A substation power equipment fault diagnosis method and system

Abstract

The application belongs to the technical field of data processing, and particularly relates to a substation power equipment fault diagnosis method and system, which comprises the following steps: obtaining real-time load current of a line where the equipment is located from a SCADA system; positioning a contact area of an infrared thermal image by using an image recognition algorithm, and extracting the highest temperature of the area as an image highest temperature; calculating an average temperature value of a non-contact area as an environment temperature; reading a rated current of the equipment from a database; calculating a current thermal intensity index, a vibration dispersion factor and a composite risk index; generating a diagnosis report of an abnormal equipment; triggering an alarm pushing mechanism; and storing the collected data and updating a statistical model. The application effectively solves the technical problems that the prior art cannot effectively eliminate load interference, cannot easily capture early mechanical loosening characteristics, and results in low fault diagnosis accuracy.

Description

Technical Field

[0001] This invention relates to the field of data processing technology. More specifically, this invention relates to a method and system for diagnosing faults in substation power equipment. Background Technology

[0002] As the hub of the power system, the operational reliability of key equipment such as high-voltage disconnect switches and conductor joints in substations directly affects the safety and stability of the power grid. Currently, routine inspections of power equipment mainly rely on infrared thermal imaging and visible light video technology to identify potential faults by capturing the absolute temperature distribution or appearance changes on the equipment surface. This non-contact detection method, due to its intuitiveness and safety, has been widely used in the daily operation and maintenance of substations. It can effectively detect some serious defects in the late stages of overheating and is currently a core monitoring method in the power grid operation and maintenance system.

[0003] However, existing technologies still have significant limitations in diagnostic accuracy under complex operating conditions. On the one hand, under high-load operation, the background temperature rise of normal equipment due to the Joule effect often exceeds the preset threshold, which is easily misjudged by the system as a poor contact fault, leading to frequent false alarms. On the other hand, the early identification capability for hidden loosening is insufficient. In the early stage, loose mechanical connections mainly manifest as micro-vibration rather than significant temperature rise. Traditional thermal imaging technology is extremely insensitive to such hidden faults, resulting in the loss of the optimal maintenance window. More importantly, existing monitoring methods lack the evaluation and analysis of the coupling mechanism of electro-thermal-mechanical multi-physics fields, making it difficult to clarify the nonlinear causal relationship between load fluctuations, mechanical vibration, and contact heating, resulting in poor robustness of fault diagnosis.

[0004] These limitations not only make it easy to misdiagnose faults under high load conditions due to false high temperatures, but also cause potential problems to be missed under low load conditions due to the insensitivity of traditional technologies to hidden loosening. This not only increases the ineffective workload of operation and maintenance personnel and drags down the overall operation and maintenance efficiency, but may also cause small problems to develop into major faults by missing the early maintenance window, thus creating potential risks to the safe operation of the power grid. It is necessary to further improve the accuracy of fault diagnosis under complex operating conditions and the reliability of long-term operation through intelligent diagnostic solutions that integrate multi-source data and solve load and environmental interference. Summary of the Invention

[0005] To address the technical problems of low fault diagnosis accuracy caused by the inability of existing technologies to effectively eliminate load interference and the difficulty in capturing early mechanical loosening characteristics, the present invention provides solutions in the following aspects.

[0006] In a first aspect, the present invention provides a method for diagnosing faults in substation power equipment, comprising:

[0007] The system uses a multispectral inspection robot to collect real-time infrared thermal images and high-frame-rate visible light video of the target equipment, and obtains the real-time load current of the circuit where the equipment is located from the SCADA system. Image recognition algorithms are used to locate the contact area in the infrared thermal image, and the highest temperature in that area is extracted and recorded as the highest temperature in the image. The average temperature value of the non-contact area is calculated and recorded as the ambient temperature. The spatial temperature gradient is obtained based on the ratio of the temperature difference to the distance difference from the center point to the edge point of the contact area. The rated current of the equipment is read from the database. Based on the principle that the conductor's heating power is proportional to the square of the current, the current thermal intensity index is calculated using the highest temperature in the image, the ambient temperature, and the real-time load current. The vibration dispersion factor, which characterizes the looseness of the equipment's mechanical connections, is calculated using the video's microscopic amplitude and its historical statistical characteristics. By integrating the electrothermal effect, mechanical stability, and thermal field distribution, a composite risk index is calculated based on the current thermal intensity index and the vibration dispersion factor. The health status of the target equipment is judged based on the magnitude of the composite risk index. The system summarizes the above data, generates a diagnostic report for abnormal equipment, triggers an alarm push mechanism, stores the collected data, and updates the statistical model.

