Substation fault prediction method based on multi-level device cooperation

Abstract

The present application relates to the substation operation and maintenance technical field, especially to a kind of substation fault prediction method based on multistage equipment cooperation, the method in the operation process of substation, the real-time monitoring data of multiple sources of target equipment in current cycle is collected, and the equipment real-time state feature vector is obtained;Obtain the fault feature range and feature weight corresponding to each feature in current cycle under each preset fault type, match each feature data in equipment real-time state feature vector, if at least two preset fault types are matched, then it is recorded as target fault type;For any target fault type, the feature compliance and feature weight of each feature under any target fault type are combined, and the fault matching confidence of any target fault type is obtained;According to the fault matching confidence of each target fault type, the final fault type of target equipment in current cycle is determined, which can greatly reduce the fault false alarm rate and improve the fault prediction accuracy.

Description

Technical Field

[0001] The present invention relates to the technical field of substation operation and maintenance, and in particular to a substation fault prediction method based on multi-level equipment collaboration. Background Art

[0002] With the advancement of the intelligent transformation of power systems, substations, as core hubs of the power grid, face increasingly complex challenges in their safe operation. Predicting substation equipment faults is a critical step in ensuring safe and stable power grid operation. Currently, existing technologies generally use multi-source sensing technologies (such as infrared thermal imaging, partial discharge detection, and vibration and soundprint monitoring) to collect substation equipment status data and provide fault warnings through artificial intelligence models. However, in actual deployments, a core issue facing multi-source sensing technologies is the inefficient coordination of multi-source data, resulting in a high rate of false alarms, which seriously hinders the improvement of fault prediction accuracy. False alarms in substations not only lead to inefficient consumption of operation and maintenance resources, but also potentially mask real hidden dangers in false warnings.

[0003] Therefore, most of the existing substation fault prediction methods use data alignment algorithms or feature-level fusion to improve the synergy of multi-source data. However, due to inherent defects, the proportion of false alarms caused by the failure of multi-source data collaboration is still high, which limits the improvement of fault prediction accuracy. Among them, the inherent defects are specifically manifested as follows: static fault prediction rules cannot adapt to equipment status and environmental changes (such as equipment aging, climate change), and only raw data or low-level features are transmitted between edge terminals (such as drones, inspection robots, fixed sensors) and central systems, and fault semantic-level collaborative reasoning is not realized (such as the mapping verification of the "vibration + temperature + current" combination features and the "mechanical jam" fault mode). Summary of the Invention

[0004] In view of this, an embodiment of the present invention provides a substation fault prediction method based on multi-level equipment collaboration to solve the problem that the efficiency of multi-source data collaboration is low, which limits the accuracy of substation fault prediction.

[0005] An embodiment of the present invention provides a substation fault prediction method based on multi-level equipment collaboration, the method comprising the following steps:

[0006] During the operation of the substation, multi-source real-time monitoring data of the target device in the current cycle is collected by using multi-level equipment, and feature extraction is performed on the multi-source real-time monitoring data to obtain a device real-time status feature vector composed of at least two feature data;

[0007] Obtaining the fault feature range and feature weight corresponding to each feature under each preset fault type in the current cycle, respectively, and using the fault feature range corresponding to each feature under each preset fault type to match each feature data in the device real-time state feature vector. If at least two preset fault types are matched, the matched preset fault type is recorded as the target fault type;

[0008] For any target fault type, based on the relationship between each feature data in the real-time state feature vector of the device and the fault feature range corresponding to each feature under the any target fault type, obtain the feature conformity of each feature under the any target fault type respectively, and combine the feature conformity and feature weight of each feature under the any target fault type to obtain the fault matching confidence of the any target fault type;

[0009] Obtain a fault matching confidence of each target fault type, and determine a final fault type of the target device in the current cycle based on the fault matching confidence of each target fault type.

[0010] Compared with the prior art, the embodiments of the present invention have the following beneficial effects:

[0011] The present invention converts the correlation of core physical parameters (such as the inevitable acoustic-optical-thermal effects of discharge) into a device state feature vector that can represent the operating status of the device, and then dynamically adjusts the initial feature hyperparameter set obtained based on simulation to obtain the fault feature range and feature weight corresponding to each feature under each preset fault type in the current cycle, so as to avoid the accuracy degradation of fault prediction using static fault thresholds and feature parameters due to equipment aging and environmental changes. The fault feature range and feature weight corresponding to each feature under each preset fault type in the current cycle are used to match the feature data in the real-time state feature vector of the device in the current cycle. When at least two preset fault types are matched, the final fault type is determined by analyzing the fault matching confidence of each preset fault type, which can greatly reduce the fault false alarm rate and improve the accuracy of fault prediction. BRIEF DESCRIPTION OF THE DRAWINGS

[0012] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0013] Figure 1 This is a method flow chart of a substation fault prediction method based on multi-stage equipment collaboration provided in Example 1 of the present invention. DETAILED DESCRIPTION

[0014] The embodiments of the present disclosure are described in detail below, and examples of the embodiments are shown in the accompanying drawings. The embodiments described below with reference to the accompanying drawings are exemplary and intended to be used to explain the present disclosure, but should not be understood as limiting the present disclosure.

[0015] It should be noted that the terms "first," "second," and the like in the specification of the present disclosure and the accompanying drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or precedence. It should be understood that the terms used in this manner are interchangeable where appropriate, so that the embodiments of the present disclosure described herein can be implemented in an order other than those illustrated or described herein. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present disclosure. Instead, they are merely examples of devices and methods consistent with some aspects of the present disclosure.

