Next, the connection weights of the upstream nodes of the device node are calculated based on the topology established in the previous step. Each upstream node imposes a certain risk transmission strength on the device node through its connection edges. This strength is composed of two components: the upstream node's own failure probability and the connection weight. The connection weight is derived from the results of the previous propagation path analysis and is quantified as: coupling degree × fault impact factor. The fault impact factor is derived from statistical analysis of historical operating data and aims to quantify the actual severity of the impact of upstream node failures on downstream nodes. Taking the relationship between substations and distribution cabinets as an example, the system screens typical events in historical operating records and counts the number of substation failures that significantly impacted the distribution cabinet. Assuming that there are 10 substation failure events recorded in the historical data, 8 of which resulted in protection activation or shutdown of the distribution cabinet, and the average duration of these impact events exceeds 5 minutes, the impact ratio of substation failures on the distribution cabinet can be preliminarily calculated as 8 ÷ 10 = 0.8, which is then used as the value of the fault impact factor. For example, in a factory power system, the coupling degree between the substation and the distribution cabinet is 0.75. If the impact of a substation failure on the distribution cabinet is 0.8, the connection weight is 0.75×0.8=0.6.

[0099] Based on the above data, the risk probability of a node can be calculated by weighted summation. Suppose a node N has two upstream nodes and , each with its own failure probability and , the connection weights are and , then the risk probability The expression is: .

[0100] For example, assume that the cooling pump node N has a failure probability of 0.12, the boiler node is one of its upstream nodes, has a failure probability of 0.20, and a connection weight of 0.6; the main water valve node is another upstream node, has a failure probability of 0.15, and a connection weight of 0.4. The risk probability of the cooling pump node is then: .

[0101] The probability threshold is set based on a statistical analysis of historical failure data. Specifically, operational data samples from devices that experienced failures in the system within the past three years were selected and the average risk probability values ​​for these devices in the week prior to the failures were calculated. After normalization, it was found that the risk values ​​of most device nodes prior to failures were concentrated between 0.22 and 0.28. Taking into account factors such as fluctuations in the device operating environment, sensor noise, and false alarm tolerance, and to maintain the sensitivity and stability of risk screening, the risk warning threshold was ultimately set at 0.25. In other words, if a node's risk probability value is greater than 0.25, it is marked as an abnormal node and becomes a key target for subsequent analysis.

[0102] After identifying an abnormal node, further analysis is required to determine its structural position within the fault diffusion path structure, determining whether it is located at the path's starting point, an intermediate propagation link, or a terminal load. This structural position information is crucial for assessing risk diffusion trends. To infer risk diffusion trends, the directionality of the path structure and node connection weights must be combined to analyze the length and coverage of downstream paths potentially affected by the abnormal node.

[0103] The specific operation is: traverse the identified abnormal node, take it as the starting point, search all downstream nodes connected to it in turn, and calculate the cumulative impact value on each path. In order to avoid infinite loops due to the existence of loops during the path traversal process, the system introduces explicit termination conditions when performing the sequential retrieval of downstream nodes and the calculation of cumulative impact values ​​to ensure that the path propagation process is convergent and reasonable in terms of structure and data. Specifically, the system maintains a list of access records when traversing the path. If a node that has been visited appears in the current path, it is regarded as forming a loop and the expansion operation of the path is terminated; at the same time, the system also sets the maximum number of propagation levels (such as 5 levels) to limit the path depth and prevent computational redundancy and result diffusion distortion caused by too long a path; in addition, in the process of calculating the cumulative impact value by multiplying one by one, if the cumulative result after a certain one is lower than the preset minimum impact threshold (such as 0.01), it will also be determined that the propagation effect has decayed to no actual impact, and the subsequent calculation of the path will be terminated immediately.

[0104] The cumulative impact value is calculated by multiplying the connection weight of each connecting edge along the path by the risk value of the upstream node, and adding them up until the end of the path. For example, if the risk value of abnormal node A is 0.30, the connection weight with downstream node B is 0.6, and the weight of B's ​​connection to C is 0.5, then the cumulative impact value of A on C is 0.30 × 0.6 × 0.5 = 0.09. By calculating the impact value of all downstream paths through this method, a trend chart that shows the risk diffusion intensity and path length can be drawn, thereby identifying potential multi-level transmission paths or concentrated risk areas. Ultimately, the system combines these trend inference results with the initial risk calculation results to output a structured real-time fault risk assessment result.

