Artificial intelligence-based power equipment analysis method and system

Abstract

The application discloses an electric power equipment analysis method and system based on artificial intelligence, and belongs to the technical field of fault analysis. The system comprises a data acquisition module, an intelligent analysis module, an equipment management module and a visualization module; the data acquisition module is used for collecting temperature data, operation parameters and historical logs of electric power equipment, and environmental parameters around the electric power equipment, and generating an electric power equipment model containing parts; the intelligent analysis module sets sampling points on the surface of the electric power equipment model, analyzes historical logs and filters data records; a relational expression is fitted for the sampling points according to the data records, so that abnormal points are set; the equipment management module identifies and associates parts in physical contact with the abnormal points in the electric power equipment model through an artificial intelligence algorithm, analyzes the shortest physical contact path between the two and calculates the abnormal index of each part, and filters abnormal parts; the visualization module arranges each abnormal part according to the abnormal index, sequentially displays the abnormal parts and warns the control center.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of fault analysis, in particular to an electric power equipment analysis method and system based on artificial intelligence. BACKGROUND

[0002] In recent years, with the rapid development of artificial intelligence technology, especially the breakthroughs of machine learning and deep learning algorithms in data analysis, pattern recognition and prediction, new ideas and methods have been provided to solve the above problems. Artificial intelligence technology can automatically learn and extract fault features from massive data, build accurate fault diagnosis models, and realize early warning, accurate positioning and intelligent diagnosis of thermal faults of electric power equipment.

[0003] Thermal fault is a common fault type of electric power equipment, characterized by abnormal temperature rise of the equipment. If not handled in time, it may cause insulation breakdown, component damage, fire, large-scale power outage and huge economic loss. At present, multi-angle thermal imaging monitoring combined with threshold setting is usually used for risk warning. This method has some problems. On the one hand, although the existing infrared monitoring system can build a three-dimensional temperature field, it cannot associate the temperature abnormal points with specific components. The internal structure of electric power equipment is complex, and thermal faults are conducted and diffused through contact paths. The traditional method can only identify surface overheating points, and cannot trace the thermal conduction link or distinguish the fault source from healthy components. For example, when a transformer has a local short circuit, heat diffusion causes the "low temperature surrounding high temperature" illusion, which hides the real fault location and increases the risk of misdisassembly and delay. On the other hand, the three-dimensional temperature field is easily disturbed by the environment and lacks dynamic compensation. Convective heat dissipation under high wind speed reduces the temperature of the overheating point below the threshold; temperature distribution is distorted in low temperature or rain, and the temperature difference signal is drowned in noise. The change of solar radiation causes the temperature difference between the sunny and shady sides, and the fixed threshold model cannot separate the interference, causing missed faults in high wind speed, low temperature and rainy weather, or frequent false alarms in alternating sunny and cloudy weather. Therefore, there is an urgent need for an analysis scheme that integrates component physical topology identification, thermal conduction path tracking and environmental disturbance compensation to solve the problems of inaccurate fault attribution and environmental misjudgment. SUMMARY

[0004] The purpose of the present application is to provide an electric power equipment analysis method and system based on artificial intelligence to solve the problems raised in the background art.

[0005] In order to solve the above technical problems, the present application provides an electric power equipment analysis method based on artificial intelligence, comprising:

[0006] S100, collecting temperature data, operating parameters and historical logs of the electric power equipment, and environmental parameters of the surrounding environment. Analyze the operating parameters and temperature data to generate an electric power equipment model containing components, and establish a virtual scene.

[0007] The temperature data refers to the thermal imaging pictures of the power equipment collected by the infrared thermal imager at different angles.

[0008] The operating parameters refer to the power of the power equipment and the digital model, which contains various component models capable of generating heat during operation.

[0009] The historical log refers to the data record of the surface temperature distribution when the power equipment operates at different powers each time. Each data record includes power, environmental parameters, and surface temperature distribution.

[0010] The environmental parameters include different environmental factors, specifically physical indicators that quantitatively describe the environmental state.

[0011] The environmental factors refer to physical indicators that can affect the surface temperature of the power equipment, typically including environmental temperature, solar radiation, relative humidity, wind speed, and wind direction.

[0012] A virtual scene is established and the digital model of the power equipment is placed in it. The orientation and angle of the power equipment at which the different thermal imaging pictures in the temperature data are located are analyzed, and the multi-angle thermal imaging pictures are processed into three-dimensional temperature field data.

[0013] The three-dimensional temperature field data is mapped to the surface of the digital model of the power equipment, generating a digital model superimposed with temperature distribution.

[0014] Through the infrared thermal imager, multi-angle thermal imaging data of the power equipment is collected, combined with operating parameters, historical logs, and environmental parameters. A virtual scene is established, and the three-dimensional temperature field data is mapped to the surface of the digital model of the power equipment, generating a 3D model superimposed with temperature distribution.

[0015] It provides the basis for multi-source data fusion, generates high-precision spatialized equipment models, solves the limitations of traditional single temperature monitoring, and provides reliable three-dimensional data support for subsequent anomaly analysis.

[0016] S200, set sampling points on the surface of the power equipment model in the virtual scene, and analyze the historical log to filter data records. According to the data records, fit the relationship for the sampling points and calculate the predicted temperature, and set the abnormal points by comparing the temperature data. Specifically including:

[0017] S201, set the sampling density , obtain the grid data of the surface of the power equipment model in the virtual scene, divide the total surface area by the sampling density to get the total number of sampling points .

