Power load identification method and system based on GAF and ResNet-CBAM fusion architecture

Abstract

The invention discloses a power load identification method and system based on a GAF and ResNet-CBAM fusion architecture, and relates to the technical field of power load identification, and the method comprises the steps: collecting power data, and storing the power data as a first database after initial data processing; performing event detection to determine event occurrence time; acquiring electric quantity parameter data corresponding to the steady-state data generation event before and after the event, generating a signal waveform, and extracting characteristic harmonic waves; and generating a two-dimensional feature image through a GAF algorithm, training the two-dimensional feature image by using a ResNet-CBAM neural network to generate a load type identification model, and identifying the load type by using the identification model. According to the scheme, the original signal is converted into the angle field projection through the improved GAF coding, the spatial representation of the features is realized, the improved GAF can effectively convert the power load data into an image form, the power load data can be directly processed by a traditional neural network, and the power load identification precision can be improved.

Description

Technical Field

[0001] The present application relates to the technical field of power load identification, and relates to a power load identification method and system based on a GAF and ResNet-CBAM fusion architecture. Background Art

[0002] With the rapid development of industrial automation, smart grid and new energy grid connection technologies, problems such as frequent load fluctuations, an increase in non-linear loads, and malignant loads causing overheating of wires and internal short circuits in electrical appliances leading to fires have become increasingly prominent.

[0003] Some existing power load identification systems are difficult to extract effective distinguishing features from power data during use. The reasons are as follows: on the one hand, due to accuracy errors in the data collection process, subsequent power load identification is prone to errors; on the other hand, the electrical characteristics of some loads are not obvious and there may be similarities between the electrical characteristics of different loads. The above reasons directly increase the difficulty of power system load identification. Therefore, there is an urgent need for a high-precision and intelligent load identification method to solve the above problems. Summary of the Invention

[0004] Aiming at the problem of low power load identification accuracy in practical applications, a first object of the present application is to provide a power load identification method based on a GAF and ResNet-CBAM fusion architecture. It converts one-dimensional waveform data representing a specific load into an image through GAF technology, and then realizes the recognition and classification of specific image features through the fusion of a ResNet network and a CBAM module, ultimately achieving a higher-precision and more intelligent load type recognition effect. To implement the above power load identification method, a second object of the present application is to propose a power load identification system based on a GAF and ResNet-CBAM fusion architecture. At the same time, to facilitate the popularization and application of the above power load identification method, a third object of the present application is to propose a computer-readable storage medium on which a program module for implementing the above power load identification method is loaded. The specific solutions are as follows:

[0005] A power load identification method based on a GAF and ResNet-CBAM fusion architecture includes:

[0006] After connecting a set load to the power line, collect the power data of the power system, and store it in the first database after initial data processing;

[0007] Based on the power data in the first database, use a detection algorithm to perform event detection, determine the occurrence time of each event and store it;

[0008] Obtain and calculate the power parameter data corresponding to each event based on the steady-state data before and after the occurrence of each event, and associate and store it in the second database;

[0009] Extract the power parameter data associated with the event from the second database and generate a signal waveform, and use Fourier transform to extract the characteristic harmonics characterizing the above event;

[0010] Perform image processing on the characteristic harmonics through the GAF algorithm to generate a two-dimensional characteristic image;

[0011] Combine the power parameter data stored in the second database, and use the ResNet-CBAM neural network to train the generated two-dimensional characteristic image to generate a load type recognition model;

[0012] Obtain power data and identify and output the load type based on the load type recognition model;

[0013] Wherein, the events include the load entering the working state and the load ending the working state;

[0014] The power parameter data includes current data and voltage data.

[0015] Through the above technical solution, the processed one-dimensional power data is converted into two-dimensional characteristic image data by using the GAF algorithm, and then a recognition model for identifying the load type is obtained through training by using a neural network. By using the above load type recognition model and combining the collected power data, the load type working in the system can be accurately identified, and the accuracy and robustness of the load recognition are improved.

[0016] Further, collect and obtain the power data of the power system, and store it as the first database after initial data processing, including:

[0017] Collect the power data of the power system through a current transformer and output it to a filter circuit to eliminate high-frequency noise;

[0018] Adjust the amplitude of the current signal and / or voltage signal in the power data to make it adapt to the input range of the analog-to-digital converter;

[0019] Input the power data into the analog-to-digital converter for analog-to-digital conversion, generate digitalized power data and store it as the first database.