[0105] In step S15, based on the real-time fault risk assessment result, the optimal parameter combination is searched in the preset device parameter library, and an instruction set is generated for the optimal parameter combination, thereby obtaining a parameter instruction set for high-risk equipment, including:

[0106] Based on the real-time fault risk assessment result, a parameter combination is selected from the preset equipment parameter library, an operation state prediction is performed, a corresponding state prediction result is obtained, and an optimal parameter combination is determined according to the state prediction result;

[0107] Based on the optimal parameter combination, instructions are generated and step size is adjusted to obtain a parameter instruction set for high-risk equipment.

[0108] It's worth noting that based on the real-time fault risk assessment results calculated in the previous step S14, the system extracts device nodes marked as abnormal. For these nodes, the system first searches the preset device parameter library for the corresponding parameter ranges, which serve as the basis for subsequent control instruction generation. This preset device parameter library is established by engineering and technical personnel during the initial equipment deployment phase. It contains the equipment's factory-rated operating parameters, configuration data from installation and commissioning, operating range statistics collected during historical maintenance, and the manufacturer's allowable adjustment ranges. These parameters are collected by various monitoring sensors and stored in a database, where they are regularly manually verified and updated to ensure data authenticity and coverage. For example, if a boiler system node is marked as abnormal, the key adjustable parameter ranges are extracted from the device parameter library. For example, the boiler temperature range is 280°C to 310°C, the pressure range is 3.2MPa to 4.5MPa, and the voltage range is 220V to 240V. This range constitutes a multidimensional adjustment space, providing constraints for the subsequent search for the optimal parameter combination.

[0109] After determining the boundaries of the multidimensional parameter space, the system first performs a "distance-weighted" calculation based on multiple measurable risk factors to determine the optimal parameter combination. This calculation uses the deviation between the candidate combination's parameter values ​​and the target ideal values ​​as the basis for risk contribution, with larger deviations indicating higher risk. The system selects key factors such as temperature, pressure, voltage, and historical failure rates and sets their ideal target values. For example, the target temperature for a boiler system is 290°C, pressure 3.5 MPa, and voltage 230 V. For candidate combination A (temperature 295°C, pressure 3.8 MPa, and voltage 234 V), the normalized distances are calculated as: (295–290) / 30 = 0.167, (3.8–3.5) / 1.5 = 0.2, and (234–230) / 20 = 0.2. Assuming the risk weights are 0.4 for temperature, 0.35 for pressure, and 0.25 for voltage, the risk assessment score is 0.167×0.4+0.2×0.35+0.2×0.25=0.1868. The system performs this calculation process on each candidate combination and selects the one with the lowest score as the optimal parameter combination. To avoid infinite iterations in the search process, the system sets three types of termination conditions: (1) if the score improvement is less than 1% in 10 consecutive iterations, it is considered converged; (2) the search rounds reach the upper limit of 200 rounds to prevent resource waste; (3) if the score of a group is already less than 0.05, it means that the risk is extremely low and there is no need to continue searching.

[0110] After determining the optimal parameter combination, the system further calculates the adjustment step size for each parameter. This step size controls the rate of change as the parameter transitions from its current state to its target value, preventing rapid changes that could cause system disturbances or unstable operation. To ensure a smooth adjustment process, the system employs an adaptive control strategy during this phase, analyzing and adjusting each parameter individually.

[0111] First, the system evaluates the difference between the equipment's current real-time operating value and the optimal target value determined in the previous phase. For example, consider a boiler temperature of 300°C and a target temperature of 290°C, with a 10°C difference. The system uses this difference as the basis for an initial adjustment range and, based on the equipment's temperature control capabilities, determines a reasonable temperature change rate.

[0112] During the calculation, the system presets an upper limit for the temperature adjustment rate, for example, a maximum temperature reduction of 1°C per minute. Then, based on the aforementioned 10°C deviation, the system initially sets the adjustment step size to a temperature reduction of 1°C per minute, estimating that the adjustment will take 10 minutes. If the evaluation reveals significant temperature fluctuations, such as a temperature change exceeding ±2°C over the past five minutes, the system will reduce the step size to 0.5°C per minute to slow the adjustment rate and ensure a smooth system response to avoid further disturbances.

