Method and system for detecting hidden fault of relay protection device

Abstract

The invention belongs to the technical field of power system relay protection, and particularly relates to a relay protection device hidden fault detection method and system. The method comprises the following steps: acquiring current detection values of multi-source monitoring parameters of the relay protection device, wherein the types of the multi-source monitoring parameters comprise at least two of a power supply parameter, an environment parameter, a communication performance parameter and a logic unit parameter; the power supply parameters comprise voltage fluctuation and / or ripple coefficients, the environment parameters comprise at least one of temperature, humidity and vibration, the communication performance parameters comprise transmission delay and / or bit error rate, and the logic unit parameters comprise CPU load and / or memory occupation; determining the deviation degree of the relay protection device based on the deviation between the current detection value of the multi-source monitoring parameter and the health baseline of the multi-source monitoring parameter; whether hidden faults exist or not is judged based on the deviation degree of the relay protection device and the hidden fault threshold value of the relay protection device, parameter gradual change type faults, environment coupling type faults and logic soft faults can be recognized, and detection blind areas are avoided.

Description

A method and system for detecting latent faults in relay protection devices Technical Field

[0001] This invention belongs to the field of power system relay protection technology, specifically relating to a method and system for detecting latent faults in relay protection devices. Background Technology

[0002] As an automatic device, relay protection can detect and identify faults and abnormal operating conditions in a power system. It can further issue alarm signals and even regulate system load shedding and tripping strategies, controlling circuit breaker operation to quickly, sensitively, selectively, and reliably isolate faults and ensure the safe operation of other fault-free components. Relay protection, as a crucial part of ensuring the safe and stable operation of a system, has received widespread attention. Latent faults are defects in relay protection devices that are not apparent during normal grid operation but are triggered only during faults. These can lead to malfunctions or failures to operate, and even trigger cascading faults. Traditional relay protection device self-testing systems mainly rely on preset thresholds to determine hardware status, such as judging the normality of the power supply module based on preset voltage thresholds. In other words, traditional relay protection device self-testing systems have a single data dimension, relying solely on internal self-test signals for detection, making them prone to false alarms or missed alarms.

[0003] To improve the accuracy of fault detection, Chinese patent application CN105629097A discloses a method for detecting latent faults in relay protection devices. This method utilizes a WAMS system and a SCADA system to obtain the voltage amplitude of each node in the power grid, and combines this with state estimation to determine a reference value for latent fault detection. Then, a threshold value is determined based on measurement error, calculation error, and the reference value. When the number of times the deviation between the measured value uploaded by the protection information system and the state estimation exceeds the threshold value exceeds a maximum number, the relay protection device is considered to have a latent fault, and an alarm signal is issued. While this method can detect latent faults to a certain extent, it is limited to parameters such as voltage and current, and has a detection blind zone. That is, it cannot identify faults with gradually changing parameters (such as increased power ripple caused by capacitor aging), environmental coupling interference (temperature drift affecting sampling accuracy), and soft faults in the logic module. For example, a substation's relay protection device showed that the power module was normal during self-testing. However, a latent fault caused by aging electrolytic capacitors led to actual power supply voltage fluctuations of ±25%, resulting in abnormal sampling data and ultimately causing malfunctioning protection. Furthermore, this system cannot adapt to dynamic scenarios such as equipment aging or environmental changes, making it difficult to reflect the health status of the protection device throughout its entire lifecycle, leading to false alarms or missed alarms. Moreover, traditional state estimation algorithms have high computational complexity and cannot meet the real-time requirements of high-frequency sampling. Summary of the Invention

[0004] The purpose of this invention is to provide a method and system for detecting latent faults in relay protection devices, so as to solve the problem of detection blind spots in existing methods for detecting latent faults in relay protection devices.

[0005] To address the aforementioned technical problems, this invention provides a method for detecting latent faults in relay protection devices, comprising: acquiring the current detection values ​​of multi-source monitoring parameters of the relay protection device; the types of multi-source monitoring parameters include at least two of power supply parameters, environmental parameters, communication performance parameters, and logic unit parameters; power supply parameters include voltage fluctuation and / or ripple coefficient; environmental parameters include at least one of temperature, humidity, and vibration; communication performance parameters include transmission delay and / or bit error rate; and logic unit parameters include CPU load and / or memory usage; determining the deviation degree of the relay protection device based on the deviation between the current detection values ​​of the multi-source monitoring parameters and the health baseline of the multi-source monitoring parameters; and determining whether a latent fault exists based on the deviation degree of the relay protection device and the latent fault threshold of the relay protection device.

[0006] Furthermore, the deviation of the relay protection device is calculated using the following formula:

[0007] ;

[0008] Where ΔP is the deviation of the relay protection device, n is the total number of types of multi-source monitoring parameters, and w i Let m be the weight of the multi-source monitoring parameters corresponding to the i-th class. i P represents the number of multi-source monitoring parameters of the i-th class. current,i,j Let P be the current detected value of the j-th parameter in the multi-source monitoring parameters of the i-th class. baseline,i,j Let be the health baseline of the j-th parameter among the multi-source monitoring parameters of the i-th class.

[0009] Furthermore, the latent fault threshold of the relay protection device is a dynamically adjusted threshold; the longer the cumulative operating time of the relay protection device, the larger the latent fault threshold of the relay protection device; the greater the environmental fluctuations, the larger the latent fault threshold of the relay protection device.

[0010] Furthermore, the latent fault threshold of the relay protection device is calculated according to the following formula:

[0011] ;

[0012] Where θ(t) is the latent fault threshold at time t, θ0 is the initial latent fault threshold, k is the aging coefficient, t is the cumulative operating time of the relay protection device, and T life For the design life of the relay protection device, σ ΔP Standard deviation of historical deviation, β θ This is the threshold adjustment coefficient.

[0013] Furthermore, the method also includes, if the number of times the deviation of the relay protection device exceeds the latent fault threshold of the relay protection device reaches a first threshold, then the weight of the multi-source monitoring parameter category that contributes the most to the deviation of the relay protection device is increased, and the adjusted weight is used to calculate the deviation of the relay protection device in the next latent fault detection.

[0014] Furthermore, the method also includes automatically updating the health baseline of the multi-source monitoring parameters according to the following formula if the deviation of the relay protection device is less than the latent fault threshold of the relay protection device:

[0015] P baseline-new,i,j =A×P baseline-old,i,j +B×P current,i,j

[0016] Among them, P baseline-new,i,j For the health baseline of the j-th parameter in the updated multi-source monitoring parameters of the i-th class, P baseline-old,i,j For the healthy baseline of the j-th parameter in the i-th class of multi-source monitoring parameters before the update, P current,i,j Let A be the current detection value of the j-th parameter in the multi-source monitoring parameters of the i-th class, where A is the first weight and B is the second weight, and A is greater than B.

[0017] Furthermore, determining whether a latent fault exists includes: if the relay protection device self-test is normal, and the deviation of the relay protection device is greater than one time the latent fault threshold of the relay protection device but less than two times the latent fault threshold of the relay protection device, then a low-level alarm is generated; if the relay protection device self-test is normal, and the deviation of the relay protection device is greater than or equal to two times the latent fault threshold of the relay protection device, then a high-level alarm is generated, and an optimization algorithm is used to locate the fault.