[0008] This invention effectively addresses various issues encountered in equipment fault diagnosis, eliminating interference from current load variations and avoiding false alarms or missed diagnoses due to varying load levels. Simultaneously, it can detect early signs of mechanical loosening in equipment, preventing these potential problems from being overlooked due to the lack of obvious overheating. By integrating multiple operational information sources and comprehensively assessing equipment status, it reduces the limitations of relying on single indicators and improves the accuracy of fault diagnosis. Whether under high or low load conditions, and whether dealing with overheating or mechanical loosening issues, it provides relatively accurate identification, making maintenance work more targeted, reducing unnecessary workload due to misdiagnosis, and minimizing the risk of escalating problems due to missed diagnoses, thus making equipment fault diagnosis more aligned with actual needs.

[0009] Preferably, the real-time load current of the line where the device is located is obtained from the SCADA system; the contact area of ​​the infrared thermal image is located using an image recognition algorithm, and the highest temperature of this area is extracted and recorded as the highest temperature in the image; the average temperature value of the non-contact area is calculated and recorded as the ambient temperature; the rated current of the device is read from the database, including:

[0010] The system obtains the real-time load current of the line where the target device is located from the SCADA system; it uses an image recognition algorithm to perform target detection on the infrared thermal image acquired in real time by the multispectral inspection robot to obtain the contact area; it extracts the highest pixel temperature value of the area and records it as the highest temperature of the image; it defines the non-contact area in the infrared thermal image as the background area, calculates the average temperature value of the background area and records it as the ambient temperature; and it reads the rated current of the device from the database.

[0011] Preferably, the current thermal index satisfies the following expression:

[0012] ;

[0013] In the formula, Indicates the thermal intensity index of electric current; The highest temperature in the image; Ambient temperature; This is the real-time load current; Rated current; It is the natural logarithm function; It is a natural constant; It is the first smallest positive number, and the denominator is guaranteed to be non-zero.

[0014] This invention can isolate the influence of current fluctuations on equipment temperature, accurately reflecting the equipment's own heat-related status. Regardless of whether the equipment is operating under high or low current conditions, this index can clearly identify any abnormal heating, avoiding misjudgments of normal heating or omissions of potential heating problems due to current fluctuations. It makes the heating status of equipment under different load conditions comparable, helping to accurately grasp the equipment's heating situation in various operating scenarios and providing strong support for fault diagnosis.

[0015] Preferably, the vibration dispersion factor satisfies the following expression:

[0016] A phase-based motion amplification and estimation algorithm is applied to the acquired high frame rate visible light video to calculate its displacement peak value, which is denoted as the video micro amplitude.

[0017] The vibration discretization factor satisfies the following expression:

[0018] ;

[0019] In the formula, Represents the vibration dispersion factor; Indicates the microscopic amplitude of the video; Indicates the historical average vibration; Indicates historical vibration standard deviation; If it is the second smallest positive number, the denominator must not be 0; Represents an exponential function with the natural constant as its base; This represents the maximum value function.

[0020] The vibration dispersion factor in this invention can accurately capture signs of loosening during equipment operation, while avoiding overlooking potential overheating issues due to stable equipment operation. When abnormal equipment vibration occurs, the calculation results amplify this signal, making potential mechanical problems easier to detect; while when the equipment is operating smoothly, it does not interfere with the judgment of overheating issues, ensuring that other potential risks are not overlooked while focusing on the mechanical condition.

[0021] Preferably, the composite risk index satisfies the following expression:

[0022] ;

[0023] In the formula, Indicates a composite risk index; The electric current thermal intensity index; The vibration dispersion factor; For space temperature gradient; It is the third smallest positive number; It is the hyperbolic tangent function.

[0024] The composite risk index in this invention integrates information from multiple aspects, including equipment heating, vibration, and temperature distribution, to comprehensively reflect the equipment's health status. It avoids the one-sidedness that can result from judging based on a single indicator, considering not only the electrical condition of the equipment but also its mechanical stability and temperature distribution characteristics. This makes the assessment of equipment health status more comprehensive and closer to actual operating conditions. Through this fusion of multi-dimensional information, misjudgments or omissions caused by relying on a single indicator are reduced, providing a more reliable basis for operation and maintenance decisions and giving operation and maintenance work more direction.

[0025] Preferably, the health status of the target device is determined, including:

[0026] The preset first benchmark sensitivity coefficient is multiplied by the natural exponential function value of the historical vibration standard deviation and recorded as the first dynamic threshold. The preset second benchmark sensitivity coefficient is multiplied by the natural exponential function value of the historical vibration standard deviation and recorded as the second dynamic threshold. The first benchmark sensitivity coefficient is less than the second benchmark sensitivity coefficient. If the composite risk index is less than or equal to the first dynamic threshold, the target equipment is determined to be normal. If the composite risk index is greater than the first dynamic threshold and less than the second dynamic threshold, the target equipment is determined to be under attention. If the composite risk index is greater than or equal to the second dynamic threshold, the target equipment is determined to be abnormal and enters the subsequent fault handling process.