[0016] In order to illustrate the technical solution of the present invention, specific embodiments are provided below.

[0017] See also Figure 1 , is a method flow chart of a substation fault prediction method based on multi-stage equipment collaboration provided by the first embodiment of the present invention, such as Figure 1 As shown, the method may include:

[0018] Step S101 , during the operation of the substation, multi-source real-time monitoring data of the target device in the current cycle is collected using multi-level equipment, and features are extracted from the multi-source real-time monitoring data to obtain a device real-time status feature vector consisting of at least two feature data.

[0019] The main purpose of the present invention is to predict the fault type and the authenticity of the occurrence of each type of fault for each device in the substation through a multi-level device collaboration method, so as to improve the prediction accuracy, reduce the fault false alarm rate, thereby enhancing the emergency protection capability of the substation and reducing the operation and maintenance costs of the substation. The devices include but are not limited to heat-reliant devices (such as transformers and reactors) and mechanical operation devices (such as circuit breakers and disconnectors) in the substation, as well as other key devices with multi-physical quantity monitorable characteristics. Therefore, an embodiment of the present invention takes any device in the substation as an example and regards any device as the target device. By collecting multi-source monitoring data of the target device, the fault type is predicted, and the prediction results are transmitted to the early warning center, so that the early warning center can take corresponding operation and maintenance measures based on the prediction results.

[0020] The multi-source monitoring data includes: using drones to scan the surface of target equipment from high altitude to obtain infrared thermal imaging of the target equipment surface; using a four-legged robot dog to get close to the base of the target equipment to obtain the target equipment's vibration signal and acoustic emission signal; using a patrol robot to connect to the target equipment's oil valve to obtain the hydrogen concentration in the oil valve. The specific method for collecting multi-source monitoring data is as follows:

[0021] (1) Time synchronization

[0022] A clock synchronization protocol is used to unify the timestamps of all enabled multi-level devices (drones, quadruped robot dogs, and patrol robots), thereby unifying the timestamps of all collected parameters. Sensor data collection from multiple devices is asynchronous, and timestamp unification eliminates timing deviations caused by sampling frequency differences (e.g., 10fps for drone thermal imaging and 44.1kHz for robot dog voiceprints). For example, when a drone detects an abnormal temperature rise in a device, the timeline must be precisely aligned with the acoustic emission signal collected by the robot dog to verify whether it is the same fault event.

[0023] (2) Spatial synchronization

[0024] UAV: The local coordinate system is the local point cloud coordinate system constructed by SLAM (simultaneous localization and mapping). The global UWB (ultra-wideband) point cloud is matched using the ICP (iterative closest point) algorithm, and the 6-DOF rigid body transformation matrix is ​​solved. Based on the rigid body transformation matrix, the local coordinates of the UAV are converted to global coordinates.

[0025] Quadruped robot dog: The local coordinate system is a two-dimensional plane coordinate system based on the odometry + IMU (inertial measurement unit), and the global coordinates are directly obtained through the UWB tag on the back.

[0026] Patrol robot: The local coordinate system is the preset track coordinate system. During installation, the track starting point is calibrated by laser, the displacement is calculated by the encoder, and the local coordinates of the track patrol robot are mapped to the global coordinates based on the displacement.

[0027] In this way, even if multiple devices collect data based on different coordinate systems (such as the drone SLAM local coordinate system and the UWB global coordinate system), they can eventually convert the coordinates of the collected data into the global coordinate system to achieve spatial synchronization.

[0028] Since equipment failures do not occur in a short period of time, the set fault prediction period should not be too short. However, if the fault prediction period is too long, it will cause delays in fault prediction and may easily lead to untimely operation and maintenance. Therefore, the embodiment of the present invention sets the equipment failure prediction period to one day. There is no restriction here and it can be set according to the implementation scenario.

[0029] Based on the above-mentioned multi-source monitoring data collection method, multi-source monitoring data of the target equipment in each cycle can be collected, and then multi-source real-time monitoring data of the target equipment in the current cycle can be obtained. The multi-source real-time monitoring data includes infrared thermal imaging image sequences, vibration signals, acoustic emission signals and hydrogen concentration sequences in oil valves.

[0030] Due to the dimensional differences and imbalanced numerical scales of multi-source monitoring data, in order to avoid large-scale features dominating decision-making, it is necessary to extract features from multi-source real-time monitoring data to obtain a device real-time status feature vector composed of at least two feature data. The specific method is as follows:

[0031] Using a Sobel operator to calculate the gradient value of each pixel in each image in the infrared thermal imaging image sequence, obtain a maximum gradient value absolute value, and normalize the maximum gradient value absolute value to obtain a maximum temperature gradient value;

[0032] Performing Fourier transform on the vibration signal to obtain a spectrum graph, obtaining an entropy value of the spectrum graph, and normalizing the entropy value to obtain vibration spectrum entropy;

[0033] Acquiring an energy value corresponding to the acoustic emission signal, and normalizing the energy value to obtain acoustic emission energy;

[0034] performing a sliding average filter on the hydrogen concentration sequence to obtain a filtered sequence, calculating an average value of the filtered sequence, and normalizing the average value to obtain an average hydrogen concentration;

[0035] The maximum temperature gradient value, the vibration spectrum entropy, the acoustic emission energy and the average hydrogen concentration are used as characteristic data to form a real-time state characteristic vector of the equipment.