[0018] Furthermore, the optimization algorithm employs the firefly algorithm, where each firefly represents a set of multi-source monitoring parameters. The algorithm uses a brightness function to find the optimal firefly in each iteration, determining whether it meets the strict convergence condition. If it does, the iteration stops, and the fault type is determined based on the multi-source monitoring parameters corresponding to the optimal firefly, with emergency fault handling implemented. If the strict convergence condition is not met, the algorithm checks for a weak convergence condition. If the weak convergence condition is not met, the iteration continues until it is met, at which point the optimal firefly at the point of stopping iteration is output. If the deviation of the relay protection device under the multi-source monitoring parameters corresponding to the optimal firefly is greater than or equal to twice the latent fault threshold of the relay protection device, the fault type is determined based on the multi-source monitoring parameters corresponding to the optimal firefly, and fault handling is implemented. The strict convergence condition is determined based on the principle of ensuring both tight aggregation and high deviation of the firefly swarm, while the weak convergence condition is determined based on the principle of preventing infinite loops.

[0019] Furthermore, the brightness function is:

[0020] ;

[0021] Among them, I i Let be the brightness of the i-th firefly, α be the self-test state difference coefficient, β be the deviation coefficient, and D be the value of the firefly. self,i D* represents the self-test status value of the relay protection device. self,i ΔP is the ideal value for self-testing of the relay protection device. i Let be the deviation of the relay protection device under the multi-source monitoring parameters corresponding to the i-th firefly.

[0022] Furthermore, during the iteration process, if the triggering condition is met, a reinforcement iteration is initiated to expand the firefly population size and increase the attraction coefficient in the attraction function used for firefly location updates. The triggering condition is that the deviation of the relay protection device under the multi-source monitoring parameters corresponding to the optimal firefly is greater than or equal to twice the latent fault threshold of the relay protection device. The attraction function is:

[0023] ;

[0024] Where, β ij (γ) represents the attraction between the i-th firefly and the j-th firefly, β0 is the attraction coefficient, γ is the light absorption coefficient, and ΔP j Let be the deviation of the relay protection device under the multi-source monitoring parameters corresponding to the j-th firefly.

[0025] To address the aforementioned technical problems, this invention also provides a latent fault detection system for relay protection devices, comprising a processor. The processor is used to execute a latent fault detection method for relay protection devices, comprising: acquiring the current detection values ​​of multi-source monitoring parameters of the relay protection device, wherein the types of multi-source monitoring parameters include at least two of power supply parameters, environmental parameters, communication performance parameters, and logic unit parameters; the power supply parameters include voltage fluctuation and / or ripple coefficient; the environmental parameters include at least one of temperature, humidity, and vibration; the communication performance parameters include transmission delay and / or bit error rate; and the logic unit parameters include CPU load and / or memory usage; determining the deviation degree of the relay protection device based on the deviation between the current detection values ​​of the multi-source monitoring parameters and the health baseline of the multi-source monitoring parameters; and determining whether a latent fault exists based on the deviation degree of the relay protection device and a latent fault threshold of the relay protection device.

[0026] Furthermore, the deviation of the relay protection device is calculated using the following formula:

[0027] ;

[0028] Where ΔP is the deviation of the relay protection device, n is the total number of types of multi-source monitoring parameters, and w i Let m be the weight of the multi-source monitoring parameters corresponding to the i-th class. i P represents the number of multi-source monitoring parameters of the i-th class. current,i,j Let P be the current detected value of the j-th parameter in the multi-source monitoring parameters of the i-th class. baseline,i,j Let be the health baseline of the j-th parameter among the multi-source monitoring parameters of the i-th class.

[0029] Furthermore, the latent fault threshold of the relay protection device is a dynamically adjusted threshold; the longer the cumulative operating time of the relay protection device, the larger the latent fault threshold of the relay protection device; the greater the environmental fluctuations, the larger the latent fault threshold of the relay protection device.

[0030] Furthermore, the latent fault threshold of the relay protection device is calculated according to the following formula:

[0031] ;

[0032] Where θ(t) is the latent fault threshold at time t, θ0 is the initial latent fault threshold, k is the aging coefficient, t is the cumulative operating time of the relay protection device, and T life For the design life of the relay protection device, σ ΔP β represents the standard deviation of historical deviation. θ This is the threshold adjustment coefficient.

[0033] Furthermore, the method also includes, if the number of times the deviation of the relay protection device exceeds the latent fault threshold of the relay protection device reaches a first threshold, then the weight of the multi-source monitoring parameter category that contributes the most to the deviation of the relay protection device is increased, and the adjusted weight is used to calculate the deviation of the relay protection device in the next latent fault detection.

[0034] Furthermore, the method also includes automatically updating the health baseline of the multi-source monitoring parameters according to the following formula if the deviation of the relay protection device is less than the latent fault threshold of the relay protection device:

[0035] P baseline-new,i,j =A×P baseline-old,i,j +B×P current,i

[0036] Among them, P baseline-new,i,j For the health baseline of the j-th parameter in the updated multi-source monitoring parameters of the i-th class, P baseline-old,i,j P is the health baseline of the j-th parameter in the i-th class of multi-source monitoring parameters before the update. current,i,j Let A be the current detection value of the j-th parameter in the multi-source monitoring parameters of the i-th class, where A is the first weight and B is the second weight, and A is greater than B.

[0037] Furthermore, determining whether a latent fault exists includes: if the relay protection device self-test is normal, and the deviation of the relay protection device is greater than one time the latent fault threshold of the relay protection device but less than two times the latent fault threshold of the relay protection device, then a low-level alarm is generated; if the relay protection device self-test is normal, and the deviation of the relay protection device is greater than or equal to two times the latent fault threshold of the relay protection device, then a high-level alarm is generated, and an optimization algorithm is used to locate the fault.

[0038] Furthermore, the optimization algorithm employs the firefly algorithm, where each firefly represents a set of multi-source monitoring parameters. The algorithm uses a brightness function to find the optimal firefly in each iteration, determining whether it meets the strict convergence condition. If it does, the iteration stops, and the fault type is determined based on the multi-source monitoring parameters corresponding to the optimal firefly, with emergency fault handling implemented. If the strict convergence condition is not met, the algorithm checks for a weak convergence condition. If the weak convergence condition is not met, the iteration continues until it is met, at which point the optimal firefly at the point of stopping iteration is output. If the deviation of the relay protection device under the multi-source monitoring parameters corresponding to the optimal firefly is greater than or equal to twice the latent fault threshold of the relay protection device, the fault type is determined based on the multi-source monitoring parameters corresponding to the optimal firefly, and fault handling is implemented. The strict convergence condition is determined based on the principle of ensuring both tight aggregation and high deviation of the firefly swarm, while the weak convergence condition is determined based on the principle of preventing infinite loops.

[0039] Furthermore, the brightness function is:

[0040] ;

[0041] Among them, I i Let be the brightness of the i-th firefly, α be the self-test state difference coefficient, β be the deviation coefficient, and D be the value of the firefly. self,i D* represents the self-test status value of the relay protection device. self,i ΔP is the ideal value for self-testing of the relay protection device. i Let be the deviation of the relay protection device under the multi-source monitoring parameters corresponding to the i-th firefly.