[0027] This invention, by dynamically setting judgment criteria, can adapt to the operating characteristics and aging levels of different equipment. For equipment with a long operating history, it avoids frequent false alarms due to natural changes in its basic condition; for new equipment, it can also accurately detect minor anomalies in the early stages, avoiding missed diagnoses. This judgment method, which is tailored to the actual situation of the equipment, makes the assessment of health status more reasonable, reduces ineffective maintenance work, and can also promptly identify potential problems that require attention, improving the targeting and efficiency of maintenance and allowing for more rational use of maintenance resources.

[0028] Preferably, generating a diagnostic report for the malfunctioning device includes:

[0029] If the target power equipment is in an abnormal health state, then the fault location and composite risk index will be summarized to generate a diagnostic report for this abnormal equipment.

[0030] Preferably, the alarm push mechanism includes:

[0031] The system triggers the audible and visual alarm through the substation integrated automation interface, and the alarm's push mechanism pushes alarm information to the handheld terminal APP of the operation and maintenance personnel in real time.

[0032] Preferably, storing the collected data and updating the statistical model includes:

[0033] If the health status of the target equipment is determined to be normal, the historical vibration mean and historical vibration standard deviation of the equipment are recalculated and updated using new data to achieve dynamic updating and optimization of the diagnostic model; if it is determined to be in an abnormal state, no update is performed.

[0034] Secondly, the present invention provides a substation power equipment fault diagnosis system, including a processor and a memory, wherein the memory stores computer program instructions, and when the computer program instructions are executed by the processor, the above-mentioned substation power equipment fault diagnosis method is implemented.

[0035] By adopting the above technical solution, a computer program for diagnosing faults in substation power equipment is generated and stored in a memory for loading and execution by a processor. Terminal equipment is then created based on the memory and processor for convenient use.

[0036] The beneficial effects of this invention are as follows: It provides practical support for the daily operation and maintenance of substation equipment. By collecting data, analyzing status, and issuing warnings in a scientific manner, it helps maintenance personnel to promptly grasp the equipment's operating status, identify potential problems in advance, and address them. This helps reduce the frequency of equipment failures, minimize losses caused by failures, and ensure the stable operation of the power grid. Simultaneously, reasonable judgment criteria and model update mechanisms allow diagnostic work to adapt to the characteristics of different equipment and long-term operational changes, reducing unnecessary maintenance investment and improving resource utilization efficiency. Its overall design aligns with the actual operational needs of substations, can be integrated into daily operation and maintenance processes, provides continuous technical support for the safe and stable operation of the power grid, and can play a positive role in practical applications, contributing to the reliable operation of the power system. Attached Figure Description

[0037] Figure 1 This schematically illustrates a flowchart of a substation power equipment fault diagnosis method according to the present invention;

[0038] Figure 2 This diagram schematically illustrates the multi-level coupling transmission model of vibration to the composite risk index in this invention.

[0039] Figure 3 The diagram illustrates the dynamic threshold determination logic based on historical vibration standard deviation in this invention. Detailed Implementation

[0040] This invention discloses a method for diagnosing faults in substation power equipment, referring to... Figure 1 This includes steps S1-S4:

[0041] S1: The multispectral inspection robot collects infrared thermal images and high frame rate visible light videos of the target equipment in real time, and obtains the real-time load current of the line where the equipment is located from the SCADA system; the contact area of ​​the infrared thermal image is located using an image recognition algorithm, and the highest temperature of the area is extracted and recorded as the highest temperature of the image; the average temperature value of the non-contact area is calculated and recorded as the ambient temperature; the spatial temperature gradient is obtained based on the ratio of the temperature difference from the center point to the edge point of the contact area to the distance difference; the rated current of the equipment is read from the database.

[0042] It is important to note that in actual substation operation, equipment condition assessment often faces the challenge of single-dimensional distortion. While infrared thermography can capture temperature anomalies, temperature is a lag variable with thermal inertia and cannot reflect transient mechanical shocks in real time. Visible light video, though intuitive, is difficult for the naked eye to detect micron-level mechanical loosening and vibration. Current data from SCADA systems only represents load conditions and cannot directly indicate equipment health. Analyzing these data in isolation can easily lead to misjudgments. For example, normal temperature rise under high load may be misjudged as overheating, while poor contact under low load may be missed due to insignificant temperature rise. From the perspective of multi-physics coupling, faults often begin with mechanical loosening, evolve into increased contact resistance, and ultimately manifest as localized heating. Therefore, it is necessary to establish a spatiotemporally synchronized multi-source data acquisition mechanism to align the temperature field representing the result, the video micro-amplitude field representing the cause, and the current field representing the background of the operating conditions. This step is not just about collecting data, but also about solving various interference factors in a unified physical coordinate system, ensuring the multidimensionality and correlation of the diagnostic model's input data, and thus laying a solid data foundation for capturing fleeting early fault characteristics.