[0020] Through the above technical solution, the power data can be converted into a digital signal, eliminating noise interference and facilitating subsequent data processing steps.

[0021] Further, use a detection algorithm for event detection, including:

[0022] Calculate the aggregated power time series data based on the collected power data, calculate the power difference sequence data after filtering processing, and store it as the input data of the RRCF algorithm;

[0023] Calculate the anomaly score for each data point in the power difference sequence data, and compare the calculated anomaly score with a set threshold to achieve event detection;

[0024] Among them, when the anomaly score is greater than the set threshold, it is determined that an event has occurred;

[0025] When the anomaly score is not greater than the set threshold, it is determined that no event has occurred.

[0026] Through the above technical solution, using the power difference as input data, abnormal data points are monitored in real time, so as to know when an event occurs, that is, when the load (electrical appliance) is turned on to work or when it is turned off to end work; and the RRCF algorithm also has a self-check processing effect. When a possible event is detected, the power difference threshold is used to further suppress misdetected events.

[0027] Furthermore, calculate and generate the power parameter data corresponding to each event according to the steady-state data before and after each event occurs, and its calculation formula is:

[0028] V = (V

[0032] ,

[0031] , off , on , off ,

[0030] , on ,

[0029] ,

[0035] ,

[0028] ,

[0034] ,

[0033] , , on , off , on , off , , , , , , + V on ) / 2;

[0029] I = I on - I off ;

[0030] Among them, V is the steady-state voltage of the power line, and the V off is the voltage of the power line after the load ends work, V on is the voltage of the power line after the load enters the working state, I is the steady-state current of the power line, I on is the current of the power line after the load enters the working state, I off is the current of the power line after the load ends work;

[0031] Obtaining the power parameter data of the above events according to the steady-state data before and after the event occurs also includes:

[0032] Based on the already obtained power parameter data, use the DDIM algorithm model to perform data augmentation;

[0033] Compare and evaluate the augmented data with the original data to form the second database.

[0034] Through the above technical solution, not only can the power parameter data corresponding to the event be accurately obtained, including voltage and current data, but also the data can be augmented, which is beneficial to improving the training accuracy of the subsequent recognition model.

[0035] Furthermore, the two-dimensional feature image is an RGB image, and generating the two-dimensional feature image through the GAF algorithm includes:

[0036] Sample the characteristic harmonics to obtain harmonic sampling data, which is used as one-dimensional time series data;

[0037] Normalize the above harmonic sampling data, perform PA transformation, and then represent it using polar coordinates;

[0038] Encode based on the GAF algorithm to obtain a single-channel image of time and frequency spectrum, and correspond the single-channel image to the three channels of the RGB color space to generate an RGB image.

[0039] Furthermore, the power load identification method further includes:

[0040] After connecting the load, collect the power data of the power system at different positions on the power line, and associate and store them with the sampling positions as multiple power data groups;

[0041] Use the load type identification model to identify the data in multiple power data groups and output multiple load type identification results;

[0042] Determine the identification accuracy rate of each sampling position according to multiple load type identification results;

[0043] Mark the identification accuracy rate of each sampling position.

[0044] Through the above technical solutions, the best data sampling positions in the power line can be found, which is convenient for engineers to configure relevant power data sampling devices at the above sampling positions, and improve the accuracy of subsequent load type identification.

[0045] A power load identification system based on the GAF and ResNet-CBAM fusion architecture, including:

[0046] A data storage module configured to store set data and provide a data interface for data retrieval or query;

[0047] A data acquisition module, including a data collection unit and a data initial processing unit, configured to collect and obtain the power data of the power system after the load is connected to the power line, and output it to the data storage module after initial data processing and store it as a first database;

[0048] An event detection module configured to be data-connected to the data storage module, used to perform event detection according to a set detection algorithm, determine the occurrence time of each event and output and store it;

[0049] A data augmentation module, configured to be data-connected to the event detection module and the data storage module. After event detection is completed, it determines the power parameter data of the above event using the steady-state data before and after the event, and performs data augmentation using the DDIM algorithm model. The augmented data is output to the data storage module after being compared and evaluated with the original data to form a second database;

[0050] A feature waveform extraction module, configured to extract the power parameter data associated with the event from the second database and generate a signal waveform, and use Fourier transform to extract the characteristic harmonics characterizing the above event;

[0051] A feature image generation module, configured to perform image processing on the characteristic harmonics through the GAF algorithm to generate a two-dimensional feature image;

[0052] An identification model training module, configured to use the ResNet-CBAM neural network to train the generated two-dimensional feature image in combination with the power parameter data stored in the second database to generate a load type identification model;

[0053] A load type identification model, configured to be data-connected to the data acquisition module, used to acquire power data and identify and output the load type based on the load type identification model;

[0054] Wherein, the events include the load entering the working state and the load ending the working state;

[0055] The power parameter data includes current data and voltage data.

