Real-time monitoring method and system for safe operation state of power grid equipment

Abstract

The invention provides a real-time monitoring method and system for a safe operation state of power grid equipment, and relates to the technical field of equipment safety, and the method comprises the steps: obtaining a predicted residual life YS of target power grid equipment according to a state feature vector ZT and a residual life prediction model of the target power grid equipment; obtaining a first state type corresponding to the YS according to the YS and a first preset life-state type mapping table; obtaining a fusion feature vector RT according to a time domain feature vector YT and a frequency domain feature vector ET corresponding to the target power grid equipment in the target time window, a time domain fusion weight YTQ corresponding to the first state type and a frequency domain fusion weight ETQ corresponding to the current state type; and inputting the RT into the anomaly prediction model to obtain the current safe operation state of the target power grid equipment. According to the invention, stable operation of power grid equipment is powerfully guaranteed, and negative effects on social economy and daily life are reduced.

Description

Background Art

[0002] With the increasing requirements for the stability and reliability of power supply in modern society, the safe operation of power grid equipment is of vital importance. The stable operation of power grid equipment is directly related to the reliability of the entire power system, power quality, and power supply continuity, and plays a key role in ensuring the normal operation of social economy and people's daily lives.

[0003] In the field of analyzing the safe operation status of power grid equipment, early research mainly focused on analysis methods based on time-domain data. Time-domain analysis infers the operation status of power grid equipment by directly observing and processing signal data that changes over time, such as the real-time fluctuations of parameters like current and voltage. However, this analysis method based solely on time-domain data has obvious limitations. Time-domain data can only reflect the change characteristics of signals in the time dimension, with insufficient data dimensions and unable to comprehensively capture the complex information during the operation of power grid equipment. For example, some potential faults or abnormalities inside the equipment may not be clearly reflected in the intuitive changes of time-domain signals, making it difficult to accurately analyze the safe operation status of the equipment based only on time-domain data and unable to detect some hidden fault hazards in a timely manner, thus affecting the effective evaluation of the safe operation status of power grid equipment.

[0004] To overcome the deficiencies of time-domain analysis, researchers began to introduce frequency-domain analysis into the monitoring of the safe operation status of power grid equipment. Frequency-domain analysis uses mathematical means such as Fourier transform on time-domain signals to convert the signals from the time domain to the frequency domain and reveals the distribution characteristics of signals at different frequency components. Combining time-domain data and frequency-domain data can obtain information on the operation status of power grid equipment from multiple dimensions and theoretically can provide a more comprehensive and in-depth understanding of the equipment's operation conditions. For example, certain faults may produce unique changes in frequency components within a specific frequency range, and these subtle changes can be captured through frequency-domain analysis, providing more abundant clues for equipment fault diagnosis.

[0005] However, the existing technologies still face major challenges when using time-domain data and frequency-domain data to monitor the safe operation status of power grid equipment. When power grid equipment is in different operation states, there are significant differences in the influence degrees of each time-domain data and frequency-domain data on the prediction results of the safe operation status. Different types of equipment faults may have different correlation strengths with specific time-domain or frequency-domain characteristics. Most of the existing technologies fail to fully consider these differences in data influence degrees, resulting in a large deviation between the prediction results and the actual operation status and being unable to accurately judge the safe operation status of power grid equipment. Summary of the Invention

[0006] To address the above technical problems, the present application provides a real-time monitoring method and system for the safe operation status of power grid equipment, which at least partially solves the problems existing in the prior art.

[0007] In the first aspect of the present application, a real-time monitoring method for the safe operation state of power grid equipment is provided. The method includes:

[0008] S100. According to the state feature vector ZT of the target power grid equipment and the remaining life prediction model, the predicted remaining life YS of the target power grid equipment is obtained; wherein, ZT is used to characterize the equipment information, historical operation state, current operation state, and environmental state of the target power grid equipment;

[0009] S200. According to YS and the first preset life-state type mapping table, the first state type corresponding to YS is obtained; wherein, the first preset life-state type mapping table includes several remaining life intervals and the first state type corresponding to each remaining life interval; each first state type has a corresponding first influence factor weight list; wherein, the first influence factor weight list includes a time-domain feature fusion weight and a frequency-domain feature fusion weight;

[0010] S300. According to the time-domain feature vector YT, frequency-domain feature vector ET, time-domain fusion weight YTQ corresponding to the first state type, and frequency-domain fusion weight ETQ corresponding to the current state type of the target power grid equipment within the target time window, the fusion feature vector RT is obtained; wherein, ET is obtained by performing short-time Fourier transform on YT and then performing feature extraction;

[0011] S400. The RT is input into the anomaly prediction model to obtain the current safe operation state of the target power grid equipment.

[0012] In the second aspect of the present application, a real-time monitoring system for the safe operation state of power grid equipment is provided. The system includes:

[0013] A prediction unit, configured to obtain the predicted remaining life YS of the target power grid equipment according to the state feature vector ZT of the target power grid equipment and the remaining life prediction model; wherein, ZT is used to characterize the equipment information, historical operation state, current operation state, and environmental state of the target power grid equipment;

[0014] A state type determination unit, configured to obtain the first state type corresponding to YS according to YS and the first preset life-state type mapping table; wherein, the first preset life-state type mapping table includes several remaining life intervals and the first state type corresponding to each remaining life interval; each first state type has a corresponding first influence factor weight list; wherein, the first influence factor weight list includes a time-domain feature fusion weight and a frequency-domain feature fusion weight;

[0015] A fusion vector determination unit, configured to obtain a fusion feature vector RT according to a time-domain feature vector YT, a frequency-domain feature vector ET, a time-domain fusion weight YTQ corresponding to a first state type, and a frequency-domain fusion weight ETQ corresponding to a current state type of a target power grid device within a target time window; wherein, ET is obtained by performing feature extraction after performing short-time Fourier transform on YT;

[0016] An operating state determination unit, configured to input RT into an anomaly prediction model to obtain the current safe operating state of the target power grid device.

