Transformer substation fault identification method and system based on image detection

Abstract

The invention discloses a transformer substation fault identification method and system based on image detection, and relates to the technical field of transformer substation fault identification, and the method comprises the steps: collecting historical imaging data in a transformer substation, and generating an original data set in combination with weather data of imaging acquisition time; calculating a global influence coefficient of the weather data on the imaging data based on the original data set; collecting equipment fault historical diagnosis data in the transformer substation, and generating a training data set in combination with the original data set; establishing a fault identification model by using the global influence coefficient and the training data set; imaging data collected by the imaging device are obtained in real time, a fault recognition result is obtained after the imaging data pass through the fault recognition model, and the fault recognition result is forwarded to operation and maintenance personnel in the transformer substation. According to the method, the fault recognition period is greatly shortened, the training amount of the fault recognition model is greatly reduced, and therefore the consumption of computing resources is reduced while the recognition precision is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of substation fault identification, and in particular to a substation fault identification method and system based on image detection. Background Art

[0002] At present, infrared imaging technology plays a crucial role in substation equipment fault identification. By detecting the infrared radiation emitted by power equipment and converting it into a visible image, the temperature distribution of the equipment can be identified. Then, through deep learning algorithms, it can be determined whether there is a fault in the equipment based on the temperature information. Temperature anomalies are usually precursors to equipment faults, so infrared imaging technology can effectively prevent faults from occurring. However, infrared imaging technology is greatly affected by weather (such as temperature and light). The infrared images of the same equipment at noon and at night are inconsistent. To achieve accurate fault identification of a single equipment, a large amount of imaging data and annotation data are required. It can be imagined how huge the workload and data volume are for fault identification of equipment in a substation, which will inevitably lead to the model consuming a large amount of computing resources and computing time. Therefore, how to improve the efficiency of the algorithm and reduce resource consumption on the premise of accurate and real-time fault identification has become a major issue that researchers in this field urgently need to solve. Summary of the Invention

[0003] The present invention provides a substation fault identification method based on image detection, including:

[0004] Step1: Collect historical imaging data in the substation and generate an original data set by combining the weather data at the time of imaging acquisition;

[0005] Step2: Calculate the global influence coefficient of weather data on imaging data based on the original data set;

[0006] Step3: Collect historical diagnostic data of equipment faults in the substation and generate a training data set by combining the original data set;

[0007] Step4: Establish a fault identification model using the global influence coefficient and the training data set;

[0008] Step5: Real-time obtain the imaging data collected by the imaging device, obtain the fault identification result through the fault identification model, and forward the fault identification result to the operation and maintenance personnel in the substation.

[0009] For the above-mentioned substation fault identification method based on image detection, calculating the global influence coefficient of weather data on imaging data based on the original data set is specifically divided into the following sub-steps:

[0010] Extract the color features of each imaging in the original data set to replace the imaging data in the original data pair;

[0011] Deduplicate and sort the replaced data pairs according to the values of each weather data item;

[0012] Calculate the global influence coefficient of weather data pairs on imaging data based on the deviation generated between adjacent data pairs.

[0013] A substation fault identification method based on image detection as described above, wherein historical diagnostic data of equipment faults in the substation is collected, and a training data set is generated in combination with the original data set, which is specifically divided into the following sub-steps:

[0014] Vectorize the historical diagnostic data of equipment faults;

[0015] Use the vectorized diagnostic data as the output, and the color features of the historical imaging and the weather data at that time as the input to form input-output pairs one by one;

[0016] Organize these input-output pairs into a data set to form a training data set.

[0017] A substation fault identification method based on image detection as described above, wherein a fault identification model is established using the global influence coefficient and the training data set, which is specifically divided into the following sub-steps:

[0018] Create a fault identification model based on the global influence coefficient;

[0019] Use the training data set to train and optimize the fault identification model;

[0020] Verify the accuracy rate of the fault identification model, and put it into the production environment after reaching the preset standard.

