Transformer partial discharge multi-source positioning system based on digital twinning

Abstract

The invention belongs to the technical field of scene simulation, and particularly discloses a transformer partial discharge multi-source positioning system based on digital twinning. A generative adversarial network digital twinborn simulation data model of the transformer is constructed, and multi-source positioning features are obtained through processing by utilizing transformer image data of the transformer in the digital twinborn simulation data model and sensing positioning features collected and fused by a transformer partial discharge positioning sensor. Constructing a partial discharge multi-source positioning model by using the multi-source positioning characteristics and a discharge positioning method in a corresponding digital twin simulation data model; image data generated by a generative adversarial network improved based on transformer image data is combined with a fusion signal acquired by a multi-source sensor to obtain a transformer partial discharge positioning condition, adaptive expansion is performed on transformer partial discharge positioning data in a digital twinning state, and a transformer partial discharge positioning result is obtained. A multi-source positioning model is constructed more accurately, the cost of model training is reduced, and the precision of partial discharge positioning is improved.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of scene simulation, and particularly relates to a transformer partial discharge multi-source positioning system based on digital twinning. BACKGROUND

[0002] The transformer is one of the most important electrical equipment in the power system, and its operation condition is directly related to the safe operation of the power system. Once a fault occurs, it will certainly cause large-scale power outages. In recent years, the main transformer equipment of the substation has failed frequently, exposing the problems of insufficient detection, early warning and protection means, insufficient equipment intrinsic safety, and insufficient monitoring technical means.

[0003] The transformer partial discharge (PD) is mainly the discharge of the internal insulation of the transformer under the action of high voltage. The partial discharge only exists in the local position of the insulation and does not immediately form a complete insulation through breakdown or flashover. The partial discharge cannot be directly observed by the operator, and only a special partial discharge detection device can detect it. The detection of the partial discharge signal mainly detects various physicochemical phenomena such as electricity, sound, light, temperature and gas generated in the discharge process through corresponding sensors, and describes them through physical quantities that can represent these partial discharge phenomena. The detection method can be divided into electrical measurement and non-electrical measurement. The electrical measurement uses the pulse of the electric signal generated by the partial discharge to measure, which can reliably find problems in the early stage of the defect, including the pulse current method, the ultra-high frequency method, etc. The non-electrical measurement includes oil gas analysis, light measurement, ultrasonic detection method, etc.

[0004] The current methods for determining the position of the transformer partial discharge mainly include the ultra-high frequency method, the ultrasonic detection method and the high-frequency pulse current method.

[0005] The ultra-high frequency method is to analyze the electromagnetic wave above 300MHz excited by the partial discharge. Its advantage is that it can locate the position of the partial discharge, and the detection frequency band is very high, which can effectively avoid the electromagnetic interference in the environment, but it requires high detection hardware. The ultrasonic detection method is to install ultrasonic sensors on the transformer body to collect the ultrasonic signals generated when the partial discharge occurs inside the transformer for analysis. The biggest advantage of this method is that the electromagnetic interference in the field will not interfere with the ultrasonic signals, and it is relatively convenient in spatial positioning. The disadvantage is that the propagation of ultrasonic waves is greatly affected by the properties of the medium, which will reduce the measurement sensitivity. Both of the above two methods need to install multiple sensors on the transformer body when detecting and positioning the transformer partial discharge. Multiple sensors need to realize synchronous measurement of signals, and a spatial equation is established based on the time of the partial discharge signal reaching each sensor to determine the position of the partial discharge. Since multiple signals need to be time-synchronized, the device wiring and installation are complex.