[0043] Specifically, a multispectral inspection robot acquires infrared thermal images of the target equipment in real time and obtains the real-time load current of the circuit where the equipment is located from the SCADA system; an image recognition algorithm is used to locate the contact area of ​​the infrared thermal image, and the highest temperature in that area is extracted and recorded as the highest temperature in the image; the average temperature value of the non-contact area is calculated and recorded as the ambient temperature; the rated current of the equipment is read from the database, including:

[0044] Using a multispectral inspection robot within the substation, infrared thermal images and high-frame-rate visible light video of the target equipment are simultaneously acquired in real time, and the real-time load current of the line where the equipment is located is obtained from the SCADA system. Image recognition algorithms are used to perform target detection on the infrared thermal images, automatically locating the key connection parts of the power equipment to be inspected and segmenting the contact area. The highest pixel temperature value of the contact area is extracted and recorded as the highest temperature in the image, and the geometric coordinates of the pixel with the highest temperature in the image are located and recorded as the coordinates of the highest temperature point. The non-contact area in the infrared thermal image is defined as the background area, and the average temperature value of the background area is calculated and recorded as the ambient temperature. Combining the robot's acquisition distance, the pixel spacing of the infrared thermal image is mapped to the physical distance, and the ratio of the absolute value of the temperature difference from the center point to the edge point of the contact area to the physical distance is calculated and recorded as the spatial temperature gradient. The equipment number, rated current, historical vibration mean, and historical vibration standard deviation of the equipment under normal operating conditions are read from the database to calculate the first and second reference sensitivity coefficients of the dynamic threshold.

[0045] Thus, the real-time load current of the target device, the highest temperature in the image, the ambient temperature, the video micro-amplitude, the historical vibration mean, the historical vibration standard deviation, the spatial temperature gradient, the coordinates of the highest temperature point, the device number, the reference sensitivity coefficient, and the rated current were obtained.

[0046] S2: Based on the principle that the conductor's heating power is proportional to the square of the current, the current thermal intensity index is calculated using the highest temperature in the image, the ambient temperature, and the real-time load current. The vibration dispersion factor, which characterizes the looseness of the equipment's mechanical connections, is calculated using the video's micro-amplitude and its historical statistical characteristics. By integrating the electrothermal effect, mechanical stability, and thermal field distribution, the composite risk index is calculated based on the current thermal intensity index and the vibration dispersion factor.

[0047] It should be noted that in the operation and maintenance of power equipment, it is difficult to distinguish between normal load temperature rise and fault-induced temperature rise. According to Joule's law, the heat output of a conductor is proportional to the square of the current. During peak summer load periods, even with good equipment connections, high current can still lead to a significant temperature rise. In this case, relying solely on absolute temperature threshold alarms will result in numerous false alarms. Conversely, during low winter load periods, even with severe contact defects, the heat generated by the low current may not be sufficient to trigger a high-temperature alarm, leading to the failure to detect serious hidden dangers. From a thermodynamic perspective, we need to isolate the direct impact of current fluctuations on temperature and find a physical quantity that can characterize the essential properties of the contact interface. Therefore, this invention introduces a current thermal intensity index, which normalizes the temperature rise relative to the square of the current. Essentially, it calculates a dynamic equivalent thermal resistance. Simultaneously, considering the nonlinear aging characteristics of metallic materials under high temperature and high current—that is, the closer the current is to the rated value, the faster the aging chain reaction—a nonlinear gain term is introduced. The purpose of this is to give the algorithm load adaptive capability. Regardless of current fluctuations, this index can stably point to the true health state of the contact point, thereby effectively eliminating interference caused by load fluctuations. Furthermore, to prevent abnormal values ​​caused by the current approaching zero under power outage or extremely low load conditions, this invention only operates when the real-time load current is greater than the preset minimum sustaining current.

[0048] Specifically, based on the principle that the conductor's heating power is proportional to the square of the current, the current thermal index is calculated using the highest temperature in the image, the ambient temperature, and the real-time load current, including:

[0049] The current thermal index satisfies the following expression:

[0050] ;

[0051] In the formula, Indicates the thermal intensity index of electric current; The highest temperature in the image; Ambient temperature; This is the real-time load current; Rated current; It is the natural logarithm function; It is a natural constant; It is the first smallest positive number, and the denominator is guaranteed to be non-zero.