[0036] It should be noted that the maximum absolute value of the gradient, the entropy of the spectrogram, and the average value of the filtered sequence are continuous features and are normalized using the Z-score method. The energy value corresponding to the acoustic emission signal is a count feature and is normalized using the Min-Max method. The Sobel operator and normalization are both existing technologies and will not be discussed here.

[0037] Similarly, the device state feature vector of the target device in each cycle before the current cycle can also be obtained.

[0038] In step S102, the fault feature range and feature weight corresponding to each feature under each preset fault type in the current cycle are obtained respectively, and each feature data in the real-time status feature vector of the device is matched using the fault feature range corresponding to each feature under each preset fault type. If at least two preset fault types are matched, the matched preset fault type is recorded as the target fault type.

[0039] Before determining the fault type of the target device in the current cycle, it is necessary to obtain the fault threshold corresponding to each characteristic data involved in each fault type, so as to determine which fault type the target device belongs to based on the real-time status characteristic vector of the target device. Therefore, whether the fault threshold corresponding to each characteristic data involved in each fault type is accurate will directly affect the accuracy of the fault type judgment. In the past, most of the fault thresholds corresponding to each characteristic data involved in each fault type were set based on statistical experience, which could not match the current actual situation of the device, resulting in false alarms of the fault type. Therefore, the embodiment of the present invention establishes a digital twin model of the target device, matches the current actual situation of the target device, and simulates multiple fault types to obtain accurate fault thresholds.

[0040] Specifically, first build a digital twin model of the target device. When building the digital twin model, it is necessary to first determine the geometric model of the target device (modeling based on the actual size of the target device), physical model (including electromagnetic field, thermal field, structural mechanics field, fluid dynamics field) and boundary conditions (including ambient temperature, load current, voltage level, cooling conditions, etc.) to ensure the consistency of the digital twin model with the actual conditions of the target device, thereby ensuring the accuracy of the simulation results obtained in subsequent simulations.

[0041] However, due to the uncertainty of the target device state, such as device size deviation, thermal conductivity deviation of material properties, operating conditions and fault degree differences, a single simulation cannot cover most working conditions, which will cause the fault threshold obtained based on a single simulation to be unable to adapt to actual fluctuations, and the accuracy of the fault boundary condition setting is not high. Therefore, the embodiment of the present invention uses the Monte Carlo method to perform multiple fault simulations on the digital twin model to solve the uncertainty of the fault threshold and boundary conditions through probability sampling.

[0042] This embodiment of the present invention sets four preset fault types: partial discharge, mechanical looseness, insulation moisture, and overheating. These are not limited here and can be set based on the implementation scenario. Monte Carlo simulations are then run 100 times for each preset fault type to obtain a set of device state feature vectors for each preset fault type. Each Monte Carlo simulation outputs a single device state feature vector.

[0043] Finally, according to the set of device simulation state feature vectors under each preset fault type, the initial fault feature range corresponding to each feature under each preset fault type is obtained. Specifically, in the Monte Carlo simulation, since the input parameters (fault severity) are subjected to normal distribution perturbations, according to the central limit theorem, the output device state feature vector approximately obeys the normal distribution. As for the normal distribution, by querying the standard normal distribution table, it can be obtained that the interval of 1.96 times the standard deviation can cover 95% of the data, which means that in 100 simulations, the feature values ​​of approximately 95 samples will fall into this interval. Therefore, this interval reflects the natural fluctuation range of the device state feature vector under a specific fault type. The device state feature vector outside this range is very likely to be other faults or abnormal states. Then, in an embodiment of the present invention, for the jth feature under the i-th preset fault type, according to the feature simulation data corresponding to the jth feature in the device simulation state feature vector set under the i-th preset fault type, the average feature simulation data and the standard deviation of the feature simulation data are obtained, and the set As the initial fault feature range corresponding to the jth feature under the i-th preset fault type, where represents the average characteristic simulation data, Represents the standard deviation of the characteristic simulation data.

[0044] For example, when the fault type is mechanical looseness, after Monte Carlo simulation, the vibration spectrum entropy in the device simulation state eigenvector is significantly higher than that in the normal state, indicating that the device vibration has significantly increased. Therefore, the lower and upper limits of the fault characteristic threshold of the vibration spectrum entropy for the mechanical looseness fault type are significantly higher than those in the normal state. Similarly, when the fault type is insulation moisture, after Monte Carlo simulation, the average hydrogen concentration in the device simulation state eigenvector is significantly increased, indicating a sudden increase in hydrogen. At the same time, the maximum temperature gradient value decreases significantly, indicating abnormal equipment cooling. Therefore, the lower and upper limits of the fault characteristic threshold of the average hydrogen concentration for the insulation moisture fault type are significantly higher than those in the normal state, while the lower and upper limits of the fault characteristic threshold of the maximum temperature gradient value are lower than those in the normal state. When the fault type is overheating, after Monte Carlo simulation, the maximum temperature gradient value exceeds twice the normal value, indicating extreme equipment overheating. Therefore, the lower and upper limits of the fault characteristic threshold of the maximum temperature gradient value for the overheating fault type are higher than those in the normal state and other fault types.

[0045] At the same time, since different fault types may have overlapping features, relying solely on the fault threshold to judge the fault type may lead to confusion of fault types. For example, the judgment of multiple fault types may be met at the same time, resulting in false alarms. Therefore, it is necessary to calculate the confidence of each fault type. The feature weight is the key basis for calculating the confidence, which represents the ability of each feature corresponding to each fault type to distinguish the fault from other faults. The fault confidence is calculated by assigning a feature weight to each feature corresponding to each fault type, thereby improving the ability to make confidence decisions on fault types and thus improving the accuracy of fault prediction.