[0017] This application has at least the following beneficial effects:

[0018] By comprehensively considering various aspects of information of the target power grid device, this application realizes accurate analysis and judgment of its safe operating state. It uses a state feature vector including device information, operating state, and environmental state, combines a remaining life prediction model to estimate the remaining life of the device, and determines the corresponding state type and influence factor weights according to a preset mapping table, which covers time-domain and frequency-domain feature fusion weights. Then, it fuses the time-domain and frequency-domain feature vectors and dynamically adjusts the weights according to the device state to generate a fusion feature vector that can accurately reflect the current operating characteristics of the device. Finally, this vector is input into the anomaly prediction model to obtain the current safe operating state of the device. This comprehensive and targeted analysis method effectively overcomes the deficiencies of traditional methods that only rely on time-domain data or simply combine time-frequency domain data, fully considers the differences in the influence degrees of different types of data on the anomaly prediction of power grid devices under different remaining lives, significantly improves the prediction accuracy, effectively guarantees the stable operation of power grid devices, and reduces the negative impacts on social economy and daily life. Description of the Drawings

[0019] In order to more clearly illustrate the technical solutions in the embodiments of this application, the following will briefly introduce the drawings required for description in the embodiments. Obviously, the following drawings are only some embodiments of this application. For those of ordinary skill in the art, other drawings can be obtained without creative efforts based on these drawings.

[0020] Figure 1 It is a flowchart of a method for real-time monitoring of the safe operating state of a power grid device provided by an embodiment of this application;

[0021] Figure 2 It is a structural block diagram of a system for real-time monitoring of the safe operating state of a power grid device provided by an embodiment of this application. Detailed Embodiments

[0022] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present application without creative efforts shall fall within the protection scope of the present application.

[0023] It should be noted that the terms "first", "second", etc. in the specification and claims of the present application and the above-mentioned drawings are used to distinguish similar objects, and do not necessarily need to describe a specific order or sequence. It should be understood that such used data may be interchanged under appropriate circumstances so that the embodiments of the present application described herein can be implemented in an order different from those illustrated or described herein. In addition, the terms "comprising" and "having" and any variations thereof are intended to cover non-exclusive inclusion. For example, a process, method, system, product or server including a series of steps or units does not necessarily have to be limited to those steps or units clearly listed, but may include other steps or units not clearly listed or inherent to these processes, methods, products or devices.

[0024] It should be noted that the following describes various aspects of embodiments within the scope of the appended claims. It should be apparent that the aspects described herein can be embodied in a wide variety of forms, and any specific structure and / or function described herein is merely illustrative. Based on the present application, those skilled in the art should understand that one aspect described herein can be implemented independently of any other aspect, and two or more of these aspects can be combined in various ways. For example, any number of aspects described herein can be used to implement a device and / or practice a method. Additionally, this device and / or this method can be implemented using other structures and / or functionality in addition to one or more of the aspects described herein.

[0025] Please refer to Figure 1 As shown, an embodiment of the present application provides a method for real-time monitoring of the safe operation state of a power grid device, and the method includes:

[0026] S100. According to the state feature vector ZT of the target power grid device and the remaining life prediction model, obtain the predicted remaining life YS of the target power grid device; where ZT is used to characterize the device information, historical operation state, current operation state, and environmental state of the target power grid device.

[0027] Specifically, the target power grid device is one of several power grid devices, and the power grid devices can be: generators, transformers, distribution boxes, capacitors, etc. Among them, the state feature vector ZT of the target power grid device is used to characterize the device information, historical operating state, current operating state and environmental state of the target power grid device, and the above aspects all have different degrees of influence on the remaining life of the target power grid device. It should be noted that each target power grid device also has a corresponding standard service duration and used duration. However, due to the influence of various factors, the remaining life of the power grid device cannot be directly obtained by subtracting the used duration from the standard service duration. Therefore, in this implementation, the influence of the device information, historical operating state, current operating state and environmental state of the target power grid device on the remaining life is comprehensively considered, and then the predicted remaining life of the target power grid device is determined according to the pre-trained remaining life prediction model. Here, the predicted remaining life represents the predicted remaining working duration of the corresponding power grid device.

[0028] S200. According to YS and the first preset life-state type mapping table, obtain the first state type corresponding to YS; wherein, the first preset life-state type mapping table includes several remaining life intervals and the first state type corresponding to each remaining life interval; each first state type has a corresponding first influence factor weight list; wherein, the first influence factor weight list includes a time-domain feature fusion weight and a frequency-domain feature fusion weight.

[0029] Specifically, since different types of data have different degrees of influence on the abnormal prediction of power grid devices under different remaining life conditions. For example, during the mid-term operation of the device, the operating environment data and conventional operating state data of the device are more critical for abnormal prediction. Environmental data such as temperature, humidity, altitude, etc. will affect the heat dissipation, insulation performance, etc. of the device. For example, a high-temperature environment may accelerate the aging of the device's insulation material. By long-term monitoring of the relationship between the environmental temperature and the device's operating temperature, early signs of abnormal heat dissipation of the device can be detected. The change trends of conventional operating state data such as load current and voltage can reflect whether the device is overloaded for a long time, thereby predicting abnormalities such as overheating and insulation aging that the device may occur. In the later stage (when the remaining life is short), at this time, its aging intensifies and significant changes occur in its physical and chemical properties. At this time, the aging-related data of the device, such as the aging degree index of the insulation material (such as an increase in the dielectric loss factor, a decrease in the insulation resistance, etc.), the wear and corrosion data of metal components (such as the rust degree of the transformer core, the wear amount of the switch contact, etc.) have a higher degree of influence on abnormal prediction. These data can directly reflect the severity of the device's aging and accurately predict the upcoming failures of the device, such as insulation breakdown and poor contact.