[0052] In the formula, It characterizes the relative temperature rise induced by the square of a unit current, i.e., the equivalent thermal resistance. This term can effectively offset the natural temperature change caused by load fluctuations and separate the fault characteristics from the current thermal effect. It is a nonlinear gain term. Considering that the aging rate and heat accumulation effect of the contact surface will be accelerated nonlinearly under high current conditions, this term is used to appropriately improve the sensitivity to thermal defects under heavy load conditions and prevent missed detection caused by the large current masking early minor defects. The entire current thermal intensity index expression is constructed by multiplying the normalized equivalent thermal resistance with the load-related nonlinear gain, so that the true thermal defect intensity of the current contact point can be calculated more accurately under different load conditions.

[0053] For example, if a certain disconnecting switch , Current current Rated current , ,but This value accurately calculates the current thermal defect intensity. To retain six decimal places.

[0054] Thus, the current thermal intensity index of the target device was obtained.

[0055] It should be noted that in the joint monitoring scenarios of complex electromechanical equipment, thermal faults, such as heat generated by poor contact or insulation aging, often exhibit a non-linear coupling relationship with mechanical faults, such as vibration caused by structural loosening. In the original calculation logic, if the equipment is in an abnormally stable state that is quieter than usual, i.e., the current micro-amplitude is less than the historical average, direct calculation will result in a negative difference, which in turn makes the vibration dispersion factor negative. This treatment will cause the subsequent risk adjustment term to be less than 1, thus incorrectly reducing the thermal risk index that should be taken seriously. From the perspective of physical operation, the stability of the mechanical state does not mean that there is no overheating risk in the electrical connection parts. Stability should never be a reason to cover up overheating. This logical flaw can easily lead to the system missing reports when the equipment overheating is not obvious, i.e., stable operation masks the serious safety hazard of thermal faults. Therefore, this step must introduce a maximum value function to enforce from the algorithm's underlying logic that only vibration degradation is considered. That is, only when mechanical vibration is abnormally enhanced will it be used as a weighting factor to amplify the overall risk. When the equipment is running smoothly, this adjustment term will be automatically reset to zero, no longer having any inhibitory effect on the thermal risk index. This ensures that the system's sensitivity to thermal faults is not affected by improvements in mechanical condition, and achieves the safety monitoring principle of "better to falsely report loosening than to miss overheating."

[0056] Preferably, the vibration dispersion factor characterizing the degree of looseness of the mechanical connection of the equipment is calculated using the micro-amplitude of the video and its historical statistical characteristics, including:

[0057] A phase-based motion amplification and estimation algorithm is applied to the acquired high frame rate visible light video to calculate its displacement peak value, which is denoted as the video micro amplitude.

[0058] The vibration discretization factor satisfies the following expression:

[0059] ;

[0060] In the formula, Represents the vibration dispersion factor; Indicates the microscopic amplitude of the video; Indicates the historical average vibration; Indicates historical vibration standard deviation; If it is the second smallest positive number, the denominator must not be 0; Represents an exponential function with the natural constant as its base; This represents the maximum value function.

[0061] In the formula, by introducing ,when When the molecule is 0, it makes The adjustment coefficient in subsequent steps will remain at 1; only when hour, With an adjustment coefficient greater than 1, the risk of thermal failure accompanied by abnormal vibration is reasonably amplified.

[0062] For example, at a certain moment Historical vibration mean Historical vibration standard deviation ,but The high score clearly indicates a loosening of the mechanical structure. Round to two decimal places.

[0063] Thus, the vibration dispersion factor of the target equipment was obtained.

[0064] It should be noted that in the fault evolution process of substation equipment, thermal defects and mechanical defects do not exist independently. Mechanical loosening leads to a reduction in the effective contact area, resulting in a sharp increase in contact resistance and triggering local overheating. Conversely, the thermal expansion and metal oxidation caused by local high temperatures further exacerbate mechanical loosening. This electro-thermal-mechanical coupling effect means that when vibration and heat generation occur simultaneously, the combined risk is not a simple linear addition, but rather a nonlinear multiplication. Using a simple weighted summation would severely underestimate the danger of this coupled fault. Furthermore, the shape of the thermal field distribution, i.e., the spatial temperature gradient, is also an important basis for judging the nature of the fault; point-like concentrated overheating is usually more destructive than area-like overall temperature rise. Therefore, this invention constructs a nonlinear fusion model, using a hyperbolic tangent function to simulate the modulation effect of mechanical loosening on thermal risk. When the mechanical state is stable, the risk mainly depends on heat; when mechanical loosening intensifies, it multiplies the weight of thermal risk, thereby accurately reflecting the actual system crisis level under multi-physics coupling and avoiding the one-sidedness of a single indicator.