[0006] High-frequency pulse current method (HF): This method works by detecting and analyzing high-frequency current signals at locations such as the transformer core, grounding wires of clamps, bushing end screens, and surge arrester counter leads when partial discharge occurs inside the transformer. Currently, Rogowski coils are commonly used for measurement. The advantages of this method are high sensitivity, direct detection of the discharge charge, easy installation, accurate reflection of amplitude changes in partial discharge within the transformer, and wide applicability. The disadvantages are poor anti-interference capability, susceptibility to interference from various signals in the grounding grid, and inability to locate partial discharges. In engineering applications, the test data from the high-frequency pulse current method is generally used as a reference for ultra-high frequency or ultrasonic data.

[0007] Based on the above analysis, the problems and shortcomings of the existing technology are as follows: Existing transformer partial discharge monitoring methods lack multi-source data fusion diagnostic analysis methods. Furthermore, real-world transformers require the installation of multiple sensors for analysis, and the parameters collected by each sensor are analyzed piecemeal to pinpoint the partial discharge location. This approach fails to accurately locate the partial discharge. Additionally, when training models, the commonly used digital twin models only utilize existing partial discharge data from real-world transformers without adaptive data expansion to compensate for insufficient model applicability. Inaccurate or incomplete model parameters result in inadequate model classification accuracy. Consequently, it cannot leverage image data to simultaneously process the classified location area, thus failing to achieve the desired effect of region classification followed by precise localization. Summary of the Invention

[0008] To address the aforementioned technical problems, this invention proposes a multi-source localization system for partial discharge of transformers based on digital twins.

[0009] In a first aspect of the invention, a multi-source localization method for partial discharge in transformers based on digital twins is provided, the method comprising: A generative adversarial network digital twin simulation data model of a transformer is constructed, transformer image data of the transformer is collected, and sensing and localization features are obtained by fusion of data from multiple sources of sensors. Based on the generative adversarial network digital twin simulation data model, transformer generated image data is merged and generated. Multi-source positioning features are obtained by processing the transformer generated image data and sensor positioning features, and the corresponding discharge positioning methods in the digital twin simulation data model are collected. A partial discharge multi-source localization model is constructed based on the multi-source localization features and discharge localization method.

[0010] Furthermore, the signals collected by the multi-source sensor include voltage signals, current signals, acoustic signals, ultra-high frequency signals, and optical signals.

[0011] Further, the sensor positioning feature obtained by fusing the signals collected by the multi-source sensor is calculated and fused by using multiple characteristics of the signals.

[0012] Further, the generative adversarial network digital twin simulation data model adopts a generative adversarial network digital twin simulation data model improved based on transformer image data, and the transformer generated image data is generated by optimizing and improving the discriminator of the model.

[0013] Further, the transformer image data uses a threshold method to obtain abnormal part image data.

[0014] Further, the generative adversarial network digital twin simulation data model improved based on transformer image data optimizes and improves the discriminator.

[0015] Further, the partial discharge multi-source positioning model adopts a multi-classification model based on Fisher criterion, the multi-source positioning feature is obtained according to the transformer generated image data and the sensor positioning feature, the feature multi-source splicing is performed by using feature engineering, and the multi-source positioning feature is obtained by taking the transformer generated image data as the main feature vector and connecting the main feature vector with the sensor positioning feature.

[0016] A transformer partial discharge multi-source positioning system based on digital twin is also provided, which includes a transformer image data collection module, a multi-source sensor data collection module, a multi-source sensor data fusion module, a generative adversarial network digital twin simulation data model module, a multi-source positioning feature acquisition module, a partial discharge multi-source positioning model construction module, and a partial discharge positioning analysis module. The transformer image data collection module is configured to collect transformer image data of a transformer. The multi-source sensor data collection module is configured to collect signal features of voltage signals, current signals, acoustic signals, ultrahigh frequency signals, and optical signals. The multi-source sensor data fusion module is configured to obtain sensor positioning features by fusing signals collected by multi-source sensors. The generative adversarial network digital twin simulation data model module is configured to receive the transformer image data and signal data collected by the multi-source sensor data collection module, construct a digital twin simulation data model, and include a generative adversarial network improved based on transformer image data to process the transformer image data to obtain transformer generated image data. The multi-source positioning feature acquisition module is configured to process the transformer generated image data and the sensor positioning feature to obtain a multi-source positioning feature. The partial discharge multi-source positioning model construction module receives the multi-source positioning features and constructs a partial discharge multi-source positioning model according to a discharge positioning method; The partial discharge positioning analysis module inputs the to-be-multi-source-positioned features of the transformer to be positioned for partial discharge in the real power grid into the digital twin model to obtain a real-time discharge positioning method. The partial discharge positioning analysis module is configured to perform positioning analysis on the partial discharge of the transformer according to the real-time discharge positioning method, accurately position the partial discharge, and improve the functional stability of the power grid.