[0056] Further, the data acquisition unit includes an instrument transformer, which collects current and voltage signals in the power line at a set sampling site on the power line;

[0057] The data initial processing unit includes a signal filtering circuit unit, a signal amplitude modulation circuit unit, and an ADC converter, which are used to receive the current and voltage signals and perform filtering and noise reduction processing. Then, the signal amplitude is adjusted to the input range suitable for the analog-to-digital converter by the signal amplitude modulation circuit unit and output to the ADC converter to be converted into a digital signal and output.

[0058] Further, the event detection module, the data augmentation module, the feature waveform extraction module, and the feature image generation module are all configured at the on-site end of the power system and are data-connected to the data acquisition module;

[0059] The data storage module includes a local memory and a cloud memory, and the identification model training module is configured in the cloud server to receive data in the cloud data device for model training;

[0060] The load type recognition model is deployed at the field end or the cloud server end, and is data-connected to the data acquisition module and the display device configured on-site.

[0061] Through the above technical solution, it is convenient for the storage of power data, can also accelerate the training and optimization of the model, and is convenient for the large-scale deployment and application of the load type recognition model.

[0062] A computer-readable storage medium has a program module loaded thereon. When the above program module is executed by a processor, it is used to implement the power load recognition method based on the GAF and ResNet-CBAM fusion architecture as described above.

[0063] Through the above technical solution, it is beneficial to the popularization and use of the above power load recognition method.

[0064] This application has at least the following beneficial effects:

[0065] (1) Through the improved GAF encoding, the original signal can be converted into a projection in the angular domain, thereby realizing the spatial representation of features. In the load recognition method of the present invention, through the improved GAF, the power load data can be effectively converted into an image form, enabling traditional CNNs to directly process it.

[0066] (2) By introducing the ResNet-CBAM neural network, the key features in the load signal can be captured more precisely, improving the recognition accuracy.

[0067] (3) Combining GAF and ResNet-CBAM can give full play to their respective advantages. GAF converts the load signal into an image to meet the input requirements of CNNs; ResNet extracts complex features through its deep network structure; CBAM further enhances the attention mechanism of the model, improving the sensitivity to key features. Through this fusion architecture, the temporal and spatial features in the load signal can be better processed, improving the accuracy and robustness of load recognition. BRIEF DESCRIPTION OF THE DRAWINGS

[0068] Figure 1 It is a schematic flow diagram of the power load recognition method of this application;

[0069] Figure 2 It is a schematic structural diagram of the functional modules of the power load recognition system of this application.

[0070] Reference Numerals: 100, data storage module; 200, data acquisition module; 210, data collection unit; 220, data initial processing unit; 221, signal filtering circuit unit; 222, signal amplitude modulation circuit unit; 223, ADC converter; 300, event detection module; 400, data expansion module; 500, characteristic waveform extraction module; 600, characteristic image generation module; 700, recognition model training module; 800, load type recognition model; 900, data communication unit. Detailed Embodiments

[0071] The following details the embodiments of the present application, and examples of the embodiments are shown in Figure 1 - Figure 2 it.

[0072] In the description of this specification, the description with reference to the terms "certain embodiments", "one embodiment", "some embodiments", "schematic embodiments", "examples", "specific examples", or "some examples" means that the specific features, structures, materials, or characteristics described in connection with the embodiments or examples are included in at least one embodiment or example of the present application. In this specification, the schematic descriptions of the above terms do not necessarily refer to the same embodiments or examples. Moreover, the specific features, structures, materials, or characteristics described can be combined in a suitable manner in any one or more embodiments or examples.

[0073] A power load identification method based on the GAF and ResNet-CBAM fusion architecture, as Figure 1 shown, mainly includes the following steps:

[0074] S100, after connecting a set load to the power line, collect and obtain the power data of the power system, and store it as the first database after initial data processing.