[0065] Preferably, by integrating electrothermal effects, mechanical stability, and thermal field distribution, and based on the current thermal intensity index and vibration dispersion factor, a composite risk index is calculated, including:

[0066] The composite risk index satisfies the following expression:

[0067] ;

[0068] In the formula, Indicates a composite risk index; The electric current thermal intensity index; The vibration dispersion factor; For space temperature gradient; It is the third smallest positive number; It is the hyperbolic tangent function.

[0069] In the formula, A physical model of vibration-modulated thermal resistance was constructed, which is used when the mechanical connection is tightened. When the value is extremely small, this factor approaches 1, and the risk index mainly depends on the thermal strength index; when there is mechanical loosening, that is... When it is very large, Approaching 1, this term approaches 2, which means that the thermal strength index is multiplied and amplified, reflecting the mutual feedback mechanism that loosening leads to the deterioration of contact resistance; The influence of heat distribution morphology was introduced, and the spatial temperature gradient of localized point overheating, i.e. typical poor contact, was much greater than that of overall overheating. This further enhanced the ability to identify local contact faults. It achieves deep nonlinear fusion of electrical, thermal, and mechanical data, and ultimately outputs a composite index that can comprehensively reflect the health risks of equipment.

[0070] For example, using the aforementioned data, , If a local high gradient is detected ,but The system successfully transformed multidimensional features into a single calculated risk indicator.

[0071] It should be noted that, Figure 2 A multi-level coupling transmission model diagram of vibration to the composite risk index is presented, illustrating the transmission relationship and numerical changes between micro-amplitude, vibration dispersion factor, and the composite risk index. When the input micro-amplitude is 0.8 mm, the corresponding vibration dispersion factor is 13.35, and the final output composite risk index is 0.0073. This transmission process demonstrates that the model can reflect the correlation changes from micro-amplitude to the composite risk index. Through the transformation of intermediate factors, it presents the intrinsic relationship between relevant data, providing a reference for equipment risk assessment.

[0072] Thus, the composite risk index of the target equipment was obtained.

[0073] S3: Based on the magnitude of the composite risk index, the health status of the target equipment is judged; the system summarizes the above data and generates a diagnostic report for abnormal equipment; and triggers an alarm push mechanism.

[0074] It's important to note that equipment in substations has varying service lives, and there are significant differences in inherent characteristics between new and old equipment. As operating time increases, mechanical gaps and oxide layers in older equipment naturally lead to higher baseline vibration and heat generation levels – a normal aging process. Using a uniform fixed threshold across the entire substation would have drawbacks: older equipment might experience frequent false alarms due to slightly higher baseline indicators, leading to complacency among maintenance personnel; while newer equipment might have weak initial fault signals below this lenient fixed threshold, resulting in missed alarms. From a lifecycle management perspective, the criteria should be dynamic, based on the equipment's historical stability, i.e., historical vibration tolerance, to set the red line. For equipment with historically unstable performance, the system should automatically relax the tolerance to filter out background noise; for consistently precise equipment, the threshold should be tightened to sensitively detect minor anomalies.

[0075] Specifically, the health status of the target equipment is assessed based on the magnitude of the composite risk index, including:

[0076] The preset first benchmark sensitivity coefficient is multiplied by the natural exponential function value of the historical vibration standard deviation and recorded as the first dynamic threshold. The preset second benchmark sensitivity coefficient is multiplied by the natural exponential function value of the historical vibration standard deviation and recorded as the second dynamic threshold. The first benchmark sensitivity coefficient is less than the second benchmark sensitivity coefficient. If the composite risk index is less than or equal to the first dynamic threshold, the target equipment is determined to be normal. If the composite risk index is greater than the first dynamic threshold and less than the second dynamic threshold, the target equipment is determined to be under attention. If the composite risk index is greater than or equal to the second dynamic threshold, the target equipment is determined to be abnormal and enters the subsequent fault handling process.

[0077] It should be noted that, Figure 3 The diagram illustrates the logic of dynamic threshold determination based on historical vibration standard deviation, showing the correspondence between historical vibration standard deviation and the composite risk index, as well as the health status determination results. For equipment with a historical vibration standard deviation of 0.1, the calculated composite risk index is 0.0073. This value exceeds the first-level dynamic threshold but does not reach the second-level dynamic threshold, resulting in a "Caution" status. This result demonstrates that this dynamic threshold determination method can combine the equipment's historical operational stability to set appropriate judgment criteria for the equipment, helping to differentiate the equipment's health status level.

[0078] At this point, the health status of the equipment has been assessed.