[0017] Further, the generative adversarial network digital twin simulation data model adopts a generative adversarial network digital twin simulation data model improved based on transformer image data, and generates transformer image data by optimizing and improving the discriminator of the model.

[0018] The present application aims at the existing transformer partial discharge monitoring, and the lack of multi-source data fusion diagnosis analysis method, discloses a kind of fusion multi-source data, and without actual in transformer installation multiple sensors are analyzed, analysis is transferred to digital twin model, and the parameter of each sensor acquisition is analyzed to locate transformer partial discharge position according to the calculation method of the positioning influence of sensor parameter, accurately position the position of transformer partial discharge, and when using model to train, the digital twin model not only uses the real transformer partial discharge related data that has been collected, but also uses the abnormal image data based on transformer image maximum value and minimum value black and white image pixel channel based on digital twin model to improve the generative adversarial network model of discriminator, to complement the scene applicability deficiency of digital twin model in classification model input feature positioning, complement and correct model parameter, thereby improve the classification accuracy of subsequent multi-classification model, and by means of image data, the positioning area obtained after classification is processed synchronously, so as to obtain the effect of accurate positioning after regional classification.

[0019] More embodiments and improvement effects of the present application will be further introduced in conjunction with the drawings and specific embodiments. BRIEF DESCRIPTION OF DRAWINGS

[0020] Figure 1 It is the flow chart of the transformer partial discharge multi-source positioning method based on digital twin of the present application; Figure 2 It is the schematic diagram of the transformer partial discharge multi-source positioning system based on digital twin of the present application; Figure 3 It is the transformer region division digital twin simulation diagram in the embodiment of the present application; Figure 4 It is the GAN model principle diagram in the embodiment of the present application; Figure 5 Fig. 1 is a schematic diagram of an electronic device structure for implementing the method of the present application. DETAILED DESCRIPTION

[0021] The present application will be further described in conjunction with the accompanying drawings and specific embodiments. In the present application, image acquisition is color three-channel image.

[0022] In a first aspect of the present application, a transformer partial discharge multi-source positioning method based on digital twinning is provided, the method comprising: constructing a generative adversarial network digital twinning simulation data model of the transformer, collecting transformer image data of the transformer, and obtaining sensor positioning features by fusing signals collected by multi-source sensors; generating transformer generated image data based on the generative adversarial network digital twinning simulation data model, processing multi-source positioning features based on the transformer generated image data and sensor positioning features, and collecting corresponding discharge positioning methods in the digital twinning simulation data model; constructing a partial discharge multi-source positioning model according to the multi-source positioning features and the discharge positioning method.

[0023] Further, the signals collected by the multi-source sensors include voltage signals, current signals, acoustic wave signals, ultrahigh frequency signals, and optical signals.

[0024] Further, the sensor positioning features obtained by fusing signals collected by multi-source sensors are calculated and fused by using multiple characteristics of the signals, and the fusion calculation algorithm is: wherein, is the sensor positioning feature, is the voltage peak value, is the voltage frequency period time, is the current value, is the current pulse repetition rate, is the maximum value of the acoustic wave collected by the acoustic wave sensor, is the minimum value of the acoustic wave collected by the acoustic wave sensor, is the optical signal intensity.