[0030] Therefore, in this embodiment, the first preset life - status type mapping table includes several remaining life intervals and the first status type corresponding to each remaining life interval. According to the remaining life interval to which the predicted remaining life of the target power grid device belongs, the corresponding first status type is determined for the target power grid device. Here, each first status type has a corresponding first influence factor weight list, and the target power grid device has a corresponding first influence factor weight list. The first influence factor weight list includes a time - domain feature fusion weight and a frequency - domain feature fusion weight. That is, according to the current predicted remaining life of the target power grid device, the importance degree of each type of data for subsequent anomaly prediction is determined. If the time - domain feature fusion weight is larger, it indicates that the time - domain data has a greater impact on the result of subsequent anomaly prediction. Conversely, if the time - domain feature fusion weight is smaller, it indicates that the time - domain data has a smaller impact on the result of subsequent anomaly prediction.

[0031] S300. According to the time - domain feature vector YT, frequency - domain feature vector ET, time - domain fusion weight YTQ corresponding to the first status type, and frequency - domain fusion weight ETQ corresponding to the current status type of the target power grid device within the target time window, obtain the fusion feature vector RT; where ET is obtained by performing short - time Fourier transform on YT and then extracting features.

[0032] S400. Input RT into the anomaly prediction model to obtain the current safe operation state of the target power grid device.

[0033] Specifically, when fusing time - domain data and frequency - domain data, according to the determined time - domain fusion weight and frequency - domain fusion weight, corresponding weights are assigned during feature fusion, thereby adjusting the influence degrees of time - domain data and frequency - domain data on the anomaly prediction result. This makes the judgment result of whether the current safe operation state of the target power grid device is abnormal more accurate.

[0034] In an exemplary embodiment of the present application, step S100 includes:

[0035] S110. Obtain the state feature vector ZT=(ZTJ, ZTL, ZTJ, ZTH) of the target power grid device; where ZTJ is the device information list of the target power grid device; ZTL is the historical operation data list of the target power grid device within the historical time window; ZTJ is the set of operation data list of the target power grid device within the preset time window; the end time of the preset time window is the current time; the end time of the historical time window is the start time of the preset time window; ZTH is the environmental data list of the environment where the target power grid device is located.

[0036] Specifically, ZTJ may include:

[0037] Equipment type coding: Digitally encode different types of power grid equipment (such as transformers, circuit breakers, insulators, etc.) to identify equipment types. Due to different working principles and structural characteristics of different types of equipment, there are differences in life - influencing factors and aging modes. For example, transformers mainly face insulation aging problems, while for circuit breakers, the key lies in contact wear, etc. Through coding, the remaining life prediction model can learn specific life prediction modes for different equipment types.

[0038] Manufacturing year of the equipment: Reflects the technical level and manufacturing process background when the equipment is put into use. Newly manufactured equipment may adopt more advanced materials and processes, and theoretically has better performance and a longer expected life. For example, transformers manufactured in recent years have continuously improved in insulation materials and heat dissipation design, and may age more slowly compared to earlier products.

[0039] Rated capacity of the equipment: Reflects the designed operating capacity of the equipment. For equipment such as transformers, the larger the rated capacity, the higher the electromagnetic stress and thermal load it may bear during operation, which has a greater impact on life. Under the same operating conditions, large - capacity transformers generate more heat in the windings and iron cores, accelerating insulation aging. Therefore, the rated capacity is an important factor affecting life.

[0040] ZTJ may also include other basic equipment information related to life changes, which will not be elaborated here.

[0041] ZTL may include: cumulative operating time, average load rate, load volatility, number of over - voltages, etc.;

[0042] Among them, cumulative operating time: The longer the equipment operates, the more fully each component experiences wear and aging processes, which is a basic indicator for measuring the aging degree of the equipment.

[0043] Average load rate: The calculation method is the ratio of the average of the actual operating load of the equipment to the rated load. Long - term operation at a high load rate will increase the equipment temperature, accelerating insulation aging and mechanical component wear. For example, when a transformer operates at a high load rate for a long time, the winding current increases, generating more heat, and the insulation material ages faster, affecting the life of the transformer.

[0044] Load volatility: Reflects the severity of the equipment load changing over time. Frequent load fluctuations will cause additional stress shocks to the equipment. For example, during the frequent start - up and stop processes of an electric motor, large impact currents and mechanical stresses will be generated, affecting the life of the motor windings and bearings. The greater the load volatility, the more serious the damage to the equipment may be.

[0045] Number of over - voltages: The cumulative number of times the equipment has suffered over - voltages during operation. Over - voltages may cause insulation breakdown of the equipment. The instantaneous high - voltage impact will cause irreversible damage to the insulation layer. Each over - voltage may become an accelerating factor for equipment insulation aging and affect the equipment life.

[0046] ZTJ may include:

[0047] List of operating temperatures within a preset time window: The real - time temperature during equipment operation. Excessive temperature will accelerate the aging process of the equipment, such as accelerating the thermal aging of insulating materials and the oxidation corrosion of metal components. For example, too high a transformer oil temperature will accelerate the aging speed of insulating oil and insulating paper. The real - time temperature reflects the current thermal stress situation of the equipment.

[0048] List of operating temperatures within a preset time window: Reflects the current actual operating current magnitude of the equipment, which is directly related to the load state of the equipment. Through the current load current, it can be judged whether the equipment is in an over - load or abnormal operating state. Over - load operation will increase the loss and heat generation of the equipment and have a negative impact on the life.

[0049] List of partial discharge amounts within a preset time window: For high - voltage electrical equipment, partial discharge is an important sign of insulation deterioration. Partial discharge will gradually corrode the insulating material, reduce the insulation performance, and may eventually lead to insulation breakdown. The larger the detected partial discharge amount, the more serious the potential problem with the equipment insulation and the greater the threat to the equipment life.