[0075] In the embodiments of the present application, the power data of the power system, including voltage and current data, is collected through the mutual inductor set on the power line. The initial data steps in the above step S100 include filtering and denoising the collected power data to eliminate high-frequency noise, and then adjusting the amplitude of the current signal and / or voltage signal in the power data to make it adapt to the input range of the analog-to-digital converter, so as to facilitate inputting the power data into the analog-to-digital converter for analog-to-digital conversion, generating digital power data and storing it as the first database.

[0076] S200, based on the power data in the first database, use the detection algorithm to perform event detection, determine the occurrence time of each event and store it.

[0077] In the embodiments of the present application, the above events include the load entering the working state and the load ending the working state. The solutions of the present application are used to detect the types of loads connected to the power system, including non-linear loads, such as arc heating equipment, DC electrical equipment with AC rectification, switching power supply equipment, etc. After the above non-linear electrical equipment is connected to the power system as a harmonic source and operates, it will have an adverse impact on the quality of the grid voltage waveform.

[0078] In step S200, the Robust Random CutForest (RRCF) algorithm is mainly used for event detection. RRCF is an open-source algorithm for anomaly detection in streaming data. The detection process includes:

[0079] S210, Calculate the aggregated power time series data based on the collected power data, and calculate the power difference sequence data after filtering processing as the input data of the RRCF algorithm;

[0080] The above analysis of the aggregated power time series data involves summarizing and analyzing the power data arranged in chronological order to understand its variation laws and characteristics. The time series data includes power measurement values recorded in chronological order, such as power consumption data per second or per minute, which is convenient for trend analysis and prediction.

[0081] In step S210, first, perform mean pooling processing on the aggregated power time series data with a high sampling rate to achieve the filtering effect. For the aggregated power time series data with a low sampling rate, since there is no violent fluctuation, filtering is not required.

[0082] S220, Calculate the anomaly score of each data point in each power difference sequence data, and compare the calculated anomaly score with the set threshold to achieve event detection:

[0083] When the anomaly score is greater than the set threshold, it is determined that an event has occurred;

[0084] When the anomaly score is not greater than the set threshold, it is determined that no event has occurred.

[0085] The RRCF algorithm uses the power difference as the input data to monitor abnormal data points in real time, so as to know when the event occurs (that is, when the electrical appliance is turned on to work or when it is turned off to end the work). And the above algorithm also has a self-checking and processing effect. When a possible event is detected, the false detection event is further suppressed through the power difference threshold.

[0086] S300, Obtain and calculate the power parameter data corresponding to each event based on the steady-state data before and after each event occurs, and store them in association in the second database.

[0087] The power parameter data includes current data and voltage data.

[0088] The power parameter data corresponding to each event is calculated and generated based on the steady-state data before and after each event occurs. The calculation formula is:

[0089] V = (V off + V on ) / 2; I = I on - I off . Among them, V is the steady-state voltage of the power line, and the V off is the voltage of the power line after the load finishes working, V on is the voltage of the power line after the load enters the working state, I is the steady-state current of the power line, I on is the current of the power line after the load enters the working state, I off is the current of the power line after the load finishes working.

[0090] Obtaining the power parameter data of the above events based on the steady-state data before and after the event also includes: Based on the already obtained power parameter data, using the DDIM algorithm model for data augmentation, and then comparing and evaluating the augmented data with the original data to form the second database.

[0091] The above DDIM algorithm model refers to the Denoising Diffusion Implicit Models, which is an improved diffusion model designed to accelerate the process of generating data.

[0092] S400, Extract the power parameter data associated with the event from the second database and generate a signal waveform, and use the Fourier transform (FFT) to extract the characteristic harmonics used to characterize the above event.

[0093] S500, Perform image processing on the characteristic harmonics through the GAF algorithm to generate a two-dimensional characteristic image.

[0094] The above GAF algorithm is the Gramian Angular Field (GAF) method, which is an effective method for converting one-dimensional time series data into a two-dimensional image representation. Its working principle is: By regarding each data point in the one-dimensional time series data as a point in the vector space and calculating the cosine values of the angles between these points, and then mapping these cosine values to the pixels of the two-dimensional image, so as to generate an image that can reflect the dynamic and periodic characteristics of the time series.