[0046] In the embodiment of the present invention, for the jth feature under the i-th preset fault type, first, based on the difference in feature simulation data of the jth feature under the i-th preset fault type and under other preset fault types, the discriminative ability index of the jth feature is obtained. The obtaining method is:

[0047] After performing 100 Monte Carlo simulations on each preset fault type, a feature simulation data set for each feature of each preset fault type is obtained. Each feature simulation data set is generated independently and is a one-dimensional array containing 100 feature simulation data. Therefore, the Wasserstein distance can be used to quantify the distribution difference between the feature simulation data set of the jth feature under the i-th preset fault type and the feature simulation data set of the jth feature under the y-th preset fault type. The Wasserstein distance can quantify the cost of moving one distribution to another. If the distribution difference (such as the mean difference and the standard deviation difference) is large, the cost of moving will be large, that is, the Wasserstein distance will be large. Then the Wasserstein distance between the feature simulation data sets corresponding to the jth feature of two different preset fault types is used as the feature separation degree of the jth feature for the two different preset fault types.

[0048] The feature simulation data set of the jth feature of the i-th preset fault type As the benchmark group, the feature simulation data set of the jth feature under the yth preset fault type is As a comparison group. Since the two sets of data need to be aligned according to their respective quantiles (i.e., from the smallest to the largest) in order to make a one-to-one comparison, the baseline group Sort in ascending order to get the target sequence; for the comparison group Sort in ascending order to obtain a reference sequence.

[0049] The absolute value of the difference between each two elements at the same position between the target sequence and the reference sequence is calculated to obtain the mean of the absolute values ​​of the difference, which is used as the feature separation degree of the j-th feature between the i-th preset fault type and the y-th preset fault type. The calculation formula of the feature separation degree is:

[0050]

[0051] in, represents the feature separation degree of the jth feature between the i-th preset fault type and the y-th preset fault type, n represents the element data in the target sequence or reference sequence, Represents the a-th data in the target sequence, represents the ath data in the reference sequence, and || represents the absolute value symbol.

[0052] It should be noted that the larger W is, that is, the greater the feature separation is, the greater the distribution difference between the two groups of data is, and the easier it is to distinguish the two groups of data; the smaller W is, that is, the smaller the feature separation is, the smaller the distribution difference between the two groups of data is, and the more difficult it is to distinguish the two groups of data. In this case, the j-th feature is more important for judging the i-th preset fault type and the y-th preset fault type, and the feature weight assigned to this feature should be greater to avoid misjudgment of the fault type.

[0053] Similarly, the separation degree of the jth feature between the i-th preset fault type and each reference type is obtained. Feature separation reflects the distribution difference of a feature between different fault types. After Monte Carlo simulation, the smaller the distribution difference of the feature simulation data for the same feature of two fault types, the higher the similarity of the two fault types in this feature, the more difficult it is to distinguish, and thus the smaller the feature separation degree. This also indicates that the feature is more important for distinguishing the two fault types. Therefore, the separation degree of the jth feature between the i-th preset fault type and each reference type is obtained, and the minimum separation degree is used as the discriminative ability index of the jth feature for the i-th preset fault type.

[0054] Finally, the discriminative ability index of each feature under the i-th preset fault type is obtained, and the discriminative ability index of all features under the i-th preset fault type is normalized using the norm function to obtain the corresponding normalized value. The difference between the constant 1 and each normalized value is used as the initial feature weight corresponding to each feature under the i-th preset fault type.

[0055] At this point, the initial fault feature range and initial feature weight corresponding to each feature under the i-th preset fault type are obtained. Similarly, the initial fault feature range and initial feature weight corresponding to each feature under each preset fault type are obtained. The initial fault feature range and initial feature weight corresponding to each feature under all preset fault types are combined into an initial feature parameter set, which is used to judge the fault type of the target device based on the actually obtained device status feature vector.

[0056] After determining the initial feature parameter set, each feature data in the device status feature vector obtained by the target device in any cycle can be compared with the initial fault feature range corresponding to each feature under each preset fault type in the initial feature parameter set, so as to match the corresponding fault type. If there is no match, it is determined that the target device is in normal operation; if at least one fault type is matched, it is necessary to analyze the fault confidence to facilitate equipment operation and maintenance based on the fault confidence.

[0057] Since the initial feature parameter set is obtained based on simulation, with factors such as equipment aging and environmental changes, the static fault threshold and feature weight will affect the accuracy of fault prediction. Therefore, an embodiment of the present invention proposes a method for updating the fault threshold and feature weight based on the initial feature parameter set. That is, after determining the final fault type of the target device in each cycle, the initial feature parameter set is dynamically updated according to the cumulative detection results of the final fault type. By adjusting the fault feature range and feature weight corresponding to each feature under the preset fault type, the problem of attenuation of prediction accuracy caused by equipment aging and environmental changes is avoided.

[0058] Specifically, first, based on the initial feature parameter set and the target device's device state feature vector for each cycle, the target device's final fault type and corresponding fault matching confidence level are obtained. The specific acquisition method is described below. Then, the number of times each preset fault type is considered the final fault type is accumulated and recorded. At least one preset fault type corresponding to the cumulative number of times meeting a preset number threshold is obtained and recorded as the optimized fault type. It is worth noting that if an optimized fault type is not obtained, the initial feature parameter set is used to predict the target device's final fault type for each cycle.