[0050] As an example: The preset time window can be 1 hour.

[0051] ZTH may include:

[0052] Operating environment humidity: A high - humidity environment is likely to cause corrosion of the equipment and a decline in insulation performance. For example, too high humidity inside a switchgear cabinet will cause metal parts to rust and affect the reliability of electrical connections. At the same time, a humid environment may also cause insulating materials to get damp, reduce the insulation resistance, increase the risk of leakage and short - circuit, and shorten the equipment life. The environmental humidity reflects an important aspect of the impact of the equipment's environment on its life.

[0053] Degree of pollution index. Used to measure the pollution situation of the environment where the equipment is located. Dust, smoke, industrial pollutants, etc. in the air will adhere to the surface of the equipment and affect its insulation performance. For example, in coastal areas or areas with severe industrial pollution, the surface of insulators is prone to accumulate pollution. In humid weather, it may form a conductive channel, trigger flashover accidents, and accelerate the aging of insulators. The higher the degree of pollution index, the greater the risk of aging and failure of the equipment due to pollution.

[0054] And perform pre - processing such as data cleaning, data standardization, and data enhancement on the above data to obtain ZT.

[0055] S120, input the ZT into the remaining life prediction model to obtain the predicted remaining life YS of the target grid equipment.

[0056] Among them, the remaining life prediction model is a trained model. As an example, it can be a physical model, a statistical model, or a model based on machine learning and deep learning.

[0057] In an exemplary embodiment of the present application, step S300 includes:

[0058] S310, obtain the time-domain feature vector YT = (YT1, YT2,..., YT i ,..., YT n ) corresponding to the target grid equipment within the target time window; i = 1, 2,..., n; where n is the number of state impact factors corresponding to the target grid equipment; YT i is the time-domain data list corresponding to the i-th state impact factor of the target grid equipment within the target time window; YT i =(YT i，1 , YT i，2 ,..., YT i，j ,..., YT i，m ); j = 1, 2,..., m; m is the number of sampling time points within the target time window; the time interval between any two sampling time points is the same; YT i，j is the amplitude corresponding to the i-th state impact factor at the j-th sampling time point;

[0059] S320, according to YT, obtain the frequency-domain feature vector PT = (PT1, PT2,..., PT i ,..., PT n ) corresponding to the target grid equipment within the target time window; where PT i is the frequency-domain data matrix corresponding to the i-th state impact factor of the target grid equipment within the target time window;

[0060]

[0061] p = 1, 2,..., q; q is the number of preset frequency components; PT i，1，p is the amplitude list corresponding to the p-th frequency component of the i-th state impact factor of the target grid equipment within the first target sub-time period; x = 1, 2,..., y; y is the number of consecutive target sub-time periods included within the target time window; the duration of any two adjacent target sub-time periods is equal; PT i，x，p is the amplitude list corresponding to the p-th frequency component of the i-th state impact factor of the target grid equipment within the x-th target sub-time period.

[0062] Specifically, due to the complex time-varying characteristics of power grid equipment data, for example, during peak and off-peak electricity consumption periods, parameters such as the voltage and current of the equipment will have obvious fluctuations. In addition, the occurrence of equipment failures is also a dynamic process. From the gestation to the development of the failure, the relevant electrical quantities will change continuously over time. Moreover, the signals generated during the operation of power grid equipment contain various frequency components. During normal operation, the signals are mainly dominated by the power frequency, but when a failure or abnormal situation occurs, different frequency components such as various harmonics and transient signals will be generated. For example, when a short-circuit fault occurs in the power grid, high-frequency transient components will be generated. The traditional Fourier transform can only provide the overall frequency information of the signal and cannot know the time distribution of each frequency component. The short-time Fourier transform divides the data into multiple short time segments and performs Fourier transform on each short time segment respectively, which can simultaneously display the time and frequency information of the signal and form time-frequency data. In this way, the change of different frequency components in the power grid equipment data over time can be clearly observed, which helps to analyze the dynamic evolution of the equipment operation state.

[0063] S330. Extract features from the PT to obtain ET=(ET1, ET2,..., ET i ,..., ET n ); where ET i is the list of eigenvalue obtained by extracting features from the PT i .

[0064] Specifically, ET i is obtained through the following steps:

[0065] S331. Convert the PT i into a grayscale image and perform normalization processing on the pixel values to obtain a normalized image.

[0066] Specifically, in order to unify the pixel value ranges of different images, the image is first normalized so that the pixel values are in the range of [0,1].

[0067] S332. Process the normalized image according to a preset adaptive histogram algorithm and frequency domain filtering algorithm to obtain an intermediate image; where the adaptive histogram algorithm is used to perform overlapping block processing and dynamic threshold adjustment processing on the normalized image; the frequency domain filtering algorithm is used to perform multi-scale frequency domain filtering processing and phase retention processing on the normalized image processed by the adaptive histogram algorithm.