[0113] Similar processing methods are also adopted for parameters such as pressure and voltage. For example, if the current pressure is 4.5 MPa and the target pressure is 3.8 MPa, the difference is 0.7 MPa. The system defaults the pressure adjustment rate to 0.1 MPa per minute, and the system initially estimates the adjustment time to be 7 minutes at this time; if the device is in a critical load state, the system will dynamically adjust the pressure step according to the load level, for example, reduce the adjustment rate to 0.05 MPa per minute to slow down the change rate.

[0114] In actual execution, the system will monitor the intermediate results of parameter adjustment in real time. If the system running state tends to be stable within a certain period of time, the system can gradually restore the initial larger step to speed up the overall adjustment progress; on the contrary, if the system state continues to fluctuate, the smaller step is maintained until the parameter is stable near the target value.

[0115] Finally, the adjustment steps of all parameters are integrated into the corresponding instruction templates to form a complete parameter instruction set. Each instruction explicitly includes parameter name, target value, initial step, and dynamic adjustment strategy.

[0116] In step S16, the parameter instruction set is executed, initial running state data is collected, and delay evaluation and parameter correction are performed on the initial running state data to obtain the final running state data of the adjusted device, including:

[0117] The parameter instruction set is executed, and initial running state data is collected;

[0118] Delay calculation is performed on the initial running state data, and if the delay is greater than the preset delay threshold, parameter correction and adjustment are performed on the initial running state data to obtain the final running state data of the adjusted device.

[0119] It is worth noting that after the parameter instruction set is completed through the device control interface, the system immediately executes the data collection unit to perform data grabbing operation. The collected initial running state data includes the voltage, current, temperature or other monitoring values of the current adjustment target device that have been set as control parameters. To ensure the real-time and accuracy of the collected data, the system sets the sampling frequency to one per second, and uses time stamp to ensure that the collected data and the instruction issuing behavior are accurately aligned. The starting point of the device response time is the system clock time when the parameter instruction is confirmed to be written in the device control interface, and the ending point is the time when the following conditions are first met: the fluctuation range of the continuous three collected results is not more than ±1% of the target parameter. The time difference between the two is the device response delay.

[0120] The delay time is calculated as follows: the instruction issuing time recorded by the control system and the first stable return time of the device data The delay time is calculated as follows: the instruction issuing time recorded by the control system and the first stable return time of the device data The delay time is calculated as follows: the instruction issuing time recorded by the control system The time difference is calculated. For example, if the instruction issuance time recorded by the device control system is 10:00:00, and the time when the stable feedback data is first received is 10:00:00.45, the response delay is calculated to be 0.45 seconds. In order to distinguish normal fluctuations from delay anomalies, the delay evaluation threshold is set to 0.3 seconds in this embodiment. This threshold is determined based on the statistical results of the average response time of historical equipment under the same working conditions, and after adding the minimum redundant delay value of the system communication link. Actual statistical data show that most devices complete feedback within 0.2 seconds. Therefore, setting 0.3 seconds as the upper limit of judgment can effectively identify performance anomalies caused by control errors or execution blocks.

[0121] If the calculated device response delay is less than or equal to 0.3 seconds, the system records and archives the collected initial operating status data as the device's final status data. If the delay exceeds this threshold, it indicates that the device has experienced timing deviation or control sluggishness during the adjustment response process. At this time, the system will invoke the parameter correction algorithm to adjust the current parameter combination.

[0122] During parameter correction, the system first analyzes the discrepancies between the actual and target values ​​for each parameter in the initial operating status data. For example, if the target temperature is 290°C and the actual recorded temperature is 294°C, this indicates insufficient adjustment. The system then calculates parameter corrections based on this deviation and, based on the current adjustment step size, makes appropriate fine-tuning adjustments to the voltage, current, and temperature. If the original step size is 1°C per minute, the system will increase the adjustment frequency to 1.5°C per minute and re-distribute the corrected parameters to the device control interface.

[0123] After parameter correction is complete, the device's current operating status data is collected again and the response time is recalculated. If the new data meets the latency requirements, the system confirms the adjusted operating status data as the final operating status data and uses it as a reference for the next round of fault warning analysis or control judgment. This delay assessment and correction process can effectively improve the actual execution effect of parameter adjustment commands on the device side, avoiding system operation deviations caused by execution delays or substandard parameters.