[0059] Preferably, the present invention sets the preset number threshold to 10, that is, when the final fault type is the same preset fault type more than 10 times, the adjustment of the fault feature range and feature weight can be triggered. It should be noted that the preset number threshold cannot be too small. If it is too small, there will be insufficient number of samples and poor data reliability, which will lead to a decrease in the prediction accuracy after the update; if the preset number threshold is too high, there will be an update gap, and the prediction accuracy will be unsatisfactory for a period of time. Specifically, the preset number threshold corresponding to each preset fault type can be adjusted according to the climate of the substation area and the actual situation of the substation. For example: if the substation is located in an area with perennial rain and humid climate, the possibility of insulation moisture fault is greater and the number of samples is sufficient. The preset number threshold can be appropriately increased so that when the fault threshold and feature weight of the insulation moisture fault are updated, the data based on it is more sufficient. There is no restriction here.

[0060] Then, for any optimized fault type, assuming that any optimized fault type is obtained after the target device determines the final fault type in the bth cycle, all fault matching confidences corresponding to any optimized fault type when it is used as the final fault type, and the device state feature vectors corresponding to the target device in all cycles are obtained before the bth cycle, and based on all device state feature vectors and all fault matching confidences corresponding to any optimized fault type, the initial fault feature range and initial feature weight corresponding to each feature under any optimized fault type are updated to obtain a new fault feature range and new feature weight corresponding to each feature under any optimized fault type. The specific updating method is as follows:

[0061] (1) According to all the device state feature vectors corresponding to any of the optimized fault types, obtain the feature data mean of each feature , calculate the fault matching confidence mean according to all fault matching confidences corresponding to any optimized fault type .

[0062] (2) For any feature under any of the optimized fault types, obtain an intermediate value in the initial fault feature range corresponding to any of the features, calculate the absolute value of the difference between the feature data mean of any of the features and the intermediate value, and use the ratio of the absolute value of the difference to the intermediate value as the weight gradient of any of the features; use the product of the weight gradient, the fault matching confidence mean and the preset weight learning rate as the weight adjustment value, and use the sum of the initial feature weight corresponding to any of the features and the weight adjustment value as the new feature weight corresponding to any of the features.

[0063] In one embodiment, taking the jth feature under any optimized fault type as an example, the calculation formula for the new feature weight corresponding to the jth feature is:

[0064]

[0065] in, represents the new feature weight corresponding to the j-th feature, represents the initial feature weight corresponding to the j-th feature, Represents the preset weight learning rate, represents the weight gradient of the jth feature, represents the mean confidence value of fault matching.

[0066] It should be noted that Used to control the weight update step size, set , which is the traditional default value; Used to quantify feature contribution, so as to accurately locate features that need to be optimized, ,in, represents the mean of the feature data corresponding to the jth feature, It represents the middle value in the initial fault feature range corresponding to the j-th feature under any optimized fault type, that is, the ideal value. The larger the value of , the greater the gap between the feature data of the j-th feature and the ideal value, the larger the corresponding weight gradient, the greater the feature contribution of the j-th feature, and the greater the corresponding adjustment degree.

[0067] (3) Obtaining the upper limit and lower limit of the initial fault feature range corresponding to any one of the features, obtaining a first difference between the feature data mean of any one of the features and the lower limit, taking the product of a preset learning rate and the first difference as the lower limit adjustment value, and taking the difference between the lower limit and the lower limit adjustment value as the new lower limit; obtaining a second difference between the feature data mean of any one of the features and the upper limit, taking the product of a preset learning rate and the second difference as the upper limit adjustment value, and taking the sum of the upper limit and the upper limit adjustment value as the new upper limit; obtaining a new fault feature range corresponding to any one of the features according to the new upper limit and the new lower limit.

[0068] In one embodiment, taking the jth feature under any optimized fault type as an example, the update calculation formula for the lower limit value of the initial fault feature range corresponding to the jth feature is:

[0069]

[0070] in, represents the new lower limit value corresponding to the j-th feature, represents the lower limit of the initial fault feature range corresponding to the jth feature, Represents the preset learning rate, Represents the mean of the feature data corresponding to the jth feature.

[0071] The update calculation formula for the upper limit of the initial fault feature range corresponding to the jth feature is:

[0072]

[0073] in, represents the new upper limit value corresponding to the j-th feature, represents the upper limit of the initial fault feature range corresponding to the jth feature, Represents the preset learning rate, Represents the mean of the feature data corresponding to the jth feature.

[0074] It should be noted that For controlling the adjustment speed of the fault characteristic range, in the embodiment of the present invention, .

[0075] The initial feature parameter set is updated using the new fault feature range and new feature weight corresponding to each feature under each of the optimized fault types to obtain a new feature parameter set. Specifically, the new fault feature range and new feature weight corresponding to each feature under each of the optimized fault types are used to replace the initial fault feature range and initial feature weight corresponding to each feature under the corresponding preset fault type in the initial feature parameter set to obtain a new feature parameter set.

[0076] For example, if the final fault type is that the number of times the insulation is damp reaches a preset threshold, after updating the initial feature parameter set, it is found that the initial feature weight corresponding to the average hydrogen concentration in the insulation damp fault type has increased slightly, which means that the recent average hydrogen concentration feature has a greater impact on the insulation damp fault type, so the update has increased the contribution of the average hydrogen concentration feature to the insulation damp fault; at the same time, the initial fault feature range corresponding to the average hydrogen concentration has become larger after the update, which means that similar faults may be more likely to occur in the future, so the initial fault feature range is expanded by updating, making it easier to detect similar faults using the updated initial feature parameter set (that is, the new feature parameter set).