[0018] S202, based on the Poisson disc distribution algorithm, uniformly iteratively generate three-dimensional sampling points on the grid surface. Establish the mapping relationship between the sampling points and the temperature field, and complete the attachment of spatial temperature data.

[0019] S203, obtaining the current power of the power equipment , all data records with the power are screened out from the historical log. According to the surface temperature distribution of each data record, the temperature of each sampling point is analyzed.

[0020] S204, analyzing the temperature change of each sampling point within the duration of the data record, intercepting each environmental factor within the time period when the temperature of each sampling point remains unchanged, and fitting the relationship between the temperature and the environmental factor for each sampling point. Specifically, it includes:

[0021] S2041, obtaining the sampling point within the duration of the data record , intercepting the time period when the temperature of the sampling point remains unchanged, and calculating the average value of each environmental factor within each time period.

[0022] S2042, intercepting the time period when the temperature of the sampling point remains unchanged in each data record in turn, and calculating the average value of each environmental factor, and counting the number of time periods e of the sampling point in all data records.

[0023] S2043, counting the number of environmental factors , taking the temperature in each time period as the dependent variable, and the environmental factors as the independent variables, and packing them as samples.

[0024] S2044, setting the intercept and the regression coefficient , and establishing a linear regression model. Taking the independent variables in the e samples as input values , and the difference between the output value of each sample and the dependent variable as the gap coefficient. The expression is as follows:

[0025] ;

[0026] S2045, adjusting the intercept and the regression coefficient until the sum of the gap coefficients of all samples is minimized, and obtaining the fitted relationship. In this way, the relationship between the temperature and the environmental factor is fitted for each sampling point.

[0027] S205, obtaining the environmental factors in the current environmental parameters, substituting them into the relationship of each sampling point to obtain the predicted temperature . Obtaining the current temperature of each sampling point, and taking the sampling point with a temperature greater than the threshold value as an abnormal point.

[0028] Based on the model of the virtual scene, sample points are uniformly generated using the Poisson disc distribution algorithm. Analyze historical logs to filter data records at the same power. Fit a linear regression relationship between temperature and environmental factors for each sample point. Calculate the predicted temperature and compare it with the actual temperature to set the abnormal point.

[0029] The quantitative analysis of temperature and environment is realized, which significantly reduces the false positive rate caused by environmental factors. By comparing the predicted temperature, local temperature anomaly points can be dynamically identified, improving the accuracy of early fault warning and reducing noise interference.

[0030] S300, through the artificial intelligence algorithm, the abnormal parts and the abnormal points in the power equipment model are identified and associated, the shortest physical contact path between them is analyzed, and the abnormal index of each part is calculated, so as to screen out abnormal parts. Specifically, it includes:

[0031] S301, analyze the power equipment model in the virtual scene, identify the physical contact between the parts and the location of the abnormal points through the artificial intelligence algorithm, and associate the parts and the abnormal points with direct or indirect physical contact relationship.

[0032] Physical contact judgment includes data input, direct contact judgment and indirect contact tracing.

[0033] Data input: import the grid data of the digital model of the power equipment, including the geometric structure of the parts and the assembly topological relationship.

[0034] Direct contact judgment: perform hierarchical bounding box collision detection algorithm on the location of the abnormal point, identify the parts that intersect with its grid, and define as direct physical contact relationship.

[0035] Indirect contact tracing: based on the assembly topological relationship, recursively search for adjacent parts connected by direct physical contact parts, and construct physical conduction link:

[0036] If the parts are connected by bolts / welds, or there is a conduction medium with thermal conductivity > 5 W / (m·K), it is defined as indirect physical contact relationship.

[0037] Thermal path map construction: store all direct and indirect contact parts in the relational graph database according to the order of heat conduction path, and label the effective contact area and intermediate material thermal conductivity between nodes.

[0038] Dynamic update mechanism: when the pose of the part changes in the virtual scene, recalculate the contact detection and thermal link in real time to ensure that the contact relationship is consistent with the physical working condition.

[0039] Through three-order analysis of collision detection + topological recursion + link modeling, it fully covers the industrial scene requirements from direct contact to indirect contact.

[0040] S302、Calculate the distance of the shortest physical contact path between the interrelated parts and the abnormal point , and the volume of each part , so as to calculate the weight index between the parts and the abnormal point. Specifically, it includes:

[0041] S3021, analyze the thermal distribution of each part related to the abnormal point when working normally, calculate the thermal distribution gradient norm , substitute the formula to calculate the heat contribution efficiency index of the part :

[0042] ;

[0043] In the formula, , , , , are adaptive weight coefficients, is the characteristic thermal diffusion length of the system.

[0044] is the path distance attenuation, which is used to quantify the thermal conduction path resistance, avoid zero distance singular points.

[0045] is the thermal distribution enhancement factor, which is used to compress the volume influence to avoid large volume monopoly, and multiply the temperature gradient to highlight the contribution of local high temperature area.

[0046] Introduce the thermal conduction attenuation model, used to associate material properties.

[0047] The heat contribution efficiency index fuses the path physical conduction characteristics + thermal distribution gradient + volume compression effect, providing a quantitative basis for the attribution of multi-heat source coupled temperature change.

[0048] The thermal distribution gradient norm calculation includes:

[0049] First, collect multi-angle thermal imaging of power equipment under rated working condition by infrared thermal imager, or call the steady-state temperature field distribution data exported by CFD simulation software.

[0050] Second, perform trilinear interpolation on the thermal imaging data to generate a three-dimensional temperature matrix matched with the digital model grid.

[0051] Then, for each grid node, use the central difference method to calculate its thermal gradient, and take the Euclidean norm of the gradient vector.