[0021] The present invention also provides a substation fault identification system based on image detection, including: an original data sorting module, a weather influence quantification module, a training data preparation module, a fault identification model establishment module, and a fault identification module;

[0022] The original data sorting module is used to collect historical imaging data in the substation and generate an original data set in combination with the weather data at the imaging acquisition time;

[0023] The weather influence quantification module is used to calculate the global influence coefficient of weather data on imaging data based on the original data set;

[0024] The training data preparation module is used to collect historical diagnostic data of equipment faults in the substation and generate a training data set in combination with the original data set;

[0025] The fault identification model establishment module is used to establish a fault identification model using the global influence coefficient and the training data set;

[0026] A fault identification module, which is used to obtain the imaging data collected by the imaging device in real time, obtain the fault identification result through the trained fault identification model, and forward the fault identification result to the operation and maintenance personnel in the substation.

[0027] The beneficial effects achieved by the present invention are as follows: Quantify the impact of weather on imaging in advance, greatly save the cycle of fault identification and the training volume of the fault identification model, so as to improve the identification accuracy while reducing the consumption of computing resources; It can achieve the effect of real-time identification, further improve the efficiency of equipment maintenance, and reduce equipment loss caused by untimely processing. Description of the Drawings

[0028] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for use in the description of the embodiments or the prior art. Obviously, the drawings in the following description are only some embodiments recorded in the present invention. For those of ordinary skill in the art, other drawings can also be obtained based on these drawings.

[0029] Figure 1 It is a flowchart of a substation fault identification method based on image detection provided in Embodiment 1 of the present application;

[0030] Figure 2 It is a schematic diagram of a substation fault identification system based on image detection provided in Embodiment 2 of the present application. Detailed Embodiments

[0031] The following combines the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, rather than all of them. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative work fall within the protection scope of the present invention.

[0032] Embodiment 1

[0033] As Figure 1 shown, Embodiment 1 of the present application provides a substation fault identification method based on image detection, including:

[0034] Step S10: Collect historical imaging data in the substation and generate an original data set in combination with the weather data at the time of imaging acquisition;

[0035] Weather is the main reason affecting the accuracy of imaging data. Different temperatures and lighting conditions will cause deviations in the imaging of equipment. In this step, the imaging data and the weather data at that time need to be processed into a data pair (p i , E i), and organize it into an original dataset to provide a data basis for subsequent calculations, where p i is the i-th historical imaging data, E i is the weather dataset at the time when the i-th historical imaging data was obtained.

[0036] Step S20: Calculate the global influence coefficient of weather data on imaging data based on the original dataset;

[0037] It is known that weather data will affect the accuracy of imaging data. Quantifying this influence in advance can greatly save the cycle of fault identification and the training volume of the fault identification model, so as to achieve the reduction of computational resource consumption while improving the identification accuracy. It is specifically divided into the following sub-steps:

[0038] Step S21: Extract the color features of each imaging in the original dataset to replace the imaging data in the original data pair;

[0039] Since infrared imaging is usually single-channel, the gray mean value is used here as the color feature of the imaging; the imaging data in the data pair in the original dataset was originally an imaging, and now it is replaced with the color feature of this imaging.

[0040] Step S22: De-duplicate and sort the replaced data pairs according to the values of each weather data item;

[0041] Remove the data pairs with the same values of each weather data item to reduce the computational amount. Sorting is for better observing the change trend of the data and determining the weight of the weather data. Either descending or ascending order is acceptable.

[0042] Step S23: Calculate the global influence coefficient of weather data on imaging data according to the deviation generated between adjacent data pairs;

[0043] The calculation formula of the global influence coefficient λ is expressed as:

[0044] where μ j is the weight (preset value) of the j-th weather data item, e jk is the value of the j-th weather data item in the k-th data pair, e jk+1 is the value of the j-th weather data item in the (k + 1)-th data pair, c k is the imaging data in the k-th data pair (which has been replaced with the color feature of the imaging), c k+1 is the imaging data in the (k + 1)-th data pair, j takes values from 1 to m, m is the total number of weather data items, k takes values from w - 1, w is the total number of data pairs in the original dataset after de-duplication, and λ is the calculation result of the global influence coefficient.