[0042] Furthermore, during the iteration process, if the triggering condition is met, a reinforcement iteration is initiated to expand the firefly population size and increase the attraction coefficient in the attraction function used for firefly location updates. The triggering condition is that the deviation of the relay protection device under the multi-source monitoring parameters corresponding to the optimal firefly is greater than or equal to twice the latent fault threshold of the relay protection device. The attraction function is:

[0043] ;

[0044] Where, βij (γ) represents the attraction between the i-th firefly and the j-th firefly, β0 is the attraction coefficient, γ is the light absorption coefficient, and ΔP j Let be the deviation of the relay protection device under the multi-source monitoring parameters corresponding to the j-th firefly.

[0045] The beneficial effects of the above technical solution are as follows: This invention determines the deviation of the relay protection device based on the deviation between the current detection value of the multi-source monitoring parameters and the health baseline of the multi-source monitoring parameters. Based on the deviation of the relay protection device and the latent fault threshold of the relay protection device, it judges whether a latent fault exists. This constructs a multi-dimensional monitoring index system covering the device's power module, logic unit, communication link, and environmental status, breaking through the traditional self-testing method that relies solely on a single data dimension of internal hardware status. It can identify parameter-gradual, environment-coupled, and logic soft faults, avoiding detection blind spots, and is suitable for the high reliability requirements of smart grids and new energy power plants. The method proposed in this invention is entirely based on software algorithms, requiring no special hardware platform for the relay protection device and not relying on dedicated AI acceleration chips or high-performance processors. All feature calculations, model inferences, and correction logic can run efficiently on existing mainstream relay protection CPU / DSP platforms, requiring only reasonable allocation of memory and computing resources. Therefore, this solution can be seamlessly integrated into existing product lines without modifying the hardware architecture, significantly reducing upgrade costs and possessing strong engineering feasibility and promotional value. Attached Figure Description

[0046] Figure 1 is a flowchart of the latent fault detection process of the relay protection device according to the embodiment of the method of the present invention;

[0047] Figure 2 is a flowchart of the latent fault location of the relay protection device according to the embodiment of the method of the present invention;

[0048] Figure 3 is a schematic diagram of the network structure of the relay protection device latent fault detection system according to the system implementation of the present invention. Detailed Implementation

[0049] To make the objectives, technical solutions, and advantages of the present invention clearer, the specific embodiments of the present invention will be further described below with reference to the accompanying drawings.

[0050] This invention identifies whether a relay protection device has a hidden fault by comprehensively determining the deviation between the current detected values ​​of power supply parameters, environmental parameters, communication performance parameters, and logic unit parameters and the health baseline. This improves the accuracy of fault detection.

[0051] Method Implementation

[0052] The present invention provides a method for detecting latent faults in a relay protection device, as shown in Figure 1. The specific process is as follows:

[0053] 1. Determine whether it is necessary to enter the hidden fault detection process of relay protection device.

[0054] State variable definition: D real Defined as the true health state of a relay protection device, where D real =1 indicates an actual anomaly, D real =0 indicates normal. D self Defined as the self-test reporting status of the relay protection device, where D self =0 indicates that the self-test is normal, D self =1 indicates a self-test error.

[0055] If the self-test module alarms, it indicates a self-test anomaly (D). self =1), proceed directly as a visible fault, without entering the hidden fault detection process. If the self-test shows normal (D self If the value is 0, the latent fault detection mechanism is triggered, and the latent fault detection process begins.

[0056] 2. Obtain the current detection values ​​of the multi-source monitoring parameters of the relay protection device.

[0057] The types of multi-source monitoring parameters include at least two of the following: power parameters, environmental parameters, communication performance parameters, and logic unit parameters; power parameters include voltage fluctuation and / or ripple coefficient; environmental parameters include at least one of temperature, humidity, and vibration; communication performance parameters include transmission delay and / or bit error rate (such as CRC error rate); and logic unit parameters include CPU load (such as logic branch coverage) and / or memory usage.

[0058] 3. Determine the deviation of the relay protection device based on the deviation between the current detection value of the multi-source monitoring parameters and the health baseline of the multi-source monitoring parameters.

[0059] The deviation of relay protection devices is used to quantify the difference between the current state and the health baseline. Multi-source monitoring parameters are categorized into several types, each containing one or more multi-source monitoring parameters, with weights w. i The parameters are assigned to the i-th type of multi-source monitoring. Therefore, the deviation of the relay protection device is calculated using the following formula:

[0060]

[0061] Where ΔP is the deviation of the relay protection device, n is the total number of types of multi-source monitoring parameters, and w i Let m be the weight of the multi-source monitoring parameters corresponding to the i-th class. i P represents the number of multi-source monitoring parameters of the i-th class. current,i,j Let P be the current detected value of the j-th parameter in the multi-source monitoring parameters of the i-th class. baseline,i,jLet be the health baseline of the j-th parameter in the i-th category of multi-source monitoring parameters. The initial health baseline of each parameter in each category of multi-source monitoring parameters can be independently established through historical health data, such as the mean of the multi-source monitoring parameters over the past 30 days or the median of the normal fluctuation range over the past 30 days.

[0062] For example, the type of multi-source monitoring parameter is environmental parameter, and the environmental parameter weight w env =0.3, the environmental parameters include two multi-source monitoring parameters: temperature (j=1) and humidity (j=2), i.e., m env =2.

[0063] First, calculate the squares of the relative deviations of temperature and humidity respectively:

[0064]

[0065]

[0066] Secondly, calculate the deviation of the environmental parameter types:

[0067] Environmental parameter deviation = 1 / 2 × (square of relative temperature deviation + square of relative humidity deviation)

[0068] Finally, the deviations in environmental parameters are included in the overall deviation calculation:

[0069]

[0070] w i A dynamic update mechanism is adopted: if the number of times the deviation of the relay protection device exceeds the latent fault threshold reaches a first threshold, the weight of the multi-source monitoring parameter category that contributes the most to the deviation of the relay protection device is increased, and the adjusted weight is used to calculate the deviation of the relay protection device in the next latent fault detection. However, its weight is ultimately limited to 0.5. For example, the weight of this parameter is automatically increased by 20% from its original value, but its weight is ultimately limited to 0.5, i.e.:

[0071]

[0072] in, The increased weight, To increase the weight before.

[0073] If the number of times the deviation of the relay protection device is less than or equal to the latent fault threshold of the relay protection device reaches the second threshold, the weight of the multi-source monitoring parameter category that contributes the most to the deviation of the relay protection device will be reduced, and the adjusted weight will be used to calculate the deviation of the relay protection device in the next latent fault detection. However, its weight will eventually be limited to the initial weight. For example, the weight will be automatically and gradually reduced to the initial value by a factor of 0.9.

[0074] For example, the initial weights for power supply parameters, environmental parameters, communication performance parameters, and logic unit parameters are all 0.3, with a first threshold of 3 and a second threshold of 5. The environmental parameter contributes the most. If ΔP is greater than the latent fault threshold of the relay protection device for three consecutive detection cycles, the weight of the environmental parameter is automatically increased from 0.3 to 0.36. The temporary weight of 0.36 is compared with the upper limit of 0.5, and the smaller value is taken as the new weight, i.e., the new weight is 0.36, enhancing the sensitivity to this parameter. If ΔP is less than or equal to the latent fault threshold of the relay protection device for five consecutive detection cycles, the weight of the environmental parameter decreases from 0.36 to 0.324. When it decreases again to 0.292, because it is less than 0.3, it remains at the initial weight of 0.3.