[0052] Finally, take the weighted average of the gradient norms of all mesh nodes belonging to the same component, and output the gradient norm of the thermal distribution of the component as a whole.

[0053] S3022, respectively calculate the weight index between each abnormal point and the associated components of each abnormal point :

[0054] ;

[0055] In the formula, is the weight index between the i-th component and the abnormal point , is the number of components associated with the abnormal point , is the gain factor. The gain factor is used to amplify the impact of key components: when

[0056] , the weight of high value is strengthened. When , the distribution is smoothed. The gain factor supports dynamic adjustment: when the load of the power equipment fluctuates, by adjusting to match the measured temperature rise distribution, the sensitivity of abnormal attribution is improved.

[0057] is the heat contribution efficiency index between the i-th component and the abnormal point . is the heat contribution efficiency index between the i-th component and the abnormal point . is the heat contribution efficiency index between the i-th component and the abnormal point .

[0058] By introducing the exponential function to simulate the nonlinear cumulative effect of heat conduction, the weight of high value components is more significant.

[0059] S3023, in the same way, calculate the weight index for all associated components for each abnormal point.

[0060] S303, according to the difference between the current temperature and the predicted temperature of each abnormal point, combined with the weight index, calculate the abnormal index of each component, and select the components with abnormal index greater than the threshold value as abnormal components. The abnormal index The calculation formula is:

[0061] ;

[0062] wherein, is the number of abnormal points associated with the component, is the weight index between the component and the th abnormal point, and are the current temperature and the predicted temperature of the th abnormal point, respectively.

[0063] The physical contact relationship is identified by an artificial intelligence algorithm, a thermal path atlas is constructed, the shortest physical path distance and the heat contribution efficiency index of the component are calculated, the abnormal index of each component is calculated based on the weight index, and the abnormal components are screened.

[0064] Fault tracing under multi-physical field coupling is realized, and the accuracy of fault positioning is improved. Through thermal path analysis, adaptability under high load working conditions is ensured, and false attribution caused by heat conduction is avoided.

[0065] S400, arrange each abnormal component according to the abnormal index, and display the abnormal component in order and prewarn to the control center.

[0066] The abnormal components are sorted from high to low according to the abnormal index. In the virtual scene power equipment model, each abnormal component is dynamically layered and highlighted:

[0067] The abnormal component with the highest abnormal index is warned with pulsed red light, the abnormal component with the second highest abnormal index is marked with an orange outline frame, and the remaining abnormal components are highlighted. At the same time, the thermal distribution cloud map is rendered by fusing the real-time three-dimensional temperature field.

[0068] When the abnormal index of the highest abnormal component exceeds the dynamic threshold, a first-level fault warning is automatically pushed to the control center, and when the abnormal index of the remaining abnormal components continuously rises for two consecutive periods, a second-level maintenance alarm is triggered.

[0069] Finally, a diagnostic report containing spatial coordinates, abnormal index and historical data is generated, and the sound and light alarm system of the control center is activated synchronously.

[0070] The abnormal components are sorted according to the abnormal index. In the virtual model, they are layered and highlighted. A diagnostic report is generated. When the index exceeds the threshold, a first-level warning or a second-level alarm is pushed. The sound and light alarm system of the control center is activated.

[0071] A multi-dimensional visualization interface is provided, which optimizes the operation and maintenance response strategy and shortens the fault diagnosis time. The hierarchical warning mechanism reduces unnecessary downtime losses, and through the sound and light alarm, immediate intervention is realized, improving the overall safety of the system.

[0072] The application also provides an artificial intelligence-based power equipment analysis system, comprising a data acquisition module, an intelligent analysis module, an equipment management module, and a visualization module.

[0073] The data acquisition module is used to acquire temperature data, operating parameters, and historical logs of the power equipment, as well as environmental parameters of the surroundings, and generate a power equipment model containing components.

[0074] The infrared thermal imager is used to acquire thermal imaging data of the power equipment from multiple angles, in combination with operating parameters, historical logs, and environmental parameters. After the digital model is established, the three-dimensional temperature field is mapped to the model surface, and temperature distribution visualization is realized.

[0075] The multi-source heterogeneous data is fused to generate a 3D equipment model with temperature distribution, solving the limitations of traditional single temperature monitoring and providing a spatialized data basis for anomaly analysis.

[0076] The intelligent analysis module sets sampling points on the surface of the power equipment model and analyzes historical log data records. According to the data records, a relationship is fitted for the sampling points, and thus abnormal points are set.

[0077] The Poisson disc algorithm is used to uniformly distribute sampling points on the model surface. Based on historical log data records, the same power data records are screened, and the mean values of environmental factors during the stable temperature period are intercepted. The relationship between the temperature of each sampling point and the environmental factor is fitted by linear regression. By comparing the predicted temperature with the actual temperature, abnormal points with a temperature difference exceeding the threshold are calibrated.

[0078] The influence of environmental factors on temperature is quantified to reduce the false alarm rate. Local temperature rise anomalies are dynamically identified to improve fault positioning accuracy.

[0079] The equipment management module identifies and associates components in the power equipment model that have physical contact with abnormal points through artificial intelligence algorithms, analyzes the shortest physical contact path between the two, calculates the abnormal index of each component, and screens abnormal components.

[0080] Bounding box collision detection and assembly topology recursion are combined to identify directly / indirectly contacted components.

[0081] A thermal path atlas is constructed to calculate the shortest path distance and heat contribution efficiency index. Based on the weight index, the temperature rise deviation of multiple abnormal points is weighted to screen components with AN exceeding the threshold.