[0025] Since each signal has its advantages and disadvantages, such as voltage and current signals are high in time resolution, but are susceptible to interference, so the amplitude and cycle time are used as part of feature fusion; the acoustic signal positioning is accurate but the propagation attenuation is large, so the maximum and minimum values of the acoustic waves collected inside and outside the sensor are compared to determine the sensing positioning feature, the optical signal is intuitive but the application is limited, the light intensity is the most intuitive data collected, which can greatly correct the sensing positioning feature, so the fusion of these signals can complement each other's advantages, enhance the applicability of the input model feature data, and improve the accuracy and reliability of the local discharge position detection.

[0026] In the data preprocessing process, the feature vectors of different parameters are processed in a dimensionless form. Based on the relationship between each parameter and the local discharge position detection output feature vector, the scalar multiplication of the feature vectors of the multi-source data parameters is a creative data multi-fusion calculation preprocessing method according to the data correlation knowledge of the person skilled in the art, which will not be repeated here.

[0027] Further, the generative adversarial network digital twin simulation data model adopts a generative adversarial network digital twin simulation data model improved based on transformer image data, and generates transformer image data by optimizing and improving the discriminator of the model.

[0028] Further, the transformer image data uses a threshold method to obtain abnormal part image data, and the threshold calculation formula is: This is because the local discharge part is obviously different from other parts in the image, so the threshold is used to simplify the division of the transformer image data, reduce the data amount, and improve the model construction accuracy. In image processing, the black pixel value is the lowest bit 0, and the white pixel value is the highest 255. The minimum value of the white color channel pixels in the transformer color image and the maximum value of the black pixels are used to calculate the threshold, so as to remove the non-abnormal part image data. The pixel part image higher than the threshold is taken as the transformer image data.

[0029] According to the basic theory of image processing, the algorithm or calculation theory of the target image is as required by the present application, so it can be determined that the transformer image obtained according to the threshold is highly associated with the local discharge part. This is a technical means that can be foreseen by the person skilled in the art in accordance with the design common sense, which will not be repeated here.

[0030] Further, the generative adversarial network digital twin simulation data model improved based on transformer image data has a discriminator improvement optimization calculation formula: wherein, is the lowest correct rate of the discriminator, is the transformer image distribution, is the transformer image distribution generated by the transformer image data improved generative adversarial network digital twin simulation data model, is the number of images generated exceeding the threshold , is the number of images generated below the threshold , since in the existing GAN, the discriminator determines the authenticity of the power grid image data using the previous general determination, and does not consider the influence of the image extreme black and white data in the sample on the discriminator, in the present application, according to the scenario that the image extreme value of the partial discharge position is serious, the ratio of the number of images generated by the image extreme black and white in the image generation distribution is used to correct the transformer image distribution data generated, thereby improving the transformer image generation data related to the positioning of the partial discharge in the data generation, and performing accurate positioning correction of the input data for subsequent transformer partial discharge positioning classification.

[0031] The improvement of the discriminator in the generative neural network is the key link to improve the model performance. The core task of the discriminator is to accurately distinguish between real data and generated data, and its optimization direction covers multiple dimensions such as architecture design, loss function, and training strategy.

[0032] Further, the partial discharge multi-source positioning model adopts a multi-classification model based on Fisher criterion, the multi-source positioning feature is obtained according to the transformer generated image data and the sensing positioning feature processing, the feature multi-source splicing is performed by using the feature engineering method, and the multi-source positioning feature is obtained by taking the transformer generated image data as the main feature vector and connecting the main feature vector with the sensing positioning feature.