[0075] If the deviation of the relay protection device is less than the hidden fault threshold of the relay protection device, the health baseline of the multi-source monitoring parameters will be automatically updated according to the following formula, so as to adapt to the natural aging of the equipment and avoid false alarm signals.

[0076] P baseline-new,i,j =A×P baseline-old,i,j +B×P current,i,j

[0077] Among them, P baseline-new,i,j For the health baseline of the j-th parameter in the updated multi-source monitoring parameters of the i-th class, P baseline-old,i,j Let A be the health baseline of the j-th parameter in the i-th class of multi-source monitoring parameters before the update, with A as the first weight and B as the second weight, where A is greater than B.

[0078] For example, if A=0.9 and B=0.1, then for the j-th parameter in the multi-source monitoring parameters of the i-th class, its health baseline update is as follows:

[0079] P baseline-new,i,j = 0.9 P baseline-old,i,j +0.1P current,i,j

[0080] The latent fault threshold of relay protection devices is a dynamically adjusted threshold. Its value follows these principles: considering the aging effect of equipment, the latent fault threshold increases accordingly with the longer the cumulative operating time of the relay protection device. Simultaneously, the latent fault threshold is adaptively adjusted based on the statistical characteristics of historical operating data. This dynamic threshold mechanism effectively adapts to changes in the equipment's state throughout its entire lifecycle, reducing the risk of false alarms and missed alarms. In other words, environmental fluctuations are determined based on the statistical characteristics of historical operating data; the longer the cumulative operating time of the relay protection device, the higher the latent fault threshold; and the greater the environmental fluctuations, the higher the latent fault threshold.

[0081] Preferably, the latent fault threshold of the relay protection device is calculated according to the following formula:

[0082]

[0083] Where θ(t) is the latent fault threshold at time t, θ0 is the initial latent fault threshold (i.e., when the relay protection device starts operating), which can be taken between 0.1 and 0.15; k is the aging coefficient, obtained by fitting historical fault data, usually taken as 0.05; t is the cumulative operating time of the relay protection device, in years; T life The design life of relay protection devices is generally 15 years; σ ΔP β is the standard deviation of historical deviations (e.g., the standard deviation of deviations over the past 30 days), used to reflect the intensity of environmental fluctuations; θ This is the threshold adjustment coefficient.

[0084] 4. Determine whether a latent fault exists based on the deviation of the relay protection device and the latent fault threshold of the relay protection device.

[0085] In one implementation, the following condition is met simultaneously: Self-test normal D self =0 and ΔP > θ(t) (D real When =1), a latent fault alarm is output.

[0086] In another implementation, if the relay protection device self-test is normal, and the deviation of the relay protection device is greater than one time the latent fault threshold of the relay protection device but less than two times the latent fault threshold of the relay protection device, that is: self-test is normal D self When =0 and 2θ(t)>ΔP>θ(t), a low-level latent fault alarm (such as a yellow alarm) is generated, prompting attention to specific parameters (such as a sudden increase in communication bit error rate); if the relay protection device self-test is normal, and the deviation of the relay protection device is greater than or equal to twice the latent fault threshold of the relay protection device, that is: self-test normal D selfWhen t = 0 and ΔP ≥ 2θ(t), a high-level latent fault alarm (such as a red alarm) is generated. At the same time, an optimization algorithm is used to locate the fault and generate a maintenance work order.

[0087] The optimization algorithm can be the firefly algorithm, genetic algorithm, particle swarm optimization algorithm, ant colony optimization algorithm, etc., and there are no restrictions here. The following detailed explanation uses the Firefly Algorithm as an example. As shown in Figure 2, each firefly in the Firefly Algorithm represents a set of multi-source monitoring parameters. The algorithm uses a brightness function to find the optimal firefly in each iteration. It then checks if the optimal firefly meets the strict convergence condition. If it does, the iteration stops, and the fault type is determined based on the multi-source monitoring parameters corresponding to the optimal firefly, and emergency fault handling is performed. If the strict convergence condition is not met, it checks if the weak convergence condition is met. If the weak convergence condition is not met, the iteration continues until it meets the weak convergence condition, at which point the optimal firefly at the point of stopping iteration is output. If the deviation of the relay protection device under the multi-source monitoring parameters corresponding to the optimal firefly is greater than or equal to twice the implicit fault threshold of the relay protection device, the fault type is determined based on the multi-source monitoring parameters corresponding to the optimal firefly, and fault handling is performed. The strict convergence condition is determined based on the principle of ensuring that the firefly swarm is both tightly clustered (similar positions) and achieves high deviation (significant fault characteristics). The weak convergence condition is determined based on the principle of preventing infinite loops, which forces the output of the current optimal solution.

[0088] The fault output includes the fault type and alarm type. Based on the optimal combination of multi-source monitoring parameters corresponding to the firefly, the fault type is determined by matching it with a predefined library of typical fault characteristics. For example, the fault characteristic library may contain criteria for typical faults (but is not limited to):

[0089] Capacitor aging: This is typically characterized by a significant increase in power supply ripple (e.g., >5%) and may be accompanied by an increase in operating temperature (e.g., temperature gradient >8℃). If the combination of high deviation parameters located by the Firefly algorithm closely matches these characteristics, it is identified as this fault.

[0090] Communication anomalies are typically characterized by an increased bit error rate (e.g., >1E-6) and a larger synchronization signal deviation (e.g., >10μs). The synchronization signal deviation refers to the difference between the device's local clock and the standard time. If the combination of high deviation parameters identified by the Firefly algorithm closely matches these characteristics, it is identified as this fault.

[0091] A dynamic search strategy is adopted, dividing the monitoring scope into two categories: key monitoring areas and routine monitoring areas. Key monitoring areas refer to operating conditions where parameters have deviated significantly and there is a risk of hidden faults; routine monitoring areas refer to operating conditions where parameters are within healthy ranges.

[0092] If the relay protection device self-test display shows normal (D) self If the value is 0 and the actual parameter deviates from the healthy value (ΔP>θ(t)), then the system enters the key monitoring zone. Within this zone, the system increases the monitoring frequency and initiates preliminary fault analysis.

[0093] If the deviation of the relay protection device further increases (ΔP≥2θ(t)) in the key monitoring area, a large amount of computing resources will be concentrated to start the Firefly optimization algorithm for deep search in order to quickly locate the fault characteristics.

[0094] If the conditions for entering the key monitoring area are not met, routine monitoring will be carried out, and the detection frequency will be reduced to save computing power.

[0095] Finally, depending on whether fault analysis is enabled and which algorithm is used, different levels of risk alarms and fault location results are output.

[0096] Specifically, the firefly algorithm optimization process consists of five steps: initializing the firefly swarm, calculating brightness, attracting and moving fireflies, updating positions, and determining convergence.