[0068] Specifically, traditional Contrast Limited Adaptive Histogram Equalization (CLAHE) divides an image into multiple non-overlapping small blocks and performs histogram equalization independently within each small block. However, this method may generate discontinuous artifacts at the boundaries of the small blocks. In this embodiment, overlapping block division and dynamic threshold adjustment are adopted for processing. The image is divided into multiple small blocks with overlapping regions. For example, the size of each small block is e×f pixels, and the number of overlapping pixels in adjacent pixel blocks is less than the minimum side length of each pixel block. This can make the transition between small blocks smoother and reduce boundary artifacts. The specific process is as follows: Starting from the upper left corner, according to the set block size and overlapping amount, the coordinates of all divided blocks are generated to obtain the coordinate set of all blocks. For each block in the coordinate set, multiple features of it are calculated respectively. These include but are not limited to statistical features such as mean, variance, skewness, kurtosis, and texture features such as contrast, correlation, energy, entropy, etc. based on the Gray Level Co-occurrence Matrix (GLCM). These features will comprehensively describe the image information within each block. When calculating features, special attention is paid to the edge regions of the block. Since the overlapping part is located at the block edge, an enhanced strategy is adopted for feature extraction of the edge region. For example, higher weights are assigned to edge pixels to calculate statistical features, or when calculating GLCM, the contribution of edge pixels to texture features is increased to highlight the information of the edge region, making the features of the overlapping part more representative. For the overlapping regions of adjacent blocks, the features are fused by weighted averaging. The determination of the weights is based on the distance from the pixels within the block to the block center. The closer the distance to the block center, the lower the weight; the farther the distance from the block center (i.e., the closer to the overlapping region), the higher the weight. For the feature value F of a certain pixel in the overlapping region, its fusion formula is where F k is the feature value of the pixel at the corresponding position in the adjacent blocks, w k is the weight calculated according to the distance, neighbors is the set of adjacent blocks related to the pixel in the current overlapping region, and the numerator part is the weighted sum of the feature values of the pixels at the corresponding positions in all adjacent blocks. This sum value synthesizes the contributions of all adjacent blocks to the feature of the pixel in the current overlapping region, and the feature values of the adjacent blocks with larger weights (i.e., the blocks closer to the overlapping region) account for a larger proportion in this sum value. The denominator part is the sum of the weight values of all adjacent blocks. The role of the denominator is to normalize the weighted sum of the numerator to ensure that the finally obtained fused feature value is within a reasonable range and avoid deviations in the fused feature value due to the sum of weights not being 1. By dividing the numerator by the denominator, the fused feature value can balance the contributions of each adjacent block and accurately reflect the comprehensive information of the pixel in the overlapping region in the overall image features.

[0069] Through the above fusion method, the features of all blocks are integrated to generate a global feature vector containing the information of the entire image. This feature vector not only contains the features of different local regions of the image, but also effectively avoids information loss through the fusion of overlapping regions, ensuring the integrity and continuity of the features.

[0070] In this embodiment, the weights are determined based on the distance from the pixel to the center of the block for fusion, which more reasonably considers the position information of the pixel within the block. Pixels closer to the center of the block contribute more to the features of this block, while pixels closer to the edge (overlapping region) contribute more to information fusion. This weighting method enables the fused features to retain the main information within the block and effectively integrate the information between adjacent blocks, improving the accuracy and comprehensiveness of the features.

[0071] Furthermore, when calculating the histogram within each small block, a dynamic threshold is introduced. Traditional CLAHE usually uses a fixed clipping threshold to limit the distribution of the histogram to avoid over-enhancement. The improved method dynamically adjusts the clipping threshold according to the statistical characteristics (such as standard deviation) of the pixel values within the small block. Specifically, for small blocks with a larger standard deviation, the clipping threshold is appropriately increased to allow more contrast enhancement; for small blocks with a smaller standard deviation, the clipping threshold is decreased to avoid noise amplification caused by over-enhancement.

[0072] After the histogram equalization of each small block is completed, the bilinear interpolation method is used to fuse the pixel values in the overlapping region to obtain the final adaptive histogram equalized image, that is, the intermediate image.

[0073] The intermediate image is decomposed at multiple scales. For example, the wavelet transform is used to decompose the image into sub-bands of different scales and directions. For each sub-band, different frequency domain filters are designed respectively. In the low-frequency sub-band, a wider band-pass filter is used to retain the main structural information of the image; in the high-frequency sub-band, a narrower band-pass filter is used to remove high-frequency noise while trying to retain the detailed information of the image. During the frequency domain filtering process, only the amplitude spectrum of the image is filtered, while the phase spectrum remains unchanged. Because the phase information is very important for the structure and feature recovery of the image, retaining the phase spectrum can avoid blurring and distortion of the filtered image.

[0074] S333, after performing contrast fine-tuning on the intermediate image, the target image is obtained.

[0075] S334, feature extraction is performed on the target image to obtain ET i 。

[0076] Specifically, the Simple Linear Iterative Clustering (SLIC) algorithm is used to segment the image into multiple superpixel regions. By clustering in the color-space joint domain, the SLIC algorithm can generate compact and regular superpixels, which can be regarded as image regions with similar features. For each superpixel region, statistical features such as mean, standard deviation, skewness, and kurtosis are calculated, as well as shape features such as area, perimeter, and circularity of the region. At the same time, the Local Binary Pattern (LBP) features of the region are extracted to describe the texture information of the region. Finally, a Multi-Scale Self-Attention Convolutional Network (MSA-CNN) is designed. The network contains convolutional layers of multiple different scales, and each scale of convolutional layer can capture features of different sizes. After the convolutional layer, a self-attention mechanism is introduced. By calculating the correlation between each position in the feature map and other positions, the weights of the feature map are adaptively adjusted to enhance the attention to important features. The image processed by superpixel segmentation is input into the MSA-CNN for feature extraction. The feature vector output by the network contains the global and local feature information of the image at different scales. Finally, feature fusion and dimensionality reduction are performed to obtain the target image.

[0077] In this embodiment, through overlapping block division and dynamic threshold adjustment, the discontinuity problem of traditional CLAHE at the small block boundaries is overcome, and the enhancement degree can be dynamically adjusted according to the statistical characteristics of the small blocks, avoiding over-enhancement or under-enhancement. The multi-scale frequency domain filtering combined with the phase retention technology can better retain the high-frequency detail information of the image while removing noise, avoiding the image blurring and distortion problems that may be caused by traditional frequency domain filtering. Finally, the features extracted from the frequency domain data can reflect the characteristics of the frequency domain data itself and the changes of the data as much as possible.

[0078] S340. Obtain RT according to YT, ET, YTQ, and ETQ, where RT meets the following conditions:

[0079] RT = (YT × YTQ, ET × ETQ).