[0095] In the implementation manner of this application, the two-dimensional characteristic image is an RGB image. Generating the two-dimensional characteristic image through the GAF algorithm includes:

[0096] S510. Sample the characteristic harmonics to obtain harmonic sampling data, which is used as one-dimensional time series data;

[0097] S520. Normalize the above harmonic sampling data, such as normalizing it to the range of [0, 1] to eliminate the influence of different dimensions on the results. After that, perform polar coordinate transformation (PA transformation) and represent it using polar coordinates;

[0098] S530. Encode based on the GAF algorithm to obtain a single-channel image of time and frequency spectrum. Corresponding the single-channel image to the three channels of the RGB color space to generate an RGB image.

[0099] In the above step S530, the steps of generating an image based on the GAF algorithm include constructing a Gram matrix based on the processed harmonic sampling data, regarding the time series data as vectors in a vector space, and calculating the inner product between these vectors. Calculate the angle between vectors at different times according to the Gram matrix and convert it into a value between 0 and 1, thereby generating a new angle matrix. Use the angle matrix as the pixel values of the image to generate a two-dimensional image. Each pixel value in the image corresponds to the cosine value of the angle between different times in the time series data, thus retaining the time information and dynamic characteristics of the data.

[0100] S600. Combine the power parameter data stored in the second database and use the ResNet-CBAM neural network to train the generated two-dimensional feature image to generate a load type recognition model.

[0101] The process of training to generate a load type recognition model includes: first, preprocess the above RGB image, then divide the data in the second database into a training set and a validation set, use the cross-entropy loss function and the Adam or SGD optimizer. Then input the image into the network and calculate the output, calculate the loss value, and update the network parameters through backpropagation. Repeat multiple training cycles until the model converges.

[0102] S700. Obtain power data and identify the output load type based on the load type recognition model.

[0103] In practical applications, affected by factors such as environmental electromagnetic interference and load distribution location, the power data collected at each sampling site on the power line has different harmonic characteristics. Therefore, the power load recognition method further includes:

[0104] S800. After connecting the load, collect the power data of the power system at different sites on the power line and store it in association with the sampling sites as multiple power data groups;

[0105] S810. Use the load type recognition model to identify the data in multiple power data groups and output multiple load type recognition results;

[0106] S820. Determine the recognition accuracy rate of each sampling site according to the recognition results of multiple load types;

[0107] S830. Mark the recognition accuracy rate of each sampling site.

[0108] By using steps S800 - S830, the best data sampling sites in the power line can be found, which is convenient for engineers to configure relevant power data sampling devices at the above sampling sites, and improve the accuracy of subsequent load type recognition.

[0109] To implement the above power load recognition method, an embodiment of the present application also discloses a power load recognition system based on the GAF and ResNet - CBAM fusion architecture. As shown in combination with Figure 2 shown, it mainly includes a data storage module 100, a data acquisition module 200, an event detection module 300, a data augmentation module 400, a feature waveform extraction module 500, a feature image generation module 600, a recognition model training module 700, and a load type recognition model 800.

[0110] The data storage module 100 is configured to store set data and provide a data interface for data retrieval or query. In the implementation manner of the present application, for the convenience of storing and retrieving power data, the data storage module 100 includes a local memory and a cloud memory, and data transmission and interaction are realized through a data communication unit 900 provided between the field end and the cloud memory.

[0111] The data acquisition module 200 includes a data collection unit 210 and a data initial processing unit 220, and is configured to collect and obtain the power data of the power system after the load is connected to the power line, and output it to the data storage module 100 after initial data processing and store it as a first database. Specifically, the data collection unit 210 includes an instrument transformer, which collects current and voltage signals in the power line at a set sampling site on the power line. The data initial processing unit 220 includes a signal filtering circuit unit 221, a signal amplitude modulation circuit unit 222, and an ADC converter 223, which are used to receive the above current and voltage signals and perform filtering and noise reduction processing, and then adjust the signal amplitude to the input range adapted to the analog - to - digital converter through the signal amplitude modulation circuit unit 222, and then output it to the ADC converter 223 to be converted into a digital signal and output.

[0112] The event detection module 300 is configured to be data - connected to the data storage module 100, and is used to perform event detection according to a set detection algorithm, determine the occurrence time of each event and output and store it. In the implementation manner of the present application, the RRCF (Robust Random Cut Forest) algorithm is used to perform event detection.

[0113] The data augmentation module 400 is configured to be data-connected to the event detection module 300 and the data storage module 100. After the event detection is completed, the steady-state data before and after the event is used to determine the power parameter data of the above event. The power parameter data includes current data and voltage data. Then, the DDIM algorithm model is used for data augmentation. The augmented data is output to the data storage module 100 after being compared and evaluated with the original data to form a second database.