[0077] Afterwards, the new feature parameter set is used as the initial feature parameter set to continue to obtain the final fault type of the target device in each cycle. When an optimized fault type occurs, the initial feature parameter set (that is, the new feature parameter set) is updated again according to the updating method of the initial feature parameter set until the final fault type of the target device in the previous cycle of the current cycle is obtained. Based on the final fault type of the target device in the previous cycle of the current cycle, the fault feature range and feature weight corresponding to each feature under each preset fault type in the current cycle are obtained.

[0078] It should be noted that if no optimized fault type is obtained based on the final fault type of the target device in the previous cycle of the current cycle, it means that the cumulative number of times that no preset fault type is the final fault type in the previous cycle up to the current cycle meets the preset number threshold, so there is no need to update the initial feature parameter set, and the initial fault feature range and initial feature weight corresponding to each feature under each preset fault type in the initial feature parameter set are directly used as the fault feature range and feature weight corresponding to each feature under each preset fault type in the current cycle; conversely, if any optimized fault type is obtained, the initial feature parameter set is updated, and the initial fault feature range and initial feature weight corresponding to each feature under each preset fault type in the updated initial feature parameter set are used as the fault feature range and feature weight corresponding to each feature under each preset fault type in the current cycle.

[0079] It should be noted that all initial feature parameter sets after the initial feature parameter set is updated for the first time refer to new feature parameter sets obtained after the update.

[0080] At this point, the fault feature range and feature weight corresponding to each feature under each preset fault type in the current cycle can be obtained. Furthermore, using the fault feature range and feature weight corresponding to each feature under each preset fault type in the current cycle, each feature data in the device real-time status feature vector obtained by the target device in the current cycle is matched, that is, to detect whether each feature data in the device real-time status feature vector is within the fault feature range of the corresponding feature under any preset fault type. If so, the match is successful, and the preset fault type that is successfully matched is recorded as the target fault type; otherwise, no match is successful. If a target fault type is matched, the matched preset fault type is used as the final fault type of the target device in the current cycle, and the fault matching confidence of the matched preset fault type is obtained; if multiple target fault types are matched, the final fault type of the target device in the current cycle is determined based on the fault confidence of each target fault model.

[0081] Step S103: For any target fault type, based on the relationship between each feature data in the real-time status feature vector of the device and the fault feature range corresponding to each feature under any target fault type, obtain the feature conformity of each feature under any target fault type respectively, and combine the feature conformity and feature weight of each feature under any target fault type to obtain the fault matching confidence of any target fault type.

[0082] When analyzing the confidence of each target fault type, the embodiment of the present invention not only considers the feature weights corresponding to each feature under the target fault type, but also needs to consider the feature conformity of each feature data in the device real-time status feature vector of the target device in the current cycle, that is, the degree of conformity of a single feature data compared to the standard feature under the target fault type. Therefore, taking any target fault type as an example, the feature conformity of each feature under any target fault type is obtained respectively according to the relationship between each feature data in the device real-time status feature vector and the fault feature range corresponding to each feature under any target fault type.

[0083] Wherein, according to the relationship between each feature data in the real-time state feature vector of the device and the fault feature range corresponding to each feature under any target fault type, respectively obtaining the feature conformity of each feature under any target fault type includes:

[0084] For any feature under any target fault type, calculate the absolute value of the difference between the feature data corresponding to any feature in the real-time state feature vector of the device and the median value within the fault feature range corresponding to any feature under any target fault type, and record it as the deviation degree;

[0085] Calculate the upper limit or lower limit of the fault feature range corresponding to any feature under any target fault type, and the absolute value of the difference between the upper limit or lower limit and the middle value, and record it as the baseline feature value; calculate the ratio of the deviation degree to the baseline feature value, and take the difference between the constant 1 and the ratio as the feature conformity of any feature.

[0086] In one embodiment, taking the jth feature under any target fault type as an example, the calculation formula for the feature conformance of the jth feature is:

[0087]

[0088] in, Indicates the feature conformity of the jth feature, 1 represents a constant, Represents the feature data corresponding to the jth feature in the real-time status feature vector of the device, represents the middle value in the fault feature range corresponding to the jth feature under any target fault type, It represents the absolute value of the difference between the upper limit (or lower limit) and the middle value within the fault feature range corresponding to the j-th feature under any target fault type, which is also the benchmark feature value.

[0089] It should be noted that Indicates the distance between the feature data of the jth feature and the median value within the fault feature range corresponding to any target fault type. The closer the distance, the The smaller the value, the better the feature conformity. The larger the value is, the higher the degree of conformity between the j-th feature and the standard feature of any target fault type, and the higher its contribution in the subsequent calculation of fault matching confidence.

[0090] Similarly, the feature conformity of each feature under any target fault type is obtained, and then the feature conformity and feature weight of each feature under any target fault type are combined to obtain the fault matching confidence of any target fault type: according to the feature weight of each feature under any target fault type, the feature conformity of all features under any target fault type is weighted and summed to obtain the fault matching confidence of any target fault type.

[0091] In one embodiment, the calculation formula for the fault matching confidence of any target fault type is:

[0092]

[0093] in, represents the fault matching confidence of any target fault type, represents the feature weight of the jth feature under any target fault type, It represents the feature conformity of the jth feature under any target fault type.

[0094] Step S104 : obtaining the fault matching confidence of each target fault type, and determining the final fault type of the target device in the current cycle according to the fault matching confidence of each target fault type.