[0124] For example, in a boiler system regulation scenario, if the target temperature is 290°C, the system records the initial time as 12:00:00 after issuing the command, and the collected feedback data is at 12:00:00.6. The system calculates a response delay of 0.6 seconds, exceeding the threshold. Subsequent system analysis reveals that the actual temperature has only dropped by 1°C, so the temperature adjustment frequency is increased and the control command is reset. The final collected data shows that the temperature has dropped to the target value at 12:00:01, and the response delay has been shortened to 0.25 seconds, indicating that the correction is effective and the status is stable.

[0125] In step S17, a state difference analysis and risk score calculation are performed based on the final operating state data, and the path weights in the fault diffusion path solution are numerically updated based on the risk score, thereby obtaining a final fault diffusion path solution, including:

[0126] Obtain historical benchmark datasets;

[0127] Based on the historical benchmark data set, calculating the state difference value of the final operating state data to obtain a state difference value;

[0128] Calculating a risk score based on the state difference value to obtain a risk score;

[0129] If the risk score is greater than a preset risk threshold, the path weights in the fault diffusion path solution are adjusted to obtain a final fault diffusion path solution.

[0130] It's worth noting that the system extracts historical baseline data sets from a database of historical equipment initial operating states. This database is derived from multiple rounds of data collected during routine inspections and mid- to long-term scheduled inspections, and undergoes a unified data screening and archiving process. For example, for transmission equipment at a power plant, the equipment management system has collected operating parameters such as temperature, pressure, and vibration frequency every five minutes for the past two years. The system then selects data segments showing stable initial operating states under non-fault and non-load shock conditions to serve as a reference for constructing the historical baseline. To define a non-fault and non-load shock state, the following two conditions must be met simultaneously: First, the instantaneous fluctuation of key monitoring parameters must be within an acceptable range, such as a vibration frequency fluctuation of less than ±5% for at least five minutes; second, the equipment's load state must be relatively stable, such as a temperature change rate of less than 1°C / min, a current fluctuation of less than 5%, or a load factor fluctuation within ±10%, with no alarms or manual intervention during this period. Only data segments that meet both of these conditions are considered by the system to be in a stable operating state (non-fault, non-load shock) and serve as candidate segments for historical benchmark data, further used to construct a reference range for equipment condition assessment. After applying time-weighted averaging, sliding window smoothing, and outlier removal to this data, the reference operating range for the equipment in a stable state is obtained. For example, a stable vibration frequency of 5.2 Hz, a temperature gradient of 3.5 degrees Celsius, and a pressure fluctuation of 0.20 MPa are obtained.

[0131] Next, the final operating state data is collected. This refers to the measured initial operating state data after the equipment has executed the adjustment instructions, such as a vibration frequency of 5.7 Hz, a temperature gradient of 4.0 degrees Celsius, and a pressure fluctuation of 0.24 MPa. The current data is then compared with the historical baseline data, and the state difference values ​​of each indicator are calculated. The difference is calculated using the absolute value difference method, indicating the degree to which the current parameter deviates from the stable range. For example, the vibration frequency difference is 0.5 Hz, the temperature gradient difference is 0.5 degrees Celsius, and the pressure fluctuation difference is 0.04 MPa.

[0132] The system assigns weights based on the impact of different parameters on the fault. For example, vibration frequency is weighted at 0.5, temperature gradient at 0.3, and pressure fluctuation at 0.2. Each difference value is multiplied by its corresponding weight and the sum is calculated to obtain a risk score. In this example, the risk score is: 0.5 times 0.5 equals 0.25, 0.5 times 0.3 equals 0.15, and 0.04 times 0.2 equals 0.008. The final score is 0.25 plus 0.15 plus 0.008, for a total of 0.408.

[0133] To ensure that the scoring results are sufficient for engineering judgment, the system sets a scoring threshold of 0.35. Over 80% of devices with a threshold score above 0.35 experienced functional abnormalities or conductive risk events within 30 minutes, making this value a reasonable threshold for practical applications.

[0134] If the current score exceeds 0.35, the system triggers a path weight update. This update uses a linear adjustment method, multiplying the original weights of the relevant device nodes in the path by the excess score and then performing a numerical correction. For example, if the original path weight from device A to device B in the current propagation path is 0.50, and the score exceeds the threshold by 0.058, and the adjustment coefficient is 1.2, the final weight is 0.50 plus 0.058 multiplied by 1.2, resulting in a total of 0.57. This ultimately generates an updated fault diffusion path plan, which serves as the final fault diffusion path plan for subsequent device linkage control and risk prediction.