[0114] The feature waveform extraction module 500 is configured to extract the power parameter data associated with the event from the second database and generate a signal waveform, and then use Fourier transform to extract the characteristic harmonics used to characterize the above event.

[0115] The feature image generation module 600 has a GAF algorithm unit built in. The GAF algorithm is used to perform image processing on the characteristic harmonics to generate a two-dimensional feature image. In the embodiment of the present application, an RGB image is generated.

[0116] The recognition model training module 700 is configured to use the ResNet-CBAM neural network to train the generated two-dimensional feature image in combination with the power parameter data stored in the second database to generate a load type recognition model 800.

[0117] The load type recognition model 800 is configured to be data-connected to the data acquisition module 200, and is used to acquire power data and identify and output the load type based on the load type recognition model 800.

[0118] Furthermore, the event detection module 300, the data augmentation module 400, the feature waveform extraction module 500, and the feature image generation module 600 are all configured at the on-site end of the power system, stored in a specific instrument device in the form of a program module, such as a PC at the on-site end, and are data-connected to the data acquisition module 200. To improve the model training speed and facilitate the continuous optimization of the recognition model, the recognition model training module 700 is configured in the cloud server and receives data in the cloud data device for model training.

[0119] The generated load type recognition model 800 after training is deployed at the on-site end or the cloud server end, and is data-connected to the data acquisition module 200 and the display device configured on-site, and is used to identify the load type in the power system according to the received power data. The above technical solution is conducive to the large-scale deployment and application of the load type recognition model 800 described in the present application.

[0120] To facilitate the popularization and application of the solution of the present application, the embodiment of the present application also discloses a computer-readable storage medium, on which a program module is loaded. When the above program module is executed by a processor, it is used to implement the power load recognition method based on the GAF and ResNet-CBAM fusion architecture as described above.

[0121] Although the embodiments of the present application have been shown and described above, it can be understood that the above embodiments are exemplary and should not be construed as limitations on the present application. Those of ordinary skill in the art can make changes, modifications, substitutions, and variations to the above embodiments within the scope of the present application.

Claims

1. A method for identifying electrical loads based on a fused architecture of GAF and ResNet-CBAM, characterized in that, Including: After connecting a set load to the power line, collect and obtain the power data of the power system, and store it in the first database after initial data processing; Based on the power data in the first database, use a detection algorithm to perform event detection, determine the occurrence time of each event and store it; Obtain and calculate the power parameter data corresponding to each event based on the steady-state data before and after the occurrence of each event, and associate and store it in the second database; Extract the power parameter data associated with the event from the second database and generate a signal waveform, and use Fourier transform to extract the characteristic harmonics characterizing the above events; Perform image processing on the characteristic harmonics through the GAF algorithm to generate a two-dimensional characteristic image; Combined with the power parameter data stored in the second database, use the ResNet-CBAM neural network to train the generated two-dimensional characteristic image to generate a load type recognition model; Obtain power data and identify and output the load type based on the load type recognition model; Wherein, the events include the load entering the working state and the load ending the working state; The power parameter data includes current data and voltage data.

2. The power load identification method according to claim 1, characterized in that Collect and obtain the power data of the power system, and store it in the first database after initial data processing, including: Collect the power data of the power system through a current transformer and output it to a filter circuit to eliminate high-frequency noise; Adjust the amplitude of the current signal and / or voltage signal in the power data to make it adapt to the input range of the analog-to-digital converter; Input the power data into the analog-to-digital converter for analog-to-digital conversion, generate digital power data and store it in the first database.

3. The power load identification method according to claim 1, wherein, Using a detection algorithm for event detection, including: Calculate the aggregated power time series data based on the collected power data, calculate the power difference sequence data after filtering, and store it as the input data of the RRCF algorithm; Calculate the anomaly score of each data point in the power difference sequence data, and compare the calculated anomaly score with a set threshold to achieve event detection; Wherein, when the anomaly score is greater than the set threshold, it is determined that an event has occurred; When the anomaly score is not greater than the set threshold, it is determined that no event has occurred.