[0082] Fault attribution under multi-physical field coupling is realized to analyze the shortest physical contact path between the two. The thermal conduction link is dynamically updated to ensure the adaptability of working conditions.

[0083] The visualization module arranges abnormal components according to the abnormal index, sequentially displays abnormal components, and warns the control center.

[0084] Parts are ranked by anomaly index, and a hierarchical warning strategy is adopted. A real-time temperature field thermal map is superimposed. First-level warning is pushed to the control center, and second-level alarm triggers maintenance reminders. A diagnostic report containing spatial coordinates is generated and an audible and visual alarm is activated.

[0085] Multi-dimensional visualization improves fault identification efficiency. The hierarchical warning mechanism optimizes the operation and maintenance response strategy, reducing downtime losses.

[0086] The system forms a diagnostic closed loop from acquisition to analysis to management to visualization. Through a group of artificial intelligence algorithms such as Poisson sampling, bounding box detection, regression model, and heat conduction formula, intelligent diagnosis of industrial scenes is achieved.

[0087] Compared with the prior art, the beneficial effects achieved by the present application are:

[0088] Multi-source data fusion modeling: Break through the limitations of traditional single temperature monitoring, integrate infrared thermal imaging, operating power, historical operating conditions, and multi-dimensional environmental parameters (temperature / wind speed / radiation, etc.), and construct a three-dimensional digital device model with temperature distribution. Through spatial mapping, the problem of disconnection between surface hot spots and component entities is solved, providing accurate spatial reference for fault tracing.

[0089] Dynamic environmental interference stripping: Based on the environmental factor regression analysis of historical logs, a temperature-environment dynamic relationship model is established for each surface sampling point. By comparing the predicted temperature with the actual temperature, environmental disturbances such as sudden wind speed changes and solar radiation are effectively stripped, avoiding fault omission and false alarms in high wind speed / low temperature scenarios.

[0090] Physical conduction path quantification: Collision detection (OBBTree) and assembly topology recursive algorithm are introduced to identify direct / indirect heat conduction paths composed of bolt connections and high thermal conductivity media (> 5W / (m·K)). Combined with the heat gradient attenuation model (HCEI formula), the volume effect and path distance influence of the component are quantified to accurately distinguish the fault source component and the conductive temperature rise.

[0091] Hierarchical warning and three-dimensional visualization: Based on the dynamic ranking of components by anomaly index, a hierarchical warning strategy using pulsed red light and outline boxes is adopted, and a real-time temperature field thermal map is superimposed. Through the hierarchical mechanism of first-level fault warning (dynamic threshold triggering) and second-level maintenance alarm (continuous deterioration judgment), the operation and maintenance response priority is optimized, and the diagnostic decision chain is shortened.

[0092] Industrial scene adaptive capability: Full-process support for dynamic adjustment, real-time update of thermal path atlas to ensure the accuracy of conduction links in changing operating conditions. The environmental factor regression model is continuously iterated to ensure the analysis robustness under different climate conditions. BRIEF DESCRIPTION OF DRAWINGS

[0093] The accompanying drawings are included to provide a further understanding of the application, and are incorporated in and constitute a part of the specification, illustrate embodiments of the application, and are used to explain the present application, but are not intended to limit the present application. In the drawings:

[0094] Figure 1 is a flowchart of the power equipment analysis method based on artificial intelligence of the present application;

[0095] Figure 2 is a structural schematic diagram of the power equipment analysis system based on artificial intelligence of the present application. DETAILED DESCRIPTION

[0096] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application. 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.

[0097] Please refer to Figure 1 The present application provides a power equipment analysis method based on artificial intelligence, comprising:

[0098] S100, collecting temperature data, operating parameters and historical logs of the power equipment, and surrounding environmental parameters. Analyzing the operating parameters and the temperature data, generating a power equipment model containing components, and establishing a virtual scene.

[0099] The temperature data refers to thermal imaging pictures of the power equipment at different angles collected by an infrared thermal imager.

[0100] The operating parameters refer to the power of the power equipment and a digital model, and the digital model contains component models capable of generating heat during operation.

[0101] The historical logs refer to data records of surface temperature distribution when the power equipment is operated at different powers each time, and each data record includes power, environmental parameters and surface temperature distribution.

[0102] The environmental parameters include different environmental factors, specifically physical indicators for quantitatively describing the environmental state.

[0103] The environmental factors refer to physical indicators that can affect the surface temperature of the power equipment, usually including environmental temperature, solar radiation, relative humidity, wind speed and wind direction (the angle between the surface of the power equipment and the wind vector).

[0104] The virtual scene is established and the digital model of the power equipment is put in. The positions and angles of the power equipment in different thermal imaging pictures in the temperature data are analyzed, and the multi-angle thermal imaging pictures are processed into three-dimensional temperature field data.

[0105] The three-dimensional temperature field data is mapped to the surface of the digital model of the power equipment, to generate a digital model superimposed with temperature distribution.

[0106] Through the infrared thermal imager, multi-angle thermal imaging data (temperature data) of the power equipment is collected, combined with operating parameters (such as power and digital model), historical logs (power-temperature distribution records), and environmental parameters (environmental factors including temperature, wind speed, etc.). A virtual scene is established, the three-dimensional temperature field data is mapped to the surface of the digital model of the power equipment, to generate a 3D model superimposed with temperature distribution.

[0107] A basis for multi-source data fusion is provided, a high-precision spatialized equipment model is generated, the limitations of traditional single temperature monitoring are solved, and reliable three-dimensional data support is provided for subsequent anomaly analysis.