[0045] Step S30: Collect the historical diagnostic data of equipment failures in the substation and generate a training dataset in combination with the original dataset;

[0046] First, vectorize the historical diagnostic data of equipment failures; subsequently, use the vectorized diagnostic data as the output, and the color features of the historical imaging and the weather data at that time as the input when diagnosing to form input-output pairs; finally, organize these input-output pairs into a dataset to form a training dataset.

[0047] Step S40: Establish a fault identification model using the global influence coefficient and the training dataset;

[0048] The global influence coefficient can enable the model to focus on learning the guiding relationship between faults and equipment imaging, greatly reducing the training volume and computational complexity while ensuring accuracy. Specifically:

[0049] Step S41: Create a fault identification model based on the global influence coefficient;

[0050] The data expression of the created fault identification model is:

[0051] where Y is the output result of the model, φ f is the output weight of the f-th layer of the model, d f is the output bias of the f-th layer of the model, x a is the input imaging data, is the j-th item of input weather data, μ j is the weight value of the j-th item of weather data, λ is the global influence coefficient, δ fl is the l-th hidden variable in the f-th layer, l takes values from 1 to L, L is the total number of hidden variables in the f-th layer, j takes values from 1 to m, m is the total number of items of input weather data, and f takes values from 1 to F, F is the total number of layers of the model.

[0052] Step S42: Use the training dataset to train and optimize the fault identification model;

[0053] First, a loss function needs to be designed for the model, and then the parameters in the model are continuously iterated during the training process using the principle of minimum loss to obtain the optimal solutions of the parameters of each layer of the model. The loss function is expressed as:

[0054] where S is the calculation result of the loss value, y uv is the actual classification label of the u-th training sample for the fault v. If the training sample u actually points to the fault v, then y uv takes the value 1, otherwise it takes the value 0, p uvThe probability that the u-th training sample is predicted as fault v, where v ranges from 1 to V (V is the number of identified faults in the substation), and u ranges from 1 to U (U is the total number of training samples).

[0055] Step S43: Verify the accuracy rate of the fault identification model, and put it into the production environment after reaching the preset standard;

[0056] Prepare a verification data set to verify the generalization ability of the model on unfamiliar samples, observe the accuracy rate of the output results. If it is lower than the preset standard, return to step S42, adjust the data training weights and then retrain and optimize; if it has reached the preset standard, put the trained model into the production environment for use.

[0057] Step S50: Obtain the imaging data collected by the imaging device in real time, get the fault identification result through the fault identification model, and forward the fault identification result to the operation and maintenance personnel in the substation;

[0058] Because the fault identification model of this application has few parameters and concentrated attention, the identification speed is much faster than that of traditional models. After obtaining the imaging data, extract the features therein and input them into the model, and the identification result can be obtained immediately, achieving the effect of real-time identification. The operation and maintenance personnel can then perform timely maintenance on the equipment in the substation, further improving the efficiency of equipment maintenance and reducing equipment losses caused by untimely processing.

[0059] Embodiment 2

[0060] As Figure 2 shown, Embodiment 2 of this application provides a substation fault identification system based on image detection, including: an original data sorting module 21, a weather impact quantification module 22, a training data preparation module 23, a fault identification model establishment module 24, and a fault identification module 25;

[0061] The original data sorting module 21 is used to collect the historical imaging data in the substation and generate an original data set in combination with the weather data at the time of imaging acquisition;

[0062] Weather is the main reason affecting the accuracy of imaging data. Different temperatures and illuminations will cause deviations in the imaging of equipment. At this step, the imaging data and the weather data at that time need to be processed into a data pair (p i , E i ), and sorted into an original data set to provide a data basis for subsequent calculations, where p i is the i-th historical imaging data, and E i is the weather data set at the time of acquisition of the i-th historical imaging data.

[0063] The weather impact quantification module 22 is used to calculate the global impact coefficient of weather data on imaging data based on the original data set; specifically including: the original data processing sub-module and the global impact coefficient calculation sub-module;

[0064] 1. The original data processing sub-module is used to extract the color features of each imaging in the original data set to replace the imaging data in the original data pair, and de-duplicate and sort the replaced data pairs according to the values of each weather data item;

[0065] Since infrared imaging is usually single-channel, the gray mean value is used here as the color feature of the imaging; the imaging data in the data pair in the original data set was originally an imaging, and now it is replaced with the color feature of this imaging. Removing the data pairs with the same values of each weather data item can reduce the calculation amount, and sorting is for better observing the change trend of the data, determining the weight of the weather data, and either descending or ascending order is acceptable.