4. The power load identification method according to claim 1, characterized in that Calculate the power parameter data corresponding to each event based on the steady-state data before and after the occurrence of each event, and its calculation formula is: V = (V off + V on ) / 2; I = I on -I off ; Among them, V is the steady-state voltage of the power line, and the V off is the voltage of the power line after the load finishes working, V on is the voltage of the power line after the load enters the working state, I is the steady-state current of the power line, I on is the current of the power line after the load enters the working state, I off is the current of the power line after the load finishes working; Obtaining the power parameter data of the above events based on the steady-state data before and after the event occurrence further includes: Based on the already obtained power parameter data, use the DDIM algorithm model for data augmentation; Compare and evaluate the augmented data with the original data to form the second database.

5. The power load identification method according to claim 1, wherein The two-dimensional characteristic image is an RGB image, and generating the two-dimensional characteristic image through the GAF algorithm includes: Sample the characteristic harmonics to obtain harmonic sampling data as one-dimensional time series data; Normalize the above harmonic sampling data and represent it in polar coordinates after PA transformation; Based on the GAF algorithm encoding, obtain a single-channel image of time and frequency spectrum, and correspond the single-channel image to the three channels of the RGB color space to generate an RGB image.

6. The power load identification method according to claim 1, characterized in that The power load identification method further includes: After connecting the load, power data of the power system is collected at different positions of the power line and associated with the sampling positions and stored as multiple power data groups; The data in the multiple power data groups is identified by using the load type identification model, and multiple load type identification results are output; The identification accuracy of each sampling position is determined according to the multiple load type identification results; The identification accuracy of each sampling position is marked.

7. A power load identification system based on the GAF and ResNet-CBAM fusion architecture, characterized in that, It includes: A data storage module (100), configured to store set data and provide a data interface for data retrieval or query; A data acquisition module (200), including a data acquisition unit (210) and a data initial processing unit (220), configured to collect and obtain the power data of the power system after the load is connected to the power line, and output it to the data storage module (100) after initial data processing and store it as a first database; An event detection module (300), configured to be data-connected to the data storage module (100), and used to perform event detection according to a set detection algorithm, determine the occurrence time of each event and output and store it; A data expansion module (400), configured to be data-connected to the event detection module (300) and the data storage module (100). After the event detection is completed, the power parameter data of the above event is determined by using the steady-state data before and after the event, and the DDIM algorithm model is used for data expansion. The expanded data is output to the data storage module (100) after being compared and evaluated with the original data to form a second database; A characteristic waveform extraction module (500), configured to extract the power parameter data associated with the event from the second database and generate a signal waveform, and use Fourier transform to extract the characteristic harmonics used to characterize the above event; A characteristic image generation module (600), configured to perform image processing on the characteristic harmonics through the GAF algorithm to generate a two-dimensional characteristic image; An identification model training module (700), configured to use the ResNet-CBAM neural network to train the generated two-dimensional characteristic image in combination with the power parameter data stored in the second database to generate a load type identification model (800); A load type identification model (800), configured to be data-connected to the data acquisition module (200), and used to obtain power data and identify and output the load type based on the load type identification model (800); Wherein, the event includes the load entering the working state and the load ending the working state; The power parameter data includes current data and voltage data.

8. The power load identification system according to claim 7, wherein The data acquisition unit (210) includes an instrument transformer, and collects current and voltage signals in the power line at a set sampling position on the power line; The data initial processing unit (220) includes a signal filtering circuit unit (221), a signal amplitude modulation circuit unit (222), and an ADC converter (223), which are used to receive the current and voltage signals and perform filtering and noise reduction processing, and then the signal amplitude modulation circuit unit (222) adjusts the signal amplitude to the input range suitable for the analog-to-digital converter and outputs it to the ADC converter (223) to be converted into a digital signal and output.

9. The power load identification system according to claim 7, wherein the event detection module (300), the data augmentation module (400), the feature waveform extraction module (500), and the feature image generation module (600) are all configured at the on-site end of the power system and are connected to the data acquisition module (200) for data connection; the data storage module (100) includes a local memory and a cloud memory, and the recognition model training module (700) is configured in a cloud server to receive data in the cloud data device for model training; the load type recognition model (800) is deployed at the on-site end or the cloud server end and is connected to the data acquisition module (200) and the display device configured on-site for data connection.

10. A computer-readable storage medium, characterized in that, A program module is loaded thereon, and when the program module is executed by a processor, it is used to implement the power load identification method based on the GAF and ResNet-CBAM fusion architecture as described in any one of claims 1-6.

Images