[0097] Initialize the firefly swarm: Randomly generate N fireflies, each firefly's location corresponding to a set of parameter combinations (such as temperature and power ripple) to scan the status of the relay protection device. Simultaneously, set the search space according to the specifications of the relay protection device, for example:

[0098] x i =[0℃, 100℃] × [0%, 5%] (Temperature × Power supply ripple)

[0099] Calculate the brightness: Calculate the brightness of each firefly using the brightness function, which is:

[0100]

[0101] Among them, I i Let be the brightness of the i-th firefly, α be the self-test state difference coefficient (which can be 0.4), β be the deviation coefficient (which can be 0.6), and D be the brightness of the firefly. self,i D* represents the self-test status value of the relay protection device. self,i ΔP is the ideal value for self-testing of the relay protection device. i Let I represent the deviation of the relay protection device under the multi-source monitoring parameters corresponding to the i-th firefly (i.e., the deviation of the relay protection device calculated by the combination of multi-source monitoring parameters represented by the i-th firefly). This value characterizes the degree of device abnormality caused by nonlinear effects such as component aging or noise interference. The brightness value, combining self-test status differences and parameter deviation, can comprehensively assess the risk of latent faults. Under the definition of the brightness function, the lower the brightness, the higher the risk of latent faults. Based on this logic, the following criterion is set: Brightness value I...i =[0.7, 1] indicates a healthy state; a green work order is generated, parameters meet the standards, and no further processing is required; brightness value I i =[0.3, 0.7), indicating a potential latent anomaly, generating a yellow work order requiring manual investigation; brightness value I i =[0, 0.3), indicating a high-risk anomaly, generates a red work order, requiring urgent handling.

[0102] Attraction and Movement: Low-brightness fireflies move towards high-brightness individuals; the attraction function is:

[0103]

[0104] Where, β ij (γ) represents the attraction between the i-th and j-th fireflies, β0 is the attraction coefficient (also the maximum attraction value, usually taken as 0.8~1.2), and γ is the light absorption coefficient. The larger γ is, the more significant the weakening effect of the difference on the attraction. It is recommended that γ ≥ (0.4, 0.5); ΔP j Let be the deviation of the relay protection device under the multi-source monitoring parameters corresponding to the j-th firefly.

[0105] Update position: If the new position exceeds the parameter range, then the boundary value is used (e.g., if the temperature is >100℃, it is forcibly set to 100℃). The position update formula is:

[0106]

[0107] The larger ΔP is, the stronger the random perturbation (enhancing local search capabilities).

[0108] Convergence criterion: The termination condition for algorithm iteration is determined by a dual criterion: there are two conditions, namely a strict convergence condition and a weak convergence condition.

[0109] Strict convergence requires both of the following conditions to be met:

[0110] First, the firefly population is highly clustered, meaning the normalized maximum population distance is less than a threshold:

[0111]

[0112] Where, x max and x min ε is the preset boundary vector for each parameter dimension, and ε is the normalized convergence threshold, with a value range of (0.01, 0.0, 2).

[0113] Secondly, the deviation of the optimal firefly is large enough:

[0114]

[0115] The purpose of strict convergence conditions is to ensure that the firefly population is both tightly clustered (positionally similar) and exhibits high deviation (significant fault characteristics). For example, after several iterations, the firefly population clusters within a small range of the parameter space, and its normalized maximum distance max‖(x) i -x j ) / (x max -x min The value is less than the threshold ε = 0.01, and the optimal firefly parameters are (temperature = 65℃, ripple = 3.2%), corresponding to a deviation ΔP = 32%. Assuming the current dynamic threshold θ(t) = 10%, then ΔP = 32% ≥ 3θ(t) = 30%. Since both the positional convergence and high deviation conditions are satisfied simultaneously, strict convergence is triggered.

[0116] For weak convergence, the number of iterations Q must satisfy the following formula:

[0117] Q≥T max

[0118] T max The maximum number of iterations, for example, T. max =100. The purpose of the weak convergence condition is to prevent infinite loops and force the output of the current optimal solution.

[0119] During the iteration process, i.e., before convergence, if the triggering condition is met, a reinforcement iteration is initiated, expanding the firefly population size (e.g., by 20%) and increasing the attraction coefficient in the attraction function used for firefly position updates (e.g., increasing β0 from 0.8 to 1.2), accelerating the gathering of fireflies towards high ΔP regions. The logical meaning of this action is to proactively enhance search density and speed when approaching faulty regions. For example, ΔP... best =25% (θ(t)=10%), the trigger condition is met, and reinforcement iteration is initiated. If the trigger condition is not met, the parameters are maintained to keep the population size and attraction coefficient unchanged, and the standard iteration is maintained to continue the regular optimization process, thereby maintaining computational efficiency in the low-risk area and avoiding resource waste. The trigger condition is that the deviation of the relay protection device under the multi-source monitoring parameters corresponding to the optimal firefly is greater than or equal to twice the latent fault threshold of the relay protection device.

[0120] For example: A relay protection device experiences intermittent communication anomalies (ΔP=22%, θ(t)=10%). The Firefly Algorithm's operation is as follows: Strict convergence condition judgment: ΔP=22%<3θ(t)=30%, not meeting the strict convergence condition. Weak convergence condition judgment: iteration count=75<100, not meeting the weak convergence condition, proceeding to trigger condition judgment. The trigger condition is ΔP=22%≥2θ=20%, which is met, initiating reinforcement iteration (expanding the population, increasing the attraction coefficient). After 10 reinforcement iterations, ΔP increases to 28%, still below 3θ, returning to the second step to recalculate brightness. A second convergence judgment (iteration count=85): ΔP=28% < 3θ, still not meeting the strict convergence condition, continuing the iteration loop. Finally, iteration count=100, meeting the weak convergence condition, outputting the current optimal solution. The optimal solution corresponds to ΔP=28%, indicating a communication anomaly. A red work order is generated, and a suspected latent fault: communication anomaly is output.

[0121] System Implementation

[0122] This invention can be implemented on an embedded system of a standard relay protection device. The entire process requires no additional hardware support, fully utilizes existing resources, and facilitates large-scale deployment and application. This invention provides a latent fault detection system for a relay protection device, including a processor for executing a latent fault detection method for a relay protection device.

[0123] Specifically, as shown in Figure 3, the system architecture consists of five layers: a multi-source data acquisition layer, a health baseline modeling layer, a parameter deviation calculation layer, an improved firefly optimization layer, and a fault location output layer. The multi-source data acquisition layer comprises a hardware layer, a software layer, and a communication layer. The hardware layer primarily acquires temperature, voltage, and vibration indicators. The software layer primarily acquires indicators such as memory usage and logical branch coverage. The communication layer primarily acquires indicators such as message latency and CRC error rate. The data flow from the multi-source data acquisition layer to the health baseline modeling layer mainly acquires real-time data such as temperature, voltage, and communication status of the relay protection device. The logical relationship is that the raw data is input into the baseline modeling unit to establish a reference model for normal equipment operation (such as the parameter range under healthy conditions). The forward input data flow from the health baseline modeling layer to the parameter deviation calculation layer mainly transmits health baseline parameters (such as standard voltage values ​​and temperature threshold ranges). The parameter deviation calculation layer internally performs dynamic threshold calculation and updates.