[0095] According to the above-mentioned method for obtaining the fault matching confidence of any target fault type, the fault matching confidence of each target fault type is obtained according to the real-time device status feature vector of the target device in the current cycle, and based on the fault matching confidence of each target fault type, the target fault type corresponding to the maximum fault matching confidence is used as the final fault type of the target device in the current cycle.

[0096] At this point, the final fault type of the target device in the current cycle and the corresponding fault matching confidence can be obtained, and then according to the updating method of the initial feature parameter set in step S102, that is, the number of times each preset fault type is the final fault type is continued to be accumulated and recorded, and at least one preset fault type corresponding to the cumulative number that meets the preset number threshold is obtained, which is recorded as the optimized fault type. The initial feature parameter set is updated according to the optimized fault type to obtain the fault feature range and feature weight corresponding to each feature under each preset fault type in the next cycle of the current cycle, and combined with the device real-time status feature vector of the target device in the next cycle of the current cycle, the final fault type of the target device in the next cycle of the current cycle is predicted.

[0097] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. These modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the various embodiments of the present invention, and should all be included in the scope of protection of the present invention.

Claims

1. A substation fault prediction method based on multi-level equipment collaboration is characterized by: The method comprises: During the operation of the substation, multi-source real-time monitoring data of the target device in the current cycle is collected by using multi-level equipment, and feature extraction is performed on the multi-source real-time monitoring data to obtain a device real-time status feature vector composed of at least two feature data; Obtaining the fault feature range and feature weight corresponding to each feature under each preset fault type in the current cycle, respectively, and using the fault feature range corresponding to each feature under each preset fault type to match each feature data in the device real-time state feature vector. If at least two preset fault types are matched, the matched preset fault type is recorded as the target fault type; For any target fault type, based on the relationship between each feature data in the real-time state feature vector of the device and the fault feature range corresponding to each feature under the any target fault type, obtain the feature conformity of each feature under the any target fault type respectively, and combine the feature conformity and feature weight of each feature under the any target fault type to obtain the fault matching confidence of the any target fault type; Obtain the fault matching confidence of each target fault type, and based on the fault matching confidence of each target fault type, use the target fault type corresponding to the maximum fault matching confidence as the final fault type of the target device in the current cycle; The step of respectively obtaining the fault feature range and feature weight corresponding to each feature under each preset fault type in the current cycle includes: Construct a digital twin model of the target device, perform multiple fault simulations on the digital twin model, obtain a set of device simulation state feature vectors under each preset fault type, and obtain an initial feature parameter set consisting of an initial fault feature range and an initial feature weight corresponding to each feature under each preset fault type based on the set of device simulation state feature vectors under each preset fault type; Obtaining, based on the initial feature parameter set and the device state feature vector of the target device in each cycle, the final fault type of the target device in each cycle and the corresponding fault matching confidence level, accumulating the number of times each preset fault type is the final fault type, and obtaining at least one preset fault type corresponding to a cumulative number that meets a preset number threshold, recording the optimized fault type; For any optimized fault type, all device state feature vectors and all fault matching confidences corresponding to the optimized fault type are obtained, and based on all device state feature vectors and all fault matching confidences corresponding to the optimized fault type, the initial fault feature range and initial feature weight corresponding to each feature under the optimized fault type are updated to obtain a new fault feature range and new feature weight corresponding to each feature under the optimized fault type; The initial feature parameter set is updated using the new fault feature range and new feature weight corresponding to each feature under each of the optimized fault types to obtain a new feature parameter set. The new feature parameter set is used as the initial feature parameter set to continue to obtain the final fault type of the target device in each cycle, and the updating method of the initial feature parameter set is repeated until the final fault type of the target device in the previous cycle of the current cycle is obtained. Based on the final fault type of the target device in the previous cycle of the current cycle, the fault feature range and feature weight corresponding to each feature under each preset fault type in the current cycle are obtained.

2. The substation fault prediction method based on multi-level equipment collaboration according to claim 1 is characterized in that: The feature extraction of the multi-source real-time monitoring data to obtain a device real-time status feature vector composed of at least two feature data includes: The multi-source real-time monitoring data includes infrared thermal imaging image sequences, vibration signals, acoustic emission signals and hydrogen concentration sequences in oil valves; Calculating the gradient value of each pixel in each image in the infrared thermal imaging image sequence to obtain a maximum gradient absolute value, and normalizing the maximum gradient absolute value to obtain a maximum temperature gradient value; Performing Fourier transform on the vibration signal to obtain a spectrum graph, obtaining an entropy value of the spectrum graph, and normalizing the entropy value to obtain vibration spectrum entropy; Acquiring an energy value corresponding to the acoustic emission signal, and normalizing the energy value to obtain acoustic emission energy; performing a sliding average filter on the hydrogen concentration sequence to obtain a filtered sequence, calculating an average value of the filtered sequence, and normalizing the average value to obtain an average hydrogen concentration; The maximum temperature gradient value, the vibration spectrum entropy, the acoustic emission energy and the average hydrogen concentration are used as characteristic data to form a real-time state characteristic vector of the equipment.