[0080] In this embodiment, when fusing the time-domain data and the frequency-domain data, according to the determined time-domain fusion weight and frequency-domain fusion weight mentioned above, corresponding weights are assigned during feature fusion, so as to adjust the influence degrees of the time-domain data and the frequency-domain data on the abnormal prediction result. This makes the judgment result of whether the current safe operation state of the target power grid equipment is abnormal more accurate.

[0081] Moreover, during image processing before feature extraction from time-domain data, through overlapping block division and dynamic threshold adjustment, the discontinuity problem of traditional CLAHE at small block boundaries is overcome, and the enhancement degree can be dynamically adjusted according to the statistical characteristics of small blocks, avoiding over-enhancement or under-enhancement. The multi-scale frequency-domain filtering combined with the phase-preserving technique better preserves the high-frequency detail information of the image while removing noise, avoiding the image blurring and distortion problems that may be caused by traditional frequency-domain filtering. Finally, the features extracted from the frequency-domain data can reflect the characteristics of the frequency-domain data itself and the changes in the data as much as possible.

[0082] In an exemplary embodiment of the present application, after step S100, the method further includes:

[0083] S500, according to YS and the second preset life-state type mapping table, to obtain the second state type corresponding to YS; wherein, the second preset life-state type mapping table includes several remaining life intervals and the second state type corresponding to each remaining life interval; each second state type has a corresponding second influence factor weight list; wherein, the second influence factor weight list includes the weight corresponding to each state influence factor among n state influence factors;

[0084] S600, according to the time-domain feature vector YT, frequency-domain feature vector ET corresponding to the target grid device within the target time window, and the weight corresponding to each state influence factor among the n state influence factors corresponding to the current state type, obtain the fused feature vector RT;

[0085] S700, input RT into the anomaly prediction model to obtain the current safe operating state of the target grid device.

[0086] Wherein, RT meets the following condition: RT = (α1 × YT1, α2 × YT2,..., α i × YT i ,..., α n × YT n , α1 × ET1, α2 × ET2,..., α i × ET i ,..., α n × ET n ); wherein, α1, α2,..., α i ,..., α n are the weights corresponding to each state influence factor in the second influence factor weight list.

[0087] In this embodiment, when fusing time-domain data and frequency-domain data, according to the weights corresponding to each state influence factor in the aforementioned second influence factor weight list, here, YS also has a corresponding second influence factor weight list, that is, when the predicted remaining life of the target power grid equipment is different, the influence degree of each state influence factor on the result of abnormal prediction is different. Therefore, in this embodiment, the same weights are assigned to the same state influence factors in the time-domain data and the frequency-domain data, so that the finally obtained abnormal prediction result is more accurate.

[0088] Please refer to Figure 2 As shown, an embodiment of the present application provides a real-time monitoring system 100 for the safe operation state of a power grid device. The system includes:

[0089] A prediction unit 110, configured to obtain the predicted remaining life YS of the target power grid device according to the state feature vector ZT of the target power grid device and the remaining life prediction model; where ZT is used to characterize the device information, historical operation state, current operation state, and environmental state of the target power grid device.

[0090] A state type determination unit 120, configured to obtain the first state type corresponding to YS according to YS and the first preset life-state type mapping table; where the first preset life-state type mapping table includes several remaining life intervals and the first state type corresponding to each remaining life interval; each first state type has a corresponding first influence factor weight list; where the first influence factor weight list includes a time-domain feature fusion weight and a frequency-domain feature fusion weight.

[0091] A fusion vector determination unit 130, configured to obtain a fusion feature vector RT according to the time-domain feature vector YT, frequency-domain feature vector ET, time-domain fusion weight YTQ corresponding to the first state type, and frequency-domain fusion weight ETQ corresponding to the current state type of the target power grid device within the target time window; where ET is obtained by performing short-time Fourier transform on YT and then performing feature extraction.

[0092] An operation state determination unit 140, configured to input RT into an abnormal prediction model to obtain the current safe operation state of the target power grid device.

[0093] An embodiment of the present application also provides a computer program product, which includes program code. When the program product runs on an electronic device, the program code is used to cause the electronic device to execute the steps in the methods according to various exemplary embodiments of the present application described above in this specification.

[0094] In addition, although the steps of the methods in this application are described in a specific order in the accompanying drawings, this does not require or imply that these steps must be performed in that specific order, or that all the steps shown must be performed to achieve the desired result. Additionally or alternatively, certain steps may be omitted, multiple steps may be combined into one step for execution, and / or one step may be decomposed into multiple steps for execution, etc.

[0095] From the description of the above embodiments, those skilled in the art can easily understand that the exemplary embodiments described herein can be implemented by software, or by a combination of software and necessary hardware. Therefore, the technical solutions according to the embodiments of this application can be embodied in the form of a software product, which can be stored in a non-volatile storage medium (which can be a CD-ROM, USB flash drive, mobile hard disk, etc.) or on a network, including several instructions to enable a computing device (which can be a personal computer, server, mobile terminal, or network device, etc.) to execute the methods according to the embodiments of this application.

[0096] In an exemplary embodiment of this application, an electronic device capable of implementing the above method is also provided.

[0097] Those skilled in the relevant technical field can understand that various aspects of this application can be implemented as a system, method, or program product. Therefore, various aspects of this application can be specifically implemented in the following forms, namely: a complete hardware implementation, a complete software implementation (including firmware, microcode, etc.), or an implementation combining hardware and software aspects, which can be collectively referred to as "circuitry", "module", or "system" here.

[0098] The electronic device according to this embodiment of this application. The electronic device is merely an example and should not impose any limitations on the functions and usage scope of the embodiments of this application.

[0099] The electronic device is presented in the form of a general-purpose computing device. The components of the electronic device may include, but are not limited to: at least one of the above-mentioned processors, at least one of the above-mentioned memories, and a bus connecting different system components (including the memory and the processor).