[0124] The positive input data flow from the parameter deviation calculation layer to the improved Firefly optimization layer refers to the deviation calculation result (e.g., ΔP=15%) being input into the Firefly optimizer. This optimizer uses the Firefly algorithm to optimize the search space and identify combination patterns of parameters with high deviation. The parameter deviation calculation layer automatically adjusts the weights of parameter types based on the comparison between the deviation and the threshold (e.g., increasing the weight of the temperature parameter from 0.3 to 0.5). This weight adjustment mechanism is an internal function of the parameter deviation calculation layer and does not require feedback from the Firefly optimization layer. Simultaneously, this layer is also responsible for calculating the deviation of each specific monitoring parameter relative to its health baseline and aggregating it according to a hierarchical formula to obtain the overall deviation. The output of the improved Firefly optimization layer to the fault location output layer includes the fault type (e.g., capacitor aging), and the logical relationship expressed is to transform the optimization results into executable location commands.

[0125] Specifically, the process of the relay protection device latent fault detection method executed by the processor includes:

[0126] 1. Determine whether it is necessary to enter the hidden fault detection process of relay protection device.

[0127] State variable definition: D real Defined as the true health state of a relay protection device, where D real =1 indicates an actual anomaly, D real =0 indicates normal. D self Defined as the self-test reporting status of the relay protection device, where D self =0 indicates that the self-test is normal, D self =1 indicates a self-test error.

[0128] If the self-test module alarms, it indicates a self-test anomaly (D). self =1), proceed directly as a visible fault, without entering the hidden fault detection process. If the self-test shows normal (D self If the value is 0, the latent fault detection mechanism is triggered, and the latent fault detection process begins.

[0129] 2. Obtain the current detection values ​​of the multi-source monitoring parameters of the relay protection device.

[0130] The types of multi-source monitoring parameters include at least two of the following: power parameters, environmental parameters, communication performance parameters, and logic unit parameters; power parameters include voltage fluctuation and / or ripple coefficient; environmental parameters include at least one of temperature, humidity, and vibration; communication performance parameters include transmission delay and / or bit error rate (such as CRC error rate); and logic unit parameters include CPU load (such as logic branch coverage) and / or memory usage.

[0131] 3. Determine the deviation of the relay protection device based on the deviation between the current detection value of the multi-source monitoring parameters and the health baseline of the multi-source monitoring parameters.

[0132] The deviation of the relay protection device is used to quantify the difference between the current state and the healthy baseline, and is calculated according to the following formula:

[0133]

[0134] Where ΔP is the deviation of the relay protection device, n is the total number of types of multi-source monitoring parameters, and w i Let m be the weight of the multi-source monitoring parameters corresponding to the i-th class. i P represents the number of multi-source monitoring parameters of the i-th class. current,i,j Let P be the current detected value of the j-th parameter in the multi-source monitoring parameters of the i-th class. baseline,i,j Let be the health baseline of the j-th parameter among the multi-source monitoring parameters of the i-th category. The initial health baseline of each parameter in the multi-source monitoring parameters of each category can be independently established through historical health data, such as the mean of the multi-source monitoring parameters over the past 30 days or the median of the normal fluctuation range over the past 30 days.

[0135] w i A dynamic update mechanism is adopted: if the number of times the deviation of the relay protection device exceeds the latent fault threshold reaches a first threshold, the weight of the multi-source monitoring parameter category that contributes the most to the deviation of the relay protection device is increased, and the adjusted weight is used to calculate the deviation of the relay protection device in the next latent fault detection. However, its weight is ultimately limited to 0.5. For example, the weight of this parameter is automatically increased by 20% from its original value, but its weight is ultimately limited to 0.5, i.e.:

[0136]

[0137] in, The increased weight, To increase the weight before.

[0138] If the number of times the deviation of the relay protection device is less than or equal to the latent fault threshold of the relay protection device reaches the second threshold, the weight of the multi-source monitoring parameter category that contributes the most to the deviation of the relay protection device will be reduced, and the adjusted weight will be used to calculate the deviation of the relay protection device in the next latent fault detection. However, its weight will eventually be limited to the initial weight. For example, the weight will be automatically and gradually reduced to the initial value by a factor of 0.9.

[0139] If the deviation of the relay protection device is less than the hidden fault threshold of the relay protection device, the health baseline of the multi-source monitoring parameters will be automatically updated according to the following formula, so as to adapt to the natural aging of the equipment and avoid false alarm signals.

[0140] P baseline-new,i,j =A×P baseline-old,i,j+B×P current,i,j

[0141] Among them, P baseline-new,i,j For the health baseline of the j-th parameter in the updated multi-source monitoring parameters of the i-th class, P baseline-old,i,j Let A be the health baseline of the j-th parameter in the i-th class of multi-source monitoring parameters before the update, with A as the first weight and B as the second weight, where A is greater than B.

[0142] For example, A=0.9, B=0.1, that is:

[0143] P baseline-new,i,j = 0.9 P baseline-old,i,j +0.1P current,i,j

[0144] The latent fault threshold of the relay protection device is a dynamically adjusted threshold. Its value follows these principles: considering the aging effect of the equipment, the latent fault threshold increases accordingly with the longer the cumulative operating time of the relay protection device. Simultaneously, the latent fault threshold is also adaptively adjusted based on the statistical characteristics of historical operating data. This dynamic threshold mechanism effectively adapts to changes in the equipment's state throughout its entire lifecycle, reducing the risk of false alarms and missed alarms. That is, environmental fluctuations are determined based on the statistical characteristics of historical operating data; the longer the cumulative operating time of the relay protection device, the larger the latent fault threshold; and the greater the environmental fluctuations, the larger the latent fault threshold. Preferably, the latent fault threshold of the relay protection device is calculated using the following formula:

[0145]

[0146] Where θ(t) is the latent fault threshold at time t, θ0 is the initial latent fault threshold (i.e., when the relay protection device starts operating), which can be taken between 0.1 and 0.15; k is the aging coefficient, obtained by fitting historical fault data, usually taken as 0.05; t is the cumulative operating time of the relay protection device, in years; T life The design life of relay protection devices is generally 15 years; σ ΔP β is the standard deviation of historical deviations (e.g., the standard deviation of deviations over the past 30 days), used to reflect the intensity of environmental fluctuations; θ This is the threshold adjustment coefficient.

[0147] 4. Determine whether a latent fault exists based on the deviation of the relay protection device and the latent fault threshold of the relay protection device.

[0148] In one implementation, the following condition is met simultaneously: Self-test normal D self =0 and ΔP > θ(t) (D real When =1), a latent fault alarm is output.

[0149] In another implementation, if the relay protection device self-test is normal, and the deviation of the relay protection device is greater than one time the latent fault threshold of the relay protection device but less than two times the latent fault threshold of the relay protection device, that is: self-test is normal D self When =0 and 2θ(t)>ΔP>θ(t), a low-level latent fault alarm (such as a yellow alarm) is generated, prompting attention to specific parameters (such as a sudden increase in communication bit error rate); if the relay protection device self-test is normal, and the deviation of the relay protection device is greater than or equal to twice the latent fault threshold of the relay protection device, that is: self-test normal D self When t = 0 and ΔP ≥ 2θ(t), a high-level latent fault alarm (such as a red alarm) is generated. Simultaneously, an optimization algorithm is used to locate the fault and generate a maintenance work order. The optimization algorithm can be any of the following: firefly algorithm, genetic algorithm, particle swarm optimization algorithm, or ant colony optimization algorithm; no specific restrictions are imposed here.