[0033] Further, the calculation formula of the multi-classification model based on Fisher criterion is: L is the multi-source positioning feature, is the output discharge positioning method, W T is the normal vector perpendicular to the hyperplane, and the value of is bounded to output the multi-classification positioning method, and the area prone to partial discharge of the transformer is divided into , , , , five areas, when is greater than -2 and less than -1, the discharge positioning method outputs area, when When greater than or equal to -1 and less than 0, the discharge positioning method outputs Region, when When equal to 0, the discharge positioning method outputs Region, when When greater than 0 and less than or equal to 1, the discharge positioning method outputs Region, when When greater than 1, the discharge positioning method outputs Region, and in the above five regions, image positioning is performed based on abnormal part image data exceeding the threshold , specifically in the digital twin data model, thereby deriving to the positioning range in the real world, facilitating monitoring and maintenance.

[0034] Also provided is a transformer partial discharge multi-source positioning system based on digital twinning, comprising a transformer image data collection module, a multi-source sensor data acquisition module, a multi-source sensor data fusion module, a generative adversarial network digital twin simulation data model module, a multi-source positioning feature acquisition module, a partial discharge multi-source positioning model construction module, and a partial discharge positioning analysis module, characterized in that: The transformer image data collection module is used to collect transformer image data of a transformer; The multi-source sensor data acquisition module is used to acquire signal features of voltage signals, current signals, sound wave signals, and ultrahigh frequency signals and optical signals; The multi-source sensor data fusion module is used to obtain sensing positioning features by fusing signals acquired by multi-source sensors; The generative adversarial network digital twin simulation data model module is used to receive the transformer image data and signal data acquired by the multi-source sensor data acquisition module, construct a digital twin simulation data model, and contain transformer image data improved generative adversarial network processing of the transformer image data to obtain transformer generated image data; The multi-source positioning feature acquisition module is used to process the transformer generated image data and the sensing positioning features to obtain multi-source positioning features; The partial discharge multi-source positioning model construction module receives the multi-source positioning features and constructs a partial discharge multi-source positioning model; The partial discharge positioning analysis module inputs multi-source positioning features of a transformer to be positioned for partial discharge in a real power grid simulated into a digital twin model into the partial discharge multi-source positioning model to obtain a real-time discharge positioning method; The partial discharge positioning analysis module is used to perform positioning analysis on transformer partial discharge according to the real-time discharge positioning method, accurately perform partial discharge positioning, and improve the functional stability of power grid operation.

[0035] Further, the generative adversarial network digital twin simulation data model adopts a generative adversarial network digital twin simulation data model improved based on transformer image data, and generates transformer generated image data by optimizing and improving the discriminator of the model.

[0036] The present application aims at the existing transformer partial discharge monitoring, and the lack of multi-source data fusion diagnosis analysis method, discloses a kind of fusion multi-source data, and without actual in transformer installation multiple sensors are analyzed, analysis is transferred to digital twin model, and the parameter of each sensor acquisition is analyzed to locate transformer partial discharge position according to the calculation method of the influence of sensor parameter positioning, well for the precise positioning of transformer partial discharge position, and when using model training, the digital twin model not only uses the real transformer partial discharge related data that has been collected, but also uses the abnormal image data based on transformer image maximum value and minimum value black and white image pixel channel based on digital twin model built-in discriminator improved generative adversarial network model, to complete the scene applicability deficiency of input feature in classification model of digital twin model and get positioning, complete correction model parameter, to improve the classification accuracy of subsequent multi-classification model, and by means of image data, the positioning area obtained after classification is processed synchronously, so as to obtain the effect of accurate positioning after regional classification.

[0037] Of course, it can be understood that each embodiment of the present application can realize one of the effects, and the combination of multiple embodiments of the present application can realize all the effects described above, but it is not required that each embodiment of the present application realizes all the advantages and effects described above, because each embodiment of the present application can constitute a separate technical solution and make one or more contributions to the prior art.

[0038] The part of module structure of the present application is not particularly clear, and the content recorded in the prior art is used as the criterion.The prior art mentioned in the foregoing background section and the specific embodiment section of the present application can be used as a part of the present application to understand the meaning of part of technical features or parameters.The protection scope of the present application is subject to the content actually recorded in the claims.