[0135] Reference Figure 2 The second embodiment of the present invention provides a power plant equipment fault diagnosis system based on the Internet of Things, including:

[0136] Data acquisition module, used to obtain data sets for monitoring the working conditions of each device;

[0137] A multi-device module is used to perform coupling analysis and abnormal triggering collection based on the working condition monitoring data sets of each device to obtain a multi-device associated status data set;

[0138] A path plan module, configured to perform node analysis and path allocation based on the multi-device association status data set to obtain a fault diffusion path plan;

[0139] A risk assessment module is used to perform risk calculation and trend reasoning based on the fault diffusion path scheme to obtain real-time fault risk assessment results;

[0140] An instruction parameter set module is used to search for an optimal parameter combination in a preset device parameter library based on the real-time fault risk assessment result, and generate an instruction set for the optimal parameter combination, thereby obtaining a parameter instruction set for high-risk equipment;

[0141] An operating status module is used to execute the parameter instruction set, collect initial operating status data, and perform delay evaluation and parameter correction on the initial operating status data to obtain final operating status data after equipment adjustment;

[0142] The final path solution module is used to perform state difference analysis and risk score calculation according to the final operating state data, and based on the risk score, numerically update the path weights in the fault diffusion path solution to obtain a final fault diffusion path solution.

[0143] It should be noted that the power plant equipment fault diagnosis system based on the Internet of Things provided in an embodiment of the present invention is used to execute all the process steps of the power plant equipment fault diagnosis method based on the Internet of Things in the above embodiment. The working principles and beneficial effects of the two correspond one to one, so they will not be repeated here.

[0144] It should be noted that the device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed across multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the present embodiment. In addition, in the drawings of the device embodiments provided by the present invention, the connection relationship between the modules indicates that there is a communication connection between them, which may be specifically implemented as one or more communication buses or signal lines. A person of ordinary skill in the art can understand and implement the present invention without inventive effort.

[0145] The specific embodiments described above further illustrate the objectives, technical solutions, and beneficial effects of the present invention. It should be understood that the above descriptions are merely specific embodiments of the present invention and are not intended to limit the scope of protection of the present invention. In particular, it should be noted that any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included within the scope of protection of the present invention for those skilled in the art.

Claims

1. A method for diagnosing power plant equipment faults based on the Internet of Things, characterized in that: include: Obtain the data set for monitoring the working condition of each device; Based on the data sets of the equipment working condition monitoring, coupling analysis and abnormal triggering collection are performed to obtain a multi-equipment associated state data set; Perform node analysis and path allocation based on the multi-device association status data set to obtain a fault diffusion path solution; According to the fault diffusion path scheme, risk calculation and trend reasoning are performed to obtain real-time fault risk assessment results; Based on the real-time fault risk assessment results, searching for an optimal parameter combination in a preset device parameter library, and generating an instruction set for the optimal parameter combination, thereby obtaining a parameter instruction set for high-risk equipment; Executing the parameter instruction set, collecting initial operating status data, and performing delay evaluation and parameter correction on the initial operating status data to obtain final operating status data after equipment adjustment; According to the final operating state data, a state difference analysis and a risk score calculation are performed, and based on the risk score, the path weights in the fault diffusion path scheme are numerically updated to obtain a final fault diffusion path scheme.

2. The method for diagnosing power plant equipment faults based on the Internet of Things according to claim 1, characterized in that: The coupling analysis and abnormal triggering collection are performed based on the working condition monitoring data sets of each device to obtain a multi-device associated state data set, including: The equipment operating condition monitoring data sets include: generator vibration frequency, boiler temperature gradient change and cooling system pressure fluctuation; Performing time synchronization processing on the equipment operating condition monitoring data set to obtain a time-synchronized operating condition data set; Performing coupling degree calculation and matrix construction on the time-synchronized working condition data set to obtain an inter-device coupling degree matrix; If the generator vibration frequency of the time-synchronized operating condition dataset exceeds a preset frequency threshold, the device corresponding to the generator vibration frequency is marked as an abnormal device, and the boiler temperature gradient change and cooling system pressure fluctuation in the current time window of the abnormal device are additionally collected as extended data; The extended data and the inter-device coupling matrix are subjected to dimensionality reduction processing to extract the main correlation features between multiple devices and obtain a multi-device correlation status data set.