[0108] S200, sample points are set on the surface of the power equipment model in the virtual scene, and data records are screened by analyzing historical logs. A relationship is fitted for the sample points according to the data records and a predicted temperature is calculated, and an abnormal point is set by comparing the temperature data. Specifically, it includes:

[0109] S201, set the sampling density , obtain the grid data of the surface of the power equipment model in the virtual scene, divide the total surface area by the sampling density to obtain the total number of sample points .

[0110] S202, based on the Poisson disc distribution algorithm, generate three-dimensional sample points on the grid surface uniformly and iteratively. A mapping relationship between the sample points and the temperature field is established, and the spatial temperature data is attached.

[0111] S203, obtain the current power of the power equipment , and screen all data records with the power of from the historical logs. According to the surface temperature distribution of each data record, the temperature of each sample point is analyzed.

[0112] S204, analyze the temperature changes of each sample point within the duration of the data record, and intercept each item of environmental factors within the time period when the temperature of each sample point remains unchanged, so as to fit the relationship between the temperature and the environmental factors for each sample point. Specifically, it includes:

[0113] S2041, obtain the temperature changes of the sample point within the duration of the data record , intercept the time period when the temperature of the sample point remains unchanged, and calculate the average value of each item of environmental factors within each time period.

[0114] S2042, intercept sample points in each data record in turn The time period when the temperature remains unchanged is calculated, and the average value of each environmental factor is calculated The number of time periods e in all data records

[0115] S2043, count the number of environmental factors Take the temperature in each time period as the dependent variable, Take each environmental factor as the independent variable, and pack it as a sample.

[0116] S2044, set the intercept And the regression coefficient And establish a linear regression model. Take the independent variable in each of the e samples as the input value The difference between the output value of each sample And the dependent variable as the gap coefficient. The expression is as follows:

[0117] ;

[0118] S2045, adjust the intercept and regression coefficient until the sum of the gap coefficients of all samples is minimized, and the fitted relationship is obtained. In this way, the relationship between temperature and environmental factors is fitted for each sample point.

[0119] S205, get the environmental factors in the current environmental parameters, and substitute them into the relationship of each sample point to get the predicted temperature Get the current temperature of each sample point The temperature Greater than the threshold value is taken as an abnormal point.

[0120] Based on the model of the virtual scene, the sample points are uniformly generated using the Poisson disc distribution algorithm. Analyze the historical log to filter data records under the same power. Fit the linear regression relationship between temperature and environmental factors for each sample point. Calculate the predicted temperature, compare the actual temperature, and set the abnormal point (temperature difference exceeds the threshold).

[0121] The quantitative analysis of temperature and environment is realized, which significantly reduces the false positive rate caused by environmental factors (such as wind speed, solar radiation). Through the comparison of the predicted temperature, local temperature anomaly points can be dynamically identified, the accuracy of early fault warning is improved, and noise interference is reduced (such as avoiding false positive alarms caused by temperature fluctuations).

[0122] S300, identify and associate the components that exist in physical contact with the abnormal points in the power equipment model through artificial intelligence algorithms, analyze the shortest physical contact path between the two, and calculate the abnormal index of each component, thereby screening out abnormal components. Specifically includes:

[0123] S301, analyze the power equipment model in the virtual scene, identify the physical contact between the parts and the position of the abnormal point by artificial intelligence algorithm, associate the parts and the abnormal point with direct or indirect physical contact relationship.

[0124] Physical contact judgment includes data input, direct contact judgment and indirect contact tracing.

[0125] Data input: import the grid data of the digital model of the power equipment, including the geometric structure of the parts and the assembly topological relationship.

[0126] Direct contact judgment: perform hierarchical bounding box collision detection algorithm (apply OBBTree or k-DOPs structure) on the position of the abnormal point, identify the parts with which the grid intersects, and define as direct physical contact relationship.

[0127] Indirect contact tracing: based on the assembly topological relationship, recursively search for the adjacent parts connected by the directly physically contacted parts, and construct the physical conduction link:

[0128] If the parts are connected by bolts / welds, or there is a conduction medium with thermal conductivity > 5W / (m·K) (such as copper bar, silicone grease), it is defined as indirect physical contact relationship.

[0129] Thermal path map construction: store all directly and indirectly contacted parts in the relational graph database according to the order of thermal conduction path, and label the effective contact area and intermediate material thermal conductivity between each node.

[0130] Dynamic update mechanism: when the pose of the parts in the virtual scene changes, recalculate the contact detection and thermal link in real time to ensure that the contact relationship is consistent with the physical working condition.

[0131] Through three-order analysis of collision detection + topological recursion + link modeling, it fully covers the industrial scene requirements from direct contact (mechanical collision) to indirect contact (thermal conduction path).

[0132] S302, calculate the distance of the shortest physical contact path between the mutually related parts and the abnormal point , and the volume of each part , so as to calculate the weight index between the parts and the abnormal point. Specifically, it includes:

[0133] S3021, analyze the thermal and force distribution of each part related to the abnormal point when working normally, calculate the thermal and force distribution gradient norm , and substitute it into the formula to calculate the heat contribution efficiency index of the part :

[0134] ;

[0135] wherein, , , , , are adaptive weight coefficients, is the system characteristic thermal diffusion length.

[0136] is the path distance attenuation, which quantifies the thermal path resistance, to avoid zero distance singularity.

[0137] is the thermal distribution enhancement factor, which is used to compress the volume effect to avoid large volume monopoly, multiplied by the temperature gradient to highlight the contribution of local high temperature area.