[0100] Among them, the memory stores program code, and the program code can be executed by the processor, so that the processor executes the steps according to various exemplary embodiments of this application described in the above "Exemplary Method" section of this specification.

[0101] The memory may include a readable medium in the form of a volatile memory, such as a random access memory (RAM) and / or a cache memory, and may further include a read-only memory (ROM).

[0102] The storage may also include program / utility with a set (at least one) of program modules, such program modules including but not limited to: operating system, one or more application programs, other program modules, and program data, and implementation of network environment may be included in each or some combination of these examples.

[0103] The bus may represent one or more of several types of bus structures, including a memory bus or memory controller, a peripheral bus, a graphics acceleration port, a processor, or a local bus using any of the various bus structures.

[0104] The electronic device may also communicate with one or more external devices (such as a keyboard, a pointing device, a Bluetooth device, etc.), and may also communicate with one or more devices that enable a user to interact with the electronic device, and / or communicate with any device that enables the electronic device to communicate with one or more other computing devices (such as a router, a modem, etc.). Such communication may be carried out through an input / output (I / O) interface. Also, the electronic device may communicate with one or more networks (such as a local area network (LAN), a wide area network (WAN), and / or a public network, such as the Internet) through a network adapter. As shown in the figure, the network adapter communicates with other modules of the electronic device through the bus. It should be understood that, although not shown in the figure, other hardware and / or software modules may be used in combination with the electronic device, including but not limited to: microcode, device drivers, redundant processors, external disk drive arrays, RAID systems, tape drives, and data backup storage systems, etc.

[0105] Through the description of the above embodiments, those skilled in the art can easily understand that the exemplary embodiments described herein can be implemented by software, or by a combination of software and necessary hardware. Therefore, the technical solutions according to the embodiments of the present application can be embodied in the form of a software product, which can be stored in a non-volatile storage medium (which may be a CD-ROM, a USB flash drive, a mobile hard disk, etc.) or on a network, and includes several instructions to enable a computing device (which may be a personal computer, a server, a terminal device, or a network device, etc.) to execute the method according to the embodiments of the present application.

[0106] In an exemplary embodiment of the present application, there is also provided a computer-readable storage medium, on which a program product capable of implementing the above method of this specification is stored. In some possible implementation manners, various aspects of the present application may also be implemented in the form of a program product, which includes program code, and when the program product runs on a terminal device, the program code is used to enable the terminal device to execute the steps according to various exemplary embodiments of the present application described in the above "Exemplary Method" section of this specification.

[0107] The program product may employ any combination of one or more readable media. The readable media may be a readable signal medium or a readable storage medium. The readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination of the foregoing. More specific examples (a non-exhaustive list) of the readable storage medium include: an electrical connection having one or more wires, a portable disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or Flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0108] The computer-readable signal medium may include a data signal propagated in a baseband or as part of a carrier wave, which carries the readable program code. Such a propagated data signal may take many forms, including but not limited to an electromagnetic signal, an optical signal, or any suitable combination of the foregoing. The readable signal medium may also be any readable medium other than the readable storage medium, which can send, propagate, or transmit a program for use by or in connection with an instruction execution system, apparatus, or device.

[0109] The program code contained on the readable medium may be transmitted using any appropriate medium, including but not limited to wireless, wireline, optical fiber cable, RF, etc., or any suitable combination of the foregoing.

[0110] The program code for performing the operations of this application may be written in any combination of one or more programming languages. The programming languages include object-oriented programming languages such as Java, C++, etc., and also include conventional procedural programming languages such as the "C" language or similar programming languages. The program code may be executed entirely on the user's computing device, partially on the user's device, executed as a stand-alone software package, partially on the user's computing device and partially on a remote computing device, or entirely on the remote computing device or server. In the case of a remote computing device, the remote computing device may be connected to the user's computing device through any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computing device (e.g., by using an Internet service provider to connect through the Internet).

[0111] In addition, the above-mentioned drawings are only schematic illustrations of the processes included in the method according to the exemplary embodiments of the present application, rather than for limiting purposes. It is easy to understand that the processes shown in the above-mentioned drawings do not indicate or limit the time sequence of these processes. Additionally, it is also easy to understand that these processes may be executed synchronously or asynchronously, for example, in multiple modules.

[0112] It should be noted that although several modules or units of the device for action execution are mentioned in the above detailed description, this division is not mandatory. In fact, according to the embodiments of the present application, the features and functions of the two or more modules or units described above can be embodied in one module or unit. Conversely, the features and functions of one module or unit described above can be further divided and embodied by multiple modules or units.

[0113] The above is only the specific embodiment of the present application, but the protection scope of the present application is not limited thereto. Any changes or substitutions that can be easily thought of by those skilled in the art within the technical scope disclosed by the present application should be covered by the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

Claims

1. A real-time monitoring method for the safe operation state of power grid equipment, characterized in that, The method includes: S100. Obtain the predicted remaining life YS of the target power grid device according to the state feature vector ZT of the target power grid device and the remaining life prediction model, where ZT is used to characterize the device information, historical operation status, current operation status, and environmental status of the target power grid device. S200. Obtain the first state type corresponding to YS according to YS and the first preset life - state type mapping table. The first preset life - state type mapping table includes several remaining life intervals and the first state type corresponding to each remaining life interval. Each first state type has a corresponding first influence factor weight list, where the first influence factor weight list includes a time - domain feature fusion weight and a frequency - domain feature fusion weight. S300. Obtain the fusion feature vector RT according to the time - domain feature vector YT, frequency - domain feature vector ET, time - domain fusion weight YTQ corresponding to the first state type, and frequency - domain fusion weight ETQ corresponding to the current state type of the target power grid device within the target time window. ET is obtained by performing short - time Fourier transform on YT and then extracting features. S400. Input RT into the anomaly prediction model to obtain the current safe operation state of the target power grid device.