3. The method for diagnosing power plant equipment faults based on the Internet of Things according to claim 2, characterized in that: The performing node analysis and path allocation based on the multi-device association status data set to obtain a fault diffusion path solution includes: Obtain historical failure mode library; Sorting the importance of the multi-device association status data set to obtain an importance ranking table of the device nodes; Based on the importance ranking table and in combination with the historical fault mode library, similarity calculation is performed one by one. If the similarity exceeds a preset similarity threshold, parameter fluctuation data of the device is collected; A time series analysis is performed on the parameter fluctuation data to determine the fault propagation time window, and a path structure is constructed and path weights are calculated in combination with the inter-device coupling matrix to obtain a fault diffusion path solution.

4. The method for diagnosing power plant equipment faults based on the Internet of Things according to claim 1, characterized in that: The risk calculation and trend reasoning are performed according to the fault diffusion path scheme to obtain a real-time fault risk assessment result, including: Mapping the fault diffusion path scheme to the corresponding device nodes and performing risk probability calculation to obtain the risk probability of each device node; If the risk probability is greater than a preset probability threshold, the corresponding device node is marked as an abnormal node, and the structural position of the abnormal node in the fault diffusion path scheme is retrieved, and risk diffusion trend reasoning is performed at the same time to obtain a real-time fault risk assessment result.

5. The method for diagnosing power plant equipment faults based on the Internet of Things according to claim 1, characterized in that: Based on the real-time fault risk assessment result, searching for an optimal parameter combination in a preset device parameter library, and generating an instruction set for the optimal parameter combination, thereby obtaining a parameter instruction set for high-risk equipment, including: Based on the real-time fault risk assessment result, a parameter combination is selected from the preset equipment parameter library, an operation state prediction is performed, a corresponding state prediction result is obtained, and an optimal parameter combination is determined according to the state prediction result; Based on the optimal parameter combination, instructions are generated and step size is adjusted to obtain a parameter instruction set for high-risk equipment.

6. The method for diagnosing power plant equipment faults based on the Internet of Things according to claim 1, characterized in that: The executing the parameter instruction set, collecting initial operating status data, and performing delay evaluation and parameter correction on the initial operating status data to obtain final operating status data after device adjustment includes: Executing the parameter instruction set and collecting initial operating status data; A delay calculation is performed on the initial operating status data. If the delay is greater than a preset delay threshold, parameter correction and adjustment are performed on the initial operating status data to obtain final operating status data after equipment adjustment.

7. The method for diagnosing power plant equipment faults based on the Internet of Things according to claim 1, characterized in that: The performing of state difference analysis and risk score calculation according to the final operating state data, and numerically updating the path weights in the fault diffusion path scheme based on the risk score, thereby obtaining a final fault diffusion path scheme, includes: Obtain historical benchmark datasets; Based on the historical benchmark data set, calculating the state difference value of the final operating state data to obtain a state difference value; Calculating a risk score based on the state difference value to obtain a risk score; If the risk score is greater than a preset risk threshold, the path weights in the fault diffusion path solution are adjusted to obtain a final fault diffusion path solution.

8. A power plant equipment fault diagnosis system based on the Internet of Things, characterized in that: include: Data acquisition module, used to obtain data sets for monitoring the working conditions of each device; A multi-device module is used to perform coupling analysis and abnormal triggering collection based on the working condition monitoring data sets of each device to obtain a multi-device associated status data set; A path plan module, configured to perform node analysis and path allocation based on the multi-device association status data set to obtain a fault diffusion path plan; A risk assessment module is used to perform risk calculation and trend reasoning based on the fault diffusion path scheme to obtain real-time fault risk assessment results; An instruction parameter set module is used to search for an optimal parameter combination in a preset device parameter library based on the real-time fault risk assessment result, and generate an instruction set for the optimal parameter combination, thereby obtaining a parameter instruction set for high-risk equipment; An operating status module is used to execute the parameter instruction set, collect initial operating status data, and perform delay evaluation and parameter correction on the initial operating status data to obtain final operating status data after equipment adjustment; The final path solution module is used to perform state difference analysis and risk score calculation according to the final operating state data, and based on the risk score, numerically update the path weights in the fault diffusion path solution to obtain a final fault diffusion path solution.

Images