[0138] The heat conduction attenuation model (analogous to Fourier's law) is introduced, to correlate material properties.

[0139] The heat contribution efficiency index fuses the path physical conduction characteristics + thermal distribution gradient + volume compression effect, providing a quantitative basis for the attribution of multi-heat source coupled temperature changes.

[0140] The thermal force distribution gradient norm calculation includes:

[0141] First, collect multi-angle thermal imaging of power equipment under rated working condition by infrared thermal imager, or call the steady-state temperature field distribution data exported by CFD simulation software.

[0142] Second, perform trilinear interpolation on the thermal imaging data to generate a three-dimensional temperature matrix matched with the digital model grid.

[0143] Then, for each grid node, calculate its thermal force gradient using the central difference method, and take the Euclidean norm of the gradient vector.

[0144] Finally, take the weighted average (weight = area ratio of the face to which the node belongs) of the gradient norm of all grid nodes belonging to the same component, and output the thermal force distribution gradient norm of the component as a whole.

[0145] S3022, respectively calculate the heat contribution efficiency index of each component correlated with the abnormal point , and substitute it into the formula to calculate the weight index between each component and the abnormal point

[0146] ;

[0147] wherein, is the first Components and anomalies The weighting index between them To and anomalies The number of interconnected parts, It is the gain factor.

[0148] Gain factor is used to amplify the influence of key components: when High-intensity time Weight of the value. Time-smooth distribution. Gain factor. Supports dynamic adjustment: When the load on power equipment fluctuates, adjustments are made... Matching the measured temperature rise distribution improves the sensitivity of anomaly attribution.

[0149] For the first Components and anomalies The thermal contribution efficiency index between them. For the first Components and anomalies The thermal contribution efficiency index between them.

[0150] By introducing the exponential function The simulation of the nonlinear cumulative effect of heat conduction (the contribution of the high-temperature region decreases exponentially with distance) makes high... The weight of components is more significant (such as components near the source of failure).

[0151] S3023, and so on, calculate the weight index for all associated components under each anomaly point.

[0152] S303. Based on the difference between the current temperature and the predicted temperature at each anomaly point, and combined with a weighted index, calculate the anomaly index for each component. Components with an anomaly index greater than a threshold are selected as anomaly components. Anomaly Index The calculation formula is:

[0153] ;

[0154] In the formula, The number of anomalies associated with the components. For components and the first The weight index between outliers and The first The current and predicted temperatures of each anomaly point.

[0155] The physical contact relationship (direct or indirect) is recognized by an artificial intelligence algorithm (such as OBBTree collision detection) to construct a heat path atlas; the shortest physical path distance and the heat contribution efficiency index (HCEI) of the components are calculated; and the abnormal index (AN) of each component is calculated based on the weight index to screen abnormal components.

[0156] The fault tracing under the coupling of multiple physical fields is realized, and the accuracy of fault positioning is improved. Through heat path analysis, the adaptability under high load working conditions is ensured, and false attribution caused by heat conduction is avoided.

[0157] S400, arrange each abnormal component according to the abnormal index, and display the abnormal components in order and prewarn to the control center.

[0158] The abnormal components are sorted from high to low according to the abnormal index. In the virtual scene power equipment model, each abnormal component is dynamically layered and highlighted:

[0159] The abnormal component with the highest abnormal index is warned with a pulse red light, the abnormal component with the second highest abnormal index is marked with an orange outline frame, and the remaining abnormal components are highlighted. At the same time, the heat distribution cloud picture is fused with the real-time three-dimensional temperature field rendering.

[0160] When the abnormal index of the highest abnormal component exceeds the dynamic threshold, a first-level fault prewarning is automatically pushed to the control center, and when the abnormal index of the remaining abnormal components continuously rises for two periods, a second-level maintenance alarm is triggered.

[0161] Finally, a diagnosis report containing spatial coordinates, abnormal index and historical data is generated, and the sound and light alarm system of the control center is activated synchronously.

[0162] The abnormal components are sorted according to the abnormal index (AN). In the virtual model, the abnormal components are highlighted in layers (such as pulse red light and orange outline). A diagnosis report is generated. When the index exceeds the threshold, a first-level prewarning or a second-level alarm is pushed. The sound and light alarm system of the control center is activated.

[0163] A multi-dimensional visualization interface is provided, which optimizes the operation and maintenance response strategy and shortens the fault diagnosis time. The hierarchical warning mechanism (first-level fault prewarning and second-level maintenance alarm) reduces unnecessary downtime loss, and immediate intervention is realized through sound and light alarm, which improves the overall safety of the system.

[0164] Please refer to Figure 2 The application also provides an artificial intelligence-based power equipment analysis system, which comprises a data acquisition module, an intelligent analysis module, an equipment management module and a visualization module.

[0165] The data acquisition module is used to acquire temperature data, operating parameters and historical logs of the power equipment, and environmental parameters around the power equipment, and generate a power equipment model containing components.

[0166] Through multi-angle acquisition of power equipment thermal imaging data by infrared thermal imager, combined with operating parameters (power, digital model), historical logs (power-temperature distribution record) and environmental parameters (temperature, humidity, wind speed, etc.), a digital model is established, and the three-dimensional temperature field is mapped to the model surface to realize temperature distribution visualization.

[0167] Fusion of multi-source heterogeneous data generates 3D equipment model with temperature distribution, solves the limitations of traditional single temperature monitoring, and provides spatialized data basis for anomaly analysis.

[0168] The intelligent analysis module sets sampling points on the surface of the power equipment model and analyzes historical log data records. According to the data records, the relationship is fitted for the sampling points, and the abnormal points are set.