3. The substation fault prediction method based on multi-level equipment collaboration according to claim 1 is characterized in that: The method of obtaining, based on the device simulation state feature vector set under each preset fault type, an initial feature parameter set consisting of an initial fault feature range and an initial feature weight corresponding to each feature under each preset fault type includes: For any feature under any preset fault type, according to the feature simulation data corresponding to the feature in the device simulation state feature vector set under any preset fault type, the average feature simulation data and the feature simulation data standard deviation are obtained, and the As the initial fault feature range corresponding to any feature under any preset fault type, represents the average characteristic simulation data, Indicates the standard deviation of characteristic simulation data; Obtaining a discriminative ability index for any feature based on a difference in feature simulation data between any one of the features under any one of the preset fault types and under other preset fault types, obtaining a discriminative ability index for each feature under any one of the preset fault types, normalizing the discriminative ability indexes of all features under any one of the preset fault types to obtain corresponding normalized values, and using the difference between a constant 1 and each normalized value as an initial feature weight corresponding to each feature under any one of the preset fault types; The initial fault feature range and initial feature weight corresponding to each feature under each preset fault type are obtained to form an initial feature parameter set.

4. The substation fault prediction method based on multi-level equipment collaboration according to claim 3 is characterized in that: The obtaining, based on a difference in feature simulation data of the any feature under the any preset fault type and under other preset fault types, a discrimination capability index of the any feature includes: Taking any preset fault type other than the preset fault type as a reference type, sorting all feature simulation data of the any feature under the any preset fault type in ascending order to obtain a target sequence, and sorting all feature simulation data of the any feature under the reference type in ascending order to obtain a reference sequence; calculating the absolute value of the difference between two elements at the same position in the target sequence and the reference sequence, and obtaining the mean of the absolute values ​​of the difference as the feature separation degree of the any feature between the any preset fault type and the reference type; According to the feature separation degree of any one of the features between any one of the preset fault types and each of the reference types, a minimum feature separation degree is obtained as the discrimination capability index of any one of the features.

5. The substation fault prediction method based on multi-level equipment collaboration according to claim 3 is characterized in that: The updating of the initial fault feature range and initial feature weight corresponding to each feature under any optimized fault type according to all device state feature vectors and all fault matching confidences corresponding to any optimized fault type to obtain a new fault feature range and new feature weight corresponding to each feature under any optimized fault type includes: Obtaining a feature data mean of each feature based on all device state feature vectors corresponding to any one of the optimized fault types, and calculating a fault matching confidence mean based on all fault matching confidences corresponding to any one of the optimized fault types; For any feature under any optimized fault type, obtain an intermediate value in the initial fault feature range corresponding to the any feature, calculate the absolute value of the difference between the feature data mean of the any feature and the intermediate value, and use the ratio of the absolute value of the difference to the intermediate value as the weight gradient of the any feature; The product of the weight gradient, the fault matching confidence mean and a preset weight learning rate is used as a weight adjustment value, and the sum of the initial feature weight corresponding to any feature and the weight adjustment value is used as a new feature weight corresponding to any feature.

6. The substation fault prediction method based on multi-level equipment collaboration according to claim 5 is characterized in that: The updating of the initial fault feature range and initial feature weight corresponding to each feature under any optimized fault type according to all device state feature vectors corresponding to any optimized fault type to obtain a new fault feature range and new feature weight corresponding to each feature under any optimized fault type further includes: Obtaining an upper limit value and a lower limit value in an initial fault feature range corresponding to any one of the features, obtaining a first difference between a feature data mean of any one of the features and the lower limit value, multiplying a preset learning rate by the first difference value as a lower limit adjustment value, and using the difference between the lower limit value and the lower limit adjustment value as a new lower limit value; Obtaining a second difference between the feature data mean of any feature and the upper limit value, multiplying a preset learning rate and the second difference as an upper limit adjustment value, and using the sum of the upper limit value and the upper limit adjustment value as a new upper limit value; A new fault feature range corresponding to any feature is obtained according to the new upper limit value and the new lower limit value.

7. The substation fault prediction method based on multi-level equipment collaboration according to claim 1 is characterized in that: The initial feature parameter set is updated using the new fault feature range and new feature weight corresponding to each feature under each optimized fault type to obtain a new feature parameter set, including: The new fault feature range and new feature weight corresponding to each feature under each optimized fault type are used to replace the initial fault feature range and initial feature weight corresponding to each feature under the preset fault type in the initial feature parameter set to obtain a new feature parameter set.

8. The substation fault prediction method based on multi-level equipment collaboration according to claim 1 is characterized in that: The obtaining, based on the relationship between each feature data in the real-time state feature vector of the device and the fault feature range corresponding to each feature under any target fault type, the feature conformance of each feature under any target fault type includes: For any feature under any target fault type, calculate the absolute value of the difference between the feature data corresponding to any feature in the real-time state feature vector of the device and the median value within the fault feature range corresponding to any feature under any target fault type, and record it as the deviation degree; Calculate the upper limit or lower limit of the fault feature range corresponding to any feature under any target fault type, and the absolute value of the difference between the upper limit or lower limit and the middle value, and record it as the baseline feature value; calculate the ratio of the deviation degree to the baseline feature value, and take the difference between the constant 1 and the ratio as the feature conformity of any feature.

9. The substation fault prediction method based on multi-level equipment collaboration according to claim 1 is characterized in that: The acquiring the fault matching confidence of any target fault type by combining the feature conformity and feature weight of each feature under any target fault type includes: According to the feature weight of each feature under any target fault type, the feature conformance of all features under any target fault type is weighted and summed to obtain the fault matching confidence of any target fault type.

10. The substation fault prediction method based on multi-level equipment collaboration according to claim 1, characterized in that: After matching each feature data in the device real-time status feature vector using the fault feature range corresponding to each feature under each preset fault type, the method further includes: If a preset fault type is matched, the matched preset fault type is used as the final fault type of the target device in the current cycle, and the fault matching confidence of the matched preset fault type is obtained.

Images