[0066] 2. The global impact coefficient calculation sub-module is used to calculate the global impact coefficient of weather data on imaging data according to the deviation generated between adjacent data pairs;

[0067] The calculation formula of the global impact coefficient λ is expressed as:

[0068] where μj is the weight (preset value) of the j-th weather data item, ejk is the value of the j-th weather data item in the k-th data pair, e jk+1 is the value of the j-th weather data item in the (k + 1)-th data pair, c k is the imaging data in the k-th data pair (which has been replaced with the color feature of the imaging at this time), ck+1 is the imaging data in the (k + 1)-th data pair, j takes values from 1 to m, m is the total number of weather data items, k takes values of w - 1, w is the total number of data pairs in the original data set after de-duplication, and λ is the calculation result of the global impact coefficient.

[0069] The training data preparation module 23 is used to collect the historical diagnostic data of equipment failures in the substation and generate a training data set in combination with the original data set;

[0070] First, vectorize the text of the historical diagnostic data of equipment failures; then, use the vectorized diagnostic data as the output, the color features of the historical imaging based on which the diagnosis is made, and the weather data at that time as the input to form input-output pairs one by one; finally, organize these input-output pairs into a data set to form a training data set.

[0071] The fault identification model establishment module 24 is used to establish a fault identification model by using the global impact coefficient and the training data set; specifically including: the model creation sub-module, the model training sub-module and the model verification sub-module;

[0072] 1. A model creation sub-module for creating a fault identification model using a global influence coefficient;

[0073] The data expression of the created fault identification model is:

[0074] where Y is the output result of the model, φ f is the output weight of the f-th layer of the model, d f is the output bias of the f-th layer of the model, x a is the input imaging data, is the j-th item of input weather data, μ j is the weight value of the j-th item of weather data, λ is the global influence coefficient, δ fl is the l-th hidden variable in the f-th layer, l takes values from 1 to L, L is the total number of hidden variables in the f-th layer, j takes values from 1 to m, m is the total number of items of input weather data, and f takes values from 1 to F, F is the total number of layers of the model.

[0075] 2. A model training sub-module for training and optimizing the fault identification model using a training data set;

[0076] First, a loss function needs to be designed for the model, and then the parameters in the model are continuously iterated during the training process using the principle of minimum loss to obtain the optimal solutions of the parameters of each layer of the model. The loss function is expressed as:

[0077] where S is the calculation result of the loss value, y uv is the actual classification label of the u-th training sample for the fault v. If the u-th training sample actually points to the fault v, then y uv takes the value 1, otherwise it takes the value 0, p uv is the probability that the u-th training sample is predicted to be the fault v. v takes values from 1 to V, V is the number of faults identified in the substation, and u takes values from 1 to U, U is the total number of training samples.

[0078] 3. A model verification sub-module for verifying the accuracy rate of the fault identification model and putting it into the production environment after reaching the preset standard;

[0079] Prepare a verification data set to verify the generalization ability of the model on unfamiliar samples, observe the accuracy rate of the output results. If it is lower than the preset standard, return to the model training sub-module, adjust the data training weights and re-train and optimize; if it has reached the preset standard, put the trained model into the production environment for use.

[0080] The fault identification module 25 is used to obtain the imaging data collected by the imaging device in real time, obtain the fault identification result through the trained fault identification model, and forward the fault identification result to the operation and maintenance personnel in the substation.

[0081] Corresponding to the above embodiments, an embodiment of the present invention provides a computer storage medium, including: at least one memory and at least one processor;

[0082] The memory is used to store one or more program instructions;

[0083] The processor is used to run one or more program instructions to execute a substation fault identification method based on image detection.

[0084] Corresponding to the above embodiments, an embodiment of the present invention provides a computer-readable storage medium. The computer storage medium contains one or more program instructions, and the one or more program instructions are used to be executed by a processor to perform a substation fault identification method based on image detection.

[0085] An embodiment disclosed by the present