[0150] The following detailed explanation uses the Firefly Algorithm as an example. Each firefly in the Firefly Algorithm represents a set of multi-source monitoring parameters. The algorithm uses a brightness function to find the optimal firefly in each iteration. It checks if the optimal firefly meets the strict convergence condition. If it does, the iteration stops, and the fault type is determined based on the multi-source monitoring parameters corresponding to the optimal firefly, and emergency fault handling is performed. If the strict convergence condition is not met, it checks if the weak convergence condition is met. If the weak convergence condition is not met, the iteration continues. If the weak convergence condition is met, the iteration stops, and the optimal firefly at the point of stopping is output. If the deviation of the relay protection device under the multi-source monitoring parameters corresponding to the optimal firefly is greater than or equal to twice the implicit fault threshold of the relay protection device, the fault type is determined based on the multi-source monitoring parameters corresponding to the optimal firefly, and fault handling is performed. The strict convergence condition is determined based on the principle of ensuring that the firefly swarm is both tightly clustered and achieves a high deviation. The weak convergence condition is determined based on the principle of preventing infinite loops. The weak convergence condition can prevent infinite loops and force the output of the current optimal solution.

[0151] The fault output includes the fault type and alarm type. Based on the optimal combination of multi-source monitoring parameters corresponding to the firefly, the fault type is determined by matching it with a predefined library of typical fault characteristics. For example, the fault characteristic library may contain criteria for typical faults (but is not limited to):

[0152] Capacitor aging: This is typically characterized by a significant increase in power supply ripple (e.g., >5%) and may be accompanied by an increase in operating temperature (e.g., temperature gradient >8℃). If the combination of high deviation parameters located by the Firefly algorithm closely matches these characteristics, it is identified as this fault.

[0153] Communication anomalies are typically characterized by an increased bit error rate (e.g., >1E-6) and a larger synchronization signal deviation (e.g., >10μs). Synchronization signal deviation refers to the difference between the device's local clock and standard time. If the combination of high-deviation parameters identified by the Firefly algorithm closely matches these characteristics, it is identified as this fault. The system employs a dynamic search strategy, dividing the monitoring range into two categories: key monitoring areas and routine monitoring areas. Key monitoring areas refer to operating states where parameters have shown significant deviations and there is a risk of hidden faults, while routine monitoring areas refer to normal operating states where parameters are within healthy ranges.

[0154] If the relay protection device shows normal self-test display (Dself=0) and the actual parameters deviate from the healthy value (ΔP>θ(t)) at the same time, it enters the key monitoring area. In this area, the system increases the monitoring frequency and initiates preliminary fault analysis.

[0155] If the deviation of the relay protection device further increases (ΔP≥2θ(t)) in the key monitoring area, a large amount of computing resources will be concentrated to start the Firefly optimization algorithm for deep search in order to quickly locate the fault characteristics.

[0156] If the conditions for entering the key monitoring area are not met, routine monitoring will be carried out, and the detection frequency will be reduced to save computing power.

[0157] Finally, depending on whether fault analysis is enabled and which algorithm is used, different levels of risk alarms and fault location results are output.

[0158] Specifically, the firefly algorithm optimization process consists of five steps: initializing the firefly swarm, calculating brightness, attracting and moving fireflies, updating positions, and determining convergence.

[0159] Initialize the firefly swarm: Randomly generate N fireflies, each firefly's location corresponding to a set of parameter combinations (such as temperature and power ripple) to scan the status of the relay protection device. Simultaneously, set the search space according to the specifications of the relay protection device, for example:

[0160] x i =[0℃, 100℃] × [0%, 5%] (Temperature × Power supply ripple)

[0161] Calculate the brightness: Calculate the brightness of each firefly using the brightness function, which is:

[0162]

[0163] Among them, I i Let be the brightness of the i-th firefly, α be the self-test state difference coefficient (which can be 0.4), β be the deviation coefficient (which can be 0.6), and D be the brightness of the firefly. self,i D* represents the self-test status value of the relay protection device.self,i ΔP is the ideal value for self-testing of the relay protection device. i This represents the deviation of the relay protection device under the multi-source monitoring parameters corresponding to the i-th firefly (i.e., the deviation of the relay protection device calculated by the combination of multi-source monitoring parameters represented by the i-th firefly). This value characterizes the degree of device abnormality caused by nonlinear effects such as component aging or noise interference. The brightness value, combining self-test status differences and parameter deviation, can comprehensively assess the risk of latent faults; lower brightness indicates a higher risk of latent faults. Based on this logic, the following criteria are set: For example, brightness value I... i =[0.7, 1] indicates a healthy state; a green work order is generated, parameters meet the standards, and no further processing is required; brightness value I i =[0.3, 0.7), indicating a potential latent anomaly, generating a yellow work order requiring manual investigation; brightness value I i =[0, 0.3), indicating a high-risk anomaly, generates a red work order, requiring urgent handling.

[0164] Attraction and Movement: Low-brightness fireflies move towards high-brightness individuals; the attraction function is:

[0165]

[0166] Where, β ij (γ) represents the attraction between the i-th and j-th fireflies, β0 is the attraction coefficient (also the maximum attraction value, usually taken as 0.8~1.2), and γ is the light absorption coefficient. The larger γ is, the more significant the weakening effect of the difference on the attraction. It is recommended that γ ≥ (0.4, 0.5); ΔP j Let be the deviation of the relay protection device under the multi-source monitoring parameters corresponding to the j-th firefly.

[0167] Update position: If the new position exceeds the parameter range, then the boundary value is used (e.g., if the temperature is >100℃, it is forcibly set to 100℃). The position update formula is:

[0168]

[0169] The larger ΔP is, the stronger the random perturbation (enhancing local search capabilities).

[0170] Convergence criterion: The termination condition for algorithm iteration is determined by a dual criterion: there are two conditions, namely a strict convergence condition and a weak convergence condition.

[0171] Strict convergence requires both of the following conditions to be met:

[0172] First, the firefly population is highly clustered, meaning the normalized maximum population distance is less than a threshold:

[0173]

[0174] Where, x max and x min ε is the preset boundary vector for each parameter dimension, and ε is the normalized convergence threshold, with a value range of (0.01, 0.0, 2).

[0175] Secondly, the deviation of the optimal firefly is large enough:

[0176]

[0177] The purpose of strict convergence conditions is to ensure that the firefly population is both tightly clustered (positionally similar) and exhibits high deviation (significant fault characteristics). For example, after several iterations, the firefly population clusters within a small range of the parameter space, and its normalized maximum distance max‖(x) i -x j ) / (x max -x min The value is less than the threshold ε = 0.01, and the optimal firefly parameters are (temperature = 65℃, ripple = 3.2%), corresponding to a deviation ΔP = 32%. Assuming the current dynamic threshold θ(t) = 10%, then ΔP = 32% ≥ 3θ(t) = 30%. Since both the positional convergence and high deviation conditions are satisfied simultaneously, strict convergence is triggered.

[0178] For weak convergence, the number of iterations Q must satisfy the following formula:

[0179] Q≥Q max

[0180] Q max This represents the maximum number of iterations. The purpose of the weak convergence condition is to prevent infinite loops and force the output of the current optimal solution.