[0169] Poisson disc algorithm is used to uniformly distribute sampling points on the model surface. Based on historical log data records, the same power data records are selected, and the average value of environmental factors during the stable temperature period is intercepted. The relationship between temperature and environmental factors at each sampling point is fitted by linear regression. By comparing the predicted temperature with the actual temperature, the abnormal points with temperature difference exceeding the threshold are calibrated.

[0170] Quantify the influence of environmental factors on temperature to reduce false positive rate. Dynamically identify local temperature rise anomalies to improve fault location accuracy.

[0171] The equipment management module identifies and associates the abnormal points and the components in physical contact in the power equipment model through artificial intelligence algorithm, analyzes the shortest physical contact path between them and calculates the abnormal index of each component, and selects abnormal components.

[0172] Combined with bounding box collision detection (such as OBBTree) and assembly topology recursion, identify directly / indirectly contacted components (bolt connection, heat conduction medium > 5W / (m·K)).

[0173] Construct heat path atlas (store contact relationship, thermal conductivity), calculate shortest path distance and heat contribution efficiency index. Based on weight index, weight the temperature rise deviation of multiple abnormal points to select AN components exceeding the threshold.

[0174] Realize fault attribution under multi-physical field coupling, analyze the shortest physical contact path between them, and dynamically update the heat conduction link to ensure the adaptability of working conditions.

[0175] The visualization module arranges each abnormal component according to the abnormal index, displays the abnormal components in order and warns the control center.

[0176] Parts are ranked by anomaly index, using a hierarchical alert strategy (pulsed red light > orange outline > highlight). Superimpose real-time temperature field heat map. First-level warning (AN super-dynamic threshold) pushes control center, second-level alarm (AN continues to rise) triggers maintenance reminder. Generate diagnostic report containing spatial coordinates and activate audible and visual alarms.

[0177] Multi-dimensional visualization improves fault identification efficiency. Hierarchical warning mechanism optimizes operation and maintenance response strategy, reducing downtime losses.

[0178] The system forms a diagnostic closed loop from collection (multi-source data integration) → analysis (environmental factor regression) → management (physical conduction modeling) → visualization (hierarchical warning).

[0179] Through a group of artificial intelligence algorithms such as Poisson sampling, bounding box detection, regression model, and heat conduction formula, intelligent diagnosis of industrial scenes is realized.

[0180] Example 1: Assuming that A1, A2, and A3 parts are respectively associated with an anomaly point, and the heat contribution efficiency indexes are 2, 1.2, and 0.5 respectively; when the gain factor is 1.5, the weight index between these parts and the anomaly point is calculated respectively:

[0181] A1 weight index: ;

[0182] A2 weight index: ;

[0183] A3 weight index: ;

[0184] Then the weight indexes between A1, A2, and A3 parts and the anomaly point are 0.71, 0.21, and 0.08 respectively.

[0185] It should be noted that in this article, relational terms such as first and second are used only to distinguish one entity or action from another entity or action, and do not necessarily require or imply any such actual relationship or order between these entities or actions. Moreover, the terms "include", "contain" or any other variant thereof are intended to cover non-exclusive inclusion, so that the process, method, article or device including a series of elements not only includes those elements, but also includes other elements not explicitly listed or inherent to such process, method, article or device.

[0186] Finally, it should be noted that the above only describes the preferred embodiments of the present application and is not intended to limit the present application. Although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art will appreciate that the technical solutions described in the foregoing embodiments can be modified or some technical features thereof can be replaced by equivalent ones. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included in the protection scope of the present application.