Claims

1. A transformer partial discharge multi-source positioning method based on digital twinning, characterized in that, The method comprises: constructing a generative adversarial network digital twin simulation data model of the transformer, collecting transformer image data of the transformer, and collecting fusion sensor positioning features by using a multi-source sensor; generating transformer generated image data based on the generative adversarial network digital twin simulation data model, processing multi-source positioning features based on the transformer generated image data and the sensor positioning features, and collecting corresponding discharge positioning methods in the digital twin simulation data model; constructing a partial discharge multi-source positioning model according to the multi-source positioning features and the discharge positioning method.

2. The transformer partial discharge multi-source positioning method based on digital twinning according to claim 1, wherein: the signals collected by the multi-source sensor include voltage signals, current signals, acoustic signals, ultrahigh frequency signals, and optical signals.

3. The transformer partial discharge multi-source positioning method based on digital twinning according to claim 2, wherein: the fusion sensor positioning features are obtained by using multiple characteristics of the signals for calculation and fusion.

4. The transformer partial discharge multi-source positioning method based on digital twinning according to claim 1 or 3, wherein: the generative adversarial network digital twin simulation data model is an improved generative adversarial network digital twin simulation data model based on transformer image data, and the generation of transformer generated image data is performed by optimizing and improving the discriminator of the model.

5. The transformer partial discharge multi-source positioning method based on digital twinning according to claim 4, wherein: the threshold method is used to obtain abnormal part image data from the transformer image data.

6. The transformer partial discharge multi-source positioning method based on digital twinning according to claim 5, wherein: the improved generative adversarial network digital twin simulation data model based on transformer image data optimizes and improves the discriminator.

7. The transformer partial discharge multi-source positioning method based on digital twinning according to claim 2, wherein: the partial discharge multi-source positioning model is a multi-classification model based on Fisher criterion, the multi-source positioning features are obtained by using feature engineering to splice multiple sources, and the multi-source positioning features are obtained by using the transformer generated image data as the main feature vector and concatenating the main feature vector with the sensor positioning features.

8. A transformer partial discharge multi-source positioning system based on digital twinning, comprising a transformer image data collection module, a multi-source sensor data acquisition module, a multi-source sensor data fusion module, a generative adversarial network digital twin simulation data model module, a multi-source positioning feature acquisition module, a partial discharge multi-source positioning model construction module, and a partial discharge positioning analysis module, wherein: the transformer image data collection module is used to collect transformer image data of the transformer; the multi-source sensor data acquisition module is used to acquire signal features of voltage signals, current signals, acoustic signals, ultrahigh frequency signals, and optical signals; The multi-source sensor data fusion module fuses the signals collected by the multi-source sensor to obtain a sensing positioning feature; The generative adversarial network digital twin simulation data model module is configured to receive the transformer image data and the signal data collected by the multi-source sensor data acquisition module, construct a digital twin simulation data model, and contain transformer image data improved generative adversarial network processing of the transformer image data to obtain transformer generated image data. The multi-source positioning feature acquisition module processes the transformer generated image data and the sensing positioning feature to obtain a multi-source positioning feature. The partial discharge multi-source positioning model construction module receives the multi-source positioning feature and a discharge positioning method to construct a partial discharge multi-source positioning model. The partial discharge positioning analysis module inputs the multi-source positioning feature of the transformer to be positioned for partial discharge in the real power grid into the partial discharge multi-source positioning model to obtain a real-time discharge positioning method. The partial discharge positioning analysis module is configured to perform positioning analysis on the transformer partial discharge according to the real-time discharge positioning method, accurately perform partial discharge positioning, and improve the functional stability of the power grid operation.

9. The transformer partial discharge multi-source positioning system based on digital twinning of claim 8, wherein: The generative adversarial network digital twin simulation data model adopts a generative adversarial network digital twin simulation data model improved based on transformer image data, and generates transformer generated image data by optimizing and improving the discriminator of the model.

Images