[0181] During the iteration process, i.e., before convergence, if the triggering condition is met, an enhanced iteration is initiated to expand the firefly population size and increase the attraction coefficient in the attraction function used for firefly position updates (e.g., β0 is increased from 0.8 to 1.2), accelerating the gathering of fireflies towards high ΔP regions. The logical meaning of this action is to proactively increase the search density and speed when approaching the fault region. For example, if ΔPbest = 25% (θ(t) = 10%), the triggering condition is met, and an enhanced iteration is initiated. If the triggering condition is not met, the parameters are maintained to keep the population size and attraction coefficient unchanged, and standard iteration is maintained to continue the regular optimization process, thereby maintaining computational efficiency in low-risk regions and avoiding resource waste. The triggering condition is that the deviation of the relay protection device under the multi-source monitoring parameters corresponding to the optimal firefly is greater than or equal to twice the latent fault threshold of the relay protection device.

[0182] This invention overcomes the detection blind spots of traditional self-inspection mechanisms for complex, latent faults by integrating multi-source data dynamic modeling and intelligent optimization algorithms. Through hierarchical modeling of parameter types and specific parameters, the accuracy and interpretability of fault detection are improved. The system constructs a dynamic health baseline covering the entire operating state of the device and, combined with an improved firefly algorithm, adaptively searches the high-dimensional parameter space, enabling accurate identification of latent defects such as capacitor aging, logic drift, and communication anomalies even when the device is in a self-inspection state without alarms. Its dynamic threshold mechanism can autonomously adjust the judgment criteria according to the degree of equipment aging and environmental fluctuations, effectively reducing the risk of false alarms and missed detections. Simultaneously, its rapid fault feature localization function provides precise maintenance guidance for operation and maintenance personnel. This solution provides innovative technical support for the safe and stable operation of power systems.

Claims

1. A method for detecting latent faults in a relay protection device, characterized in that, include: Obtain the current detection values ​​of multi-source monitoring parameters of the relay protection device. The types of multi-source monitoring parameters include at least two of power supply parameters, environmental parameters, communication performance parameters, and logic unit parameters. Power supply parameters include voltage fluctuation and / or ripple coefficient; environmental parameters include at least one of temperature, humidity, and vibration; communication performance parameters include transmission delay and / or bit error rate; and logic unit parameters include CPU load and / or memory usage. Determine the deviation degree of the relay protection device based on the deviation between the current detection values ​​of the multi-source monitoring parameters and the health baseline of the multi-source monitoring parameters. The presence of a latent fault is determined based on the deviation of the relay protection device and the latent fault threshold of the relay protection device.

2. The method for detecting latent faults in relay protection devices according to claim 1, characterized in that, The deviation of the relay protection device is calculated using the following formula: ;in, ΔP is the deviation of the relay protection device, n is the total number of types of multi-source monitoring parameters, and w i Let m be the weight of the multi-source monitoring parameters corresponding to the i-th class. i P represents the number of multi-source monitoring parameters of the i-th class. current,i,j Let P be the current detected value of the j-th parameter in the multi-source monitoring parameters of the i-th class. baseline,i,j Let be the health baseline of the j-th parameter among the multi-source monitoring parameters of the i-th class.

3. The method for detecting latent faults in relay protection devices according to claim 1 or 2, characterized in that, The latent fault threshold of the relay protection device is a dynamically adjusted threshold. The longer the cumulative operating time of the relay protection device, the higher the threshold for latent faults; the greater the environmental fluctuations, the higher the threshold for latent faults.

4. The method for detecting latent faults in relay protection devices according to claim 3, characterized in that, The latent fault threshold of the relay protection device is calculated according to the following formula: ;in, θ(t) is the latent fault threshold at time t, θ0 is the initial latent fault threshold, k is the aging coefficient, t is the cumulative operating time of the relay protection device, and T life For the design life of the relay protection device, σ ΔP β represents the standard deviation of historical deviation. θ This is the threshold adjustment coefficient.

5. The method for detecting latent faults in relay protection devices according to claim 2, characterized in that, The method also includes increasing the weight of the multi-source monitoring parameter category that contributes the most to the deviation of the relay protection device if the number of times the deviation of the relay protection device exceeds the latent fault threshold of the relay protection device reaches a first threshold, and using the adjusted weight to calculate the deviation of the relay protection device in the next latent fault detection.

6. The method for detecting latent faults in relay protection devices according to claim 1, characterized in that, The method also includes automatically updating the health baseline of the multi-source monitoring parameters according to the following formula if the deviation of the relay protection device is less than the latent fault threshold of the relay protection device: P baseline-new,i,j =A×P baseline-old,i,j +B×P current,i,j; Among them, P baseline-new,i,j For the health baseline of the j-th parameter in the updated multi-source monitoring parameters of the i-th class, P baseline-old,i,j For the healthy baseline of the j-th parameter in the i-th class of multi-source monitoring parameters before the update, P current,i,j Let A be the current detection value of the j-th parameter in the multi-source monitoring parameters of the i-th class, where A is the first weight and B is the second weight, and A is greater than B.

7. The method for detecting latent faults in relay protection devices according to claim 1, characterized in that, Determining whether a latent fault exists includes: if the relay protection device self-test is normal, and the deviation of the relay protection device is greater than one time the latent fault threshold of the relay protection device but less than two times the latent fault threshold of the relay protection device, then a low-level alarm is generated. If the relay protection device performs a normal self-test, and the deviation of the relay protection device is greater than or equal to twice the hidden fault threshold of the relay protection device, a high-level alarm is generated, and an optimization algorithm is used to locate the fault.

8. The method for detecting latent faults in a relay protection device according to claim 7, characterized in that, The optimization algorithm adopts the firefly algorithm. Each firefly in the firefly algorithm is a set of multi-source monitoring parameters. The brightness function is used to find the optimal firefly in each iteration. It is determined whether the optimal firefly meets the strict convergence condition. If the strict convergence condition is met, the iteration stops. The fault type is determined according to the multi-source monitoring parameters corresponding to the optimal firefly and emergency fault handling is performed. If the strict convergence condition is not met, then it is determined whether the weak convergence condition is met. If the weak convergence condition is not met, the iteration continues until the weak convergence condition is met, and the optimal firefly at the point of stopping iteration is output. If the deviation of the relay protection device under the multi-source monitoring parameters corresponding to the optimal firefly is greater than or equal to twice the hidden fault threshold of the relay protection device, then the fault type is determined according to the multi-source monitoring parameters corresponding to the optimal firefly and the fault is handled. The strict convergence condition is determined based on the principle of ensuring that the firefly swarm is both tightly clustered and reaches a high degree of deviation, while the weak convergence condition is determined based on the principle of preventing infinite loops.

9. The method for detecting latent faults in relay protection devices according to claim 8, characterized in that, The brightness function is: ;in, I i Let be the brightness of the i-th firefly, α be the self-test state difference coefficient, β be the deviation coefficient, and D be the value of the firefly. self,i D* represents the self-test status value of the relay protection device. self,i ΔP is the ideal value for self-testing of the relay protection device. i Let be the deviation of the relay protection device under the multi-source monitoring parameters corresponding to the i-th firefly.

10. A latent fault detection system for a relay protection device, comprising a processor, characterized in that, The processor is used to execute the method for detecting latent faults in a relay protection device as described in any one of claims 1-9.

Images