Claims

1. An artificial intelligence-based power equipment analysis method, characterized by: The method comprises: S100, collecting temperature data, operating parameters and historical logs of the power equipment, and environmental parameters of the surroundings; analyzing the operating parameters and the temperature data, generating a power equipment model containing components, and establishing a virtual scene; S200, setting sampling points on the surface of the power equipment model in the virtual scene, analyzing historical logs to screen data records; fitting a relationship for the sampling points according to the data records and calculating a predicted temperature, and setting an abnormal point by comparing the temperature data; specifically comprising: S201、Set the sampling density , obtain the surface mesh data of the power equipment model in the virtual scene, divide the total surface area by the sampling density to obtain the total number of sampling points ; S202、based on Poisson disc distribution algorithm, uniform iteration generation on the surface of the grid a three-dimensional sampling point; a mapping relationship between the sampling point and the temperature field is established, and the spatial temperature data attachment is completed; S203、Obtain the current power of the power equipment In the historical log, all data records with power are screened out; according to the surface temperature distribution of each data record, the temperature of each sampling point is analyzed; S204, analyzing temperature changes of each sampling point within the duration of the data records, intercepting each environmental factor within the time period during which the temperature of each sampling point remains unchanged, and thereby fitting a relationship between the temperature and the environmental factors for each sampling point; S205, obtaining the environmental factor in the current environmental parameter, substituting into the relational expression of each sampling point to obtain the predicted temperature ; obtaining the current temperature of each sampling point , the temperature of the sampling point greater than the threshold value is taken as an abnormal point; S300, identifying and associating components in the power equipment model that have physical contact with the abnormal point through an artificial intelligence algorithm, analyzing the shortest physical contact path between the two, and calculating an abnormality index of each component, thereby screening out abnormal components; specifically comprising: S301, analyzing the power equipment model in the virtual scene, identifying the physical contact between the components and the position of the abnormal point through an artificial intelligence algorithm, and associating components having direct or indirect physical contact with the abnormal point; S302、calculate the distance of the shortest physical contact path between the correlated parts and the abnormal point , and the volume of each part , thereby calculating the weight index between the parts and the abnormal point; S303, calculating the abnormality index of each component according to the difference between the current temperature and the predicted temperature of each abnormal point, combining a weight index, and screening out components with an abnormality index greater than a threshold value as abnormal components; S400, arranging the abnormal components according to the abnormality index, displaying the abnormal components in order, and warning the control center. 2.The artificial intelligence-based electric power equipment analysis method of claim 1, wherein: In S100, the temperature data refers to thermal imaging pictures of the power equipment at different angles collected using an infrared thermal imager; The operating parameters refer to the power of the power equipment and a digital model, which contains component models that can generate heat during operation; The historical logs refer to data records of surface temperature distribution when the power equipment operates at different powers each time, and each data record includes power, environmental parameters and surface temperature distribution; The environmental parameters include different environmental factors, specifically physical indicators that quantitatively describe the environmental state; The virtual scene is established and the digital model of the power equipment is put in; the orientation and angle of the power equipment in different thermal imaging pictures in the temperature data are analyzed, and the multi-angle thermal imaging pictures are processed into three-dimensional temperature field data; The three-dimensional temperature field data is mapped to the surface of the digital model of the power equipment, generating a digital model with superimposed temperature distribution. 3.The artificial intelligence-based electric power equipment analysis method of claim 1, wherein: S204 includes: S2041、acquire sampling points temperature change within the duration of the data record acquire sampling points time period during which the temperature remains constant, and calculate the average value of each environmental factor within each time period S2042, intercept sampling points in each data record in turn The time period of constant temperature is maintained, and the average value of each environmental factor is calculated, and the sampling points are counted The number of time periods e in all data records S2043、count the number of items of environmental factors The temperature in each time period is taken as the dependent variable, The environmental factors are taken as the independent variables respectively and packed as samples. S2044, set intercept and regression coefficient , and establish linear regression model; the independent variable in the e samples is respectively taken as input value , the output value of each sample The difference between the dependent variable and the difference coefficient; the expression is as follows: ； S2045, adjusting the intercept and regression coefficient until the sum of the gap coefficients of all samples is minimized, obtaining the fitted relationship; in this way, the relationship between the temperature and the environmental factors is fitted for each sampling point. 4.The artificial intelligence-based electric power equipment analysis method of claim 1, wherein: S302 includes: S3021、analysis with abnormal points The thermal power distribution of each part when working normally is correlated, and the thermal power distribution gradient norm is calculated The heat contribution efficiency index of the part is calculated by substituting the formula : ； wherein , , , , are adaptive weight coefficients, is the system characteristic thermal diffusion length; S3022、respectively calculate the weight index between each component and the abnormal point The heat contribution efficiency index of each component is correlated, and the formula is used to calculate the weight index between each component and the abnormal point respectively ； wherein is the weight index between the th component and the abnormal point, is the weight index between the th component and the abnormal point, is the number of components associated with the abnormal point, is the gain factor; a heat contribution efficiency index between the nth component and the anomaly point; a heat contribution efficiency index between the nth component and the anomaly point; a heat contribution efficiency index between the nth component and the anomaly point; a heat contribution efficiency index between the nth component and the anomaly point; a heat contribution efficiency index between the nth component and the anomaly point; a heat contribution efficiency index between the nth component and the anomaly point; S3023, in this way, the weight index is calculated for all associated components under each abnormal point. 5.The artificial intelligence-based electric power equipment analysis method of claim 1, wherein: In S303, the abnormality index of the component The calculation formula is: ； In the formula, is the number of abnormal points associated with the component, is the weight index between the component and the th abnormal point, and are the current temperature and the predicted temperature of the th abnormal point, respectively. 6.The artificial intelligence-based electric power equipment analysis method of claim 1, wherein: In S400, specifically comprising: The abnormal components are sorted from high to low according to the abnormality index; in the power equipment model of the virtual scene, each abnormal component is dynamically highlighted in layers: The abnormal parts with the highest abnormality index are alarmed by pulsed red light, the abnormal parts with the second highest abnormality index are marked by orange outline frame, and the rest abnormal parts are highlighted; meanwhile, the real-time three-dimensional temperature field is fused to render a thermal distribution cloud picture; When the abnormality index of the highest abnormal part exceeds the dynamic threshold, a first-level fault early warning is automatically pushed to the control center, and when the abnormality index of the rest abnormal parts continuously rises for two periods, a second-level maintenance alarm is triggered; A diagnosis report containing spatial coordinates, abnormality index and historical data is finally generated, and a sound and light alarm system in the control center is activated synchronously.

7. The power equipment analysis system based on artificial intelligence applied to the power equipment analysis method based on artificial intelligence according to claim 1, characterized in that: The system comprises a data acquisition module, an intelligent analysis module, a device management module and a visualization module; The data acquisition module is used to collect temperature data, operating parameters and historical logs of the power equipment, as well as environmental parameters around the power equipment, and generate a power equipment model containing parts; The intelligent analysis module sets sampling points on the surface of the power equipment model, analyzes historical logs to screen data records, fits a relationship for the sampling points according to the data records, and thus sets abnormal points; The device management module identifies and associates parts in physical contact with the abnormal points in the power equipment model by using an artificial intelligence algorithm, analyzes the shortest physical contact path between the two, calculates the abnormality index of each part, and screens abnormal parts; The visualization module arranges the abnormal parts according to the abnormality index, displays the abnormal parts in order and warns the control center.

Images