2. The real-time monitoring method for the safe operation state of power grid equipment according to claim 1, characterized in that Step S100 includes: S110. Obtain the state feature vector ZT=(ZTJ, ZTL, ZTJ, ZTH) of the target power grid device. Here, ZTJ is the device information list of the target power grid device; ZTL is the historical operation data list of the target power grid device within the historical time window; ZTJ is the set of operation data list sets of the target power grid device within the preset time window. The end time of the preset time window is the current time, and the end time of the historical time window is the start time of the preset time window; ZTH is the environmental data list of the environment where the target power grid device is located. S120. Input ZT into the remaining life prediction model to obtain the predicted remaining life YS of the target power grid device.

3. The real-time monitoring method for the safe operation state of power grid equipment according to claim 1, characterized in that, Step S300 includes: S310, obtain the time-domain feature vector YT = (YT1, YT2,..., YT i ,..., YT n ); i = 1, 2,..., n; where n is the number of state influence factors corresponding to the target power grid device; YT i is the time-domain data list corresponding to the i-th state influence factor of the target power grid device within the target time window; YT i = (YT i，1 , YT i，2 ,..., YT i，j ,..., YT i，m ); j = 1, 2,..., m; m is the number of sampling time points within the target time window; the time interval between any two sampling time points is the same; YT i，j is the amplitude corresponding to the i-th state influence factor at the j-th sampling time point; S320. According to YT, obtain the frequency-domain feature vector PT=(PT1, PT2, …, PT i , …, PT n ) corresponding to the target power grid device within the target time window; where PT i is the frequency-domain data matrix corresponding to the i-th state influence factor corresponding to the target power grid device within the target time window; p = 1, 2, …, q; q is the number of preset frequency components; PT i，1，p is the amplitude list corresponding to the p-th frequency component of the i-th state influence factor corresponding to the target power grid device in the first target sub-time period; x = 1, 2, …, y; y is the number of consecutive target sub-time periods included in the target time window; the durations of any two adjacent target sub-time periods are equal; PT i，x，p is the amplitude list corresponding to the p-th frequency component of the i-th state influence factor corresponding to the target power grid device in the x-th target sub-time period; S330, perform feature extraction on PT to obtain ET=(ET1, ET2,..., ET i ,..., ET n ); where ET i is the list of feature values obtained by performing feature extraction on PT i ; S340. Obtain RT according to YT, ET, YTQ, and ETQ, where RT meets the following condition: RT=(YT×YTQ, ET×ETQ).

4. The real-time monitoring method for the safe operation state of power grid equipment according to claim 3, characterized in that After step S100, the method further includes: S500. Obtain the second state type corresponding to YS according to YS and the second preset life - state type mapping table. The second preset life - state type mapping table includes several remaining life intervals and the second state type corresponding to each remaining life interval. Each second state type has a corresponding second influence factor weight list, where the second influence factor weight list includes the weight corresponding to each of the n state influence factors. S600. Obtain the fusion feature vector RT according to the time - domain feature vector YT, frequency - domain feature vector ET, and the weight corresponding to each of the n state influence factors corresponding to the current state type of the target power grid device within the target time window. S700. Input RT into the anomaly prediction model to obtain the current safe operation state of the target power grid device.

5. The real-time monitoring method for the safe operation state of power grid equipment according to claim 4, wherein, RT meets the following conditions: RT = (α1 × YT1, α2 × YT2,..., α i × YT i ,..., α n × YT n , α1 × ET1, α2 × ET2,..., α i × ET i ,..., α n × ET n ); where α1, α2,..., α i ,..., α n are the weights corresponding to each state influence factor in the second influence factor weight list.

6. The real-time monitoring method for the safe operation state of power grid equipment according to claim 3, characterized in that, ET i Obtained by the following steps: S331, convert PT i to a grayscale image and normalize the pixel values to obtain a normalized image; S332. Process the normalized image according to a preset adaptive histogram algorithm and frequency domain filtering algorithm to obtain an intermediate image. Among them, the adaptive histogram algorithm is used to perform overlapping block processing and dynamic threshold adjustment processing on the normalized image. The frequency domain filtering algorithm is used to perform multi-scale frequency domain filtering processing and phase retention processing on the normalized image processed by the adaptive histogram algorithm. S333. Fine-tune the contrast of the intermediate image to obtain a target image. S334, perform feature extraction on the target image to obtain ET i .

7. The real-time monitoring method for the safe operation state of grid equipment according to claim 6, characterized in that, In the overlapping block processing, the number of overlapping pixels of adjacent pixel blocks is less than the minimum side length of each pixel block.

8. A real-time monitoring system for the safe operation status of power grid equipment, characterized in that, The system includes: A prediction unit, configured to obtain the predicted remaining life YS of the target power grid device according to the state feature vector ZT of the target power grid device and the remaining life prediction model. Among them, ZT is used to characterize the device information, historical operating state, current operating state, and environmental state of the target power grid device. A state type determination unit, configured to obtain the first state type corresponding to YS according to YS and a first preset life-state type mapping table. Among them, the first preset life-state type mapping table includes several remaining life intervals and the first state type corresponding to each remaining life interval. Each first state type has a corresponding first influence factor weight list. Among them, the first influence factor weight list includes a time domain feature fusion weight and a frequency domain feature fusion weight. A fusion vector determination unit, configured to obtain a fusion feature vector RT according to the time domain feature vector YT, frequency domain feature vector ET corresponding to the target power grid device within the target time window, the time domain fusion weight YTQ corresponding to the first state type, and the frequency domain fusion weight ETQ corresponding to the current state type. Among them, ET is obtained by performing feature extraction after performing short-time Fourier transform on YT. An operating state determination unit, configured to input RT into an anomaly prediction model to obtain the current safe operating state of the target power grid device.

Images