[0079] It's important to note that the ultimate goal of intelligent diagnostic systems is to assist human decision-making, not to completely replace it. When the system detects an anomaly, simply issuing an alarm signal is insufficient. Maintenance personnel need to know where the problem is, its severity, and the basis for the diagnosis. In emergency repair scenarios, time is of the essence for power grid safety. A structured and visualized diagnostic report can significantly shorten troubleshooting time. By transforming abstract algorithmic outputs, such as composite risk indices, into perceptible data from the physical world, such as fault coordinates, original temperature, and vibration amplitude, the interpretability of the algorithm is achieved. This not only provides precise guidance for on-site maintenance but also offers detailed data support for subsequent fault review and responsibility determination. From a human-computer interaction perspective, this step is crucial in transforming cold, hard data into actionable maintenance strategies, ensuring efficient information flow between machines and humans.

[0080] Preferably, the system summarizes the above data and generates a diagnostic report for the malfunctioning device, including:

[0081] If the target power equipment is in an abnormal health state, then the fault location and composite risk index will be summarized to generate a diagnostic report for this abnormal equipment.

[0082] At this point, a diagnostic report for this malfunctioning device was obtained.

[0083] It is important to note that power system faults are often sudden and spread rapidly, especially overheating and loosening of high-voltage equipment. Once the critical point is crossed, they can trigger arcing, short circuits, or even explosions within a very short time. Therefore, the real-time response capability of the diagnostic system is crucial. Traditional periodic inspection and reporting models have significant time lags and cannot meet the needs of immediate loss mitigation. By establishing a multi-channel alarm push mechanism, a shift from post-analysis to real-time early warning has been achieved. On-site audible and visual alarms can immediately alert inspection personnel to stay away from dangerous areas, while push notifications from mobile terminals break spatial limitations, ensuring that remote on-duty personnel can also simultaneously perceive the on-site situation. From the perspective of closed-loop risk management, this step completes the final step of information transmission, ensuring that diagnostic results can be quickly translated into physical intervention actions, minimizing equipment damage rates and the risk of power grid outages.

[0084] Preferably, the alarm push mechanism includes:

[0085] The system triggers the audible and visual alarm through the substation integrated automation interface, and the alarm's push mechanism pushes alarm information to the handheld terminal APP of the operation and maintenance personnel in real time.

[0086] At this point, the system has implemented alarms for abnormal devices.

[0087] S4: Store the collected data and update the statistical model.

[0088] It should be noted that statistical diagnostic models are highly dependent on the quality of the sample library. With seasonal changes, adjustments to power grid operation, and the aging of equipment, the normal baseline of the equipment gradually shifts. If the model remains unchanged, its accuracy will gradually decrease over time. Therefore, the system must possess lifelong learning capabilities, continuously incorporating new operational data to correct its statistical parameters. However, blindly updating the entire sample library is dangerous. Including abnormal data from fault states in the normal sample library can contaminate the model, causing thresholds to be incorrectly raised, leading to subsequent false negatives. Therefore, this invention introduces a supervised update strategy, updating the model only when a condition is determined to be normal.

[0089] Specifically, the collected data is stored, and the statistical model is updated, including:

[0090] If the health status of the target equipment is determined to be normal, the historical vibration mean and historical vibration standard deviation of the equipment are recalculated and updated using new data to achieve dynamic updating and optimization of the diagnostic model; if it is determined to be in an abnormal state, no update is performed.

[0091] This completes the fault diagnosis of the power plant's electrical equipment.

[0092] This invention also discloses a substation power equipment fault diagnosis system, including a processor and a memory. The memory stores computer program instructions, which, when executed by the processor, implement a substation power equipment fault diagnosis method according to the present invention.

[0093] The system also includes other components well known to those skilled in the art, such as communication buses and communication interfaces, the settings and functions of which are known in the art and will not be described in detail here.

[0094] While this specification has shown and described numerous embodiments of the invention, it will be apparent to those skilled in the art that such embodiments are provided by way of example only. Many modifications, alterations, and alternatives will occur to those skilled in the art without departing from the spirit and essence of the invention. It should be understood that various alternatives to the embodiments of the invention described herein may be employed in the practice of this invention.

Claims

1. A method for diagnosing faults in substation power equipment, characterized in that, include: The multispectral inspection robot collects infrared thermal images and high-frame-rate visible light video of the target equipment in real time, and obtains the real-time load current of the line where the equipment is located from the SCADA system; the contact area of ​​the infrared thermal image is located using image recognition algorithm, and the highest temperature of the area is extracted and recorded as the highest temperature of the image; the average temperature value of the non-contact area is calculated and recorded as the ambient temperature; the spatial temperature gradient is obtained based on the ratio of the temperature difference from the center point to the edge point of the contact area to the distance difference; and the rated current of the equipment is read from the database. Based on the principle that the conductor's heating power is proportional to the square of the current, the current thermal intensity index is calculated using the highest temperature in the image, the ambient temperature, and the real-time load current; and the vibration dispersion factor, which characterizes the degree of looseness of the equipment's mechanical connections, is calculated using the video's micro-amplitude and its historical statistical characteristics. By integrating electrical thermal effects, mechanical stability, and thermal field distribution, a composite risk index is calculated based on the current thermal intensity index and vibration dispersion factor. The current thermal index satisfies: ； Indicates the thermal intensity index of electric current; The highest temperature in the image; Ambient temperature; This is the real-time load current; Rated current; It is the natural logarithm function; It is a natural constant; Given the first smallest positive number, the denominator is guaranteed to be non-zero; It characterizes the relative temperature rise induced by the square of the unit current. This item can effectively offset the natural temperature change caused by load fluctuations and separate the fault characteristics from the current thermal effect. It is a nonlinear gain term. Considering that the aging rate and heat accumulation effect of the contact surface will be accelerated nonlinearly under high current conditions, this term is used to appropriately improve the sensitivity to thermal defects under heavy load conditions and prevent missed detection caused by high current masking early minor defects. The composite risk index satisfies the following expression: ； Indicates a composite risk index; The electric current thermal intensity index; The vibration dispersion factor; For space temperature gradient; It is the third smallest positive number; It is the hyperbolic tangent function; The vibration discrete factor satisfies the following expression: A phase-based motion amplification and estimation algorithm is applied to the acquired high frame rate visible light video to calculate its displacement peak value, which is denoted as the video micro amplitude. The vibration discrete factor satisfies: ； Indicates the microscopic amplitude of the video; Indicates the historical average vibration; Indicates historical vibration standard deviation; If it is the second smallest positive number, the denominator must not be 0; Represents an exponential function with the natural constant as its base; Represents the maximum value function; The health status of the target equipment is judged based on the magnitude of the composite risk index; the system summarizes the above data, generates a diagnostic report for abnormal equipment, and triggers an alarm push mechanism; Store the collected data and update the statistical model.

2. The method for fault diagnosis of substation power equipment according to claim 1, characterized in that, The real-time load current of the line where the device is located is obtained from the SCADA system; the contact area of ​​the infrared thermal image is located using an image recognition algorithm, and the highest temperature of the area is extracted and recorded as the highest temperature of the image; the average temperature value of the non-contact area is calculated and recorded as the ambient temperature. Read the rated current of the device from the database, including: Obtain the real-time load current of the line where the target device is located from the SCADA system; Image recognition algorithms are used to perform target detection on the infrared thermal images acquired in real time by the multispectral inspection robot to obtain the contact area; the highest pixel temperature value in this area is extracted and recorded as the highest temperature of the image; the non-contact area in the infrared thermal image is defined as the background area, and the average temperature value of the background area is calculated and recorded as the ambient temperature; the rated current of the device is read from the database.

3. The method for diagnosing faults in substation power equipment according to claim 1, characterized in that, The determination of the health status of the target device includes: The preset first benchmark sensitivity coefficient is multiplied by the natural exponential function value of the historical vibration standard deviation and recorded as the first dynamic threshold. The preset second benchmark sensitivity coefficient is multiplied by the natural exponential function value of the historical vibration standard deviation and recorded as the second dynamic threshold. The first benchmark sensitivity coefficient is less than the second benchmark sensitivity coefficient. If the composite risk index is less than or equal to the first dynamic threshold, the target equipment is determined to be normal. If the composite risk index is greater than the first dynamic threshold and less than the second dynamic threshold, the target equipment is determined to be under attention. If the composite risk index is greater than or equal to the second dynamic threshold, the target equipment is determined to be abnormal and enters the subsequent fault handling process.

4. The method for diagnosing faults in substation power equipment according to claim 1, characterized in that, The diagnostic report for the generated malfunctioning device includes: If the target power equipment is in an abnormal health state, then the fault location and composite risk index will be summarized to generate a diagnostic report for this abnormal equipment.

5. The method for fault diagnosis of power equipment in a substation according to claim 1, characterized in that, The alarm push mechanism includes: The system triggers the audible and visual alarm through the substation integrated automation interface, and the alarm's push mechanism pushes alarm information to the handheld terminal APP of the operation and maintenance personnel in real time.

6. The method for diagnosing faults in substation power equipment according to claim 1, characterized in that, The process of storing the collected data and updating the statistical model includes: If the health status of the target equipment is determined to be normal, the historical vibration mean and historical vibration standard deviation of the equipment are recalculated and updated using new data to achieve dynamic updating and optimization of the diagnostic model; if it is determined to be in an abnormal state, no update is performed.

7. A fault diagnosis system for substation power equipment, characterized in that, include: A processor and a memory, wherein the memory stores computer program instructions that, when executed by the processor, implement a substation power equipment fault diagnosis method according to any one of claims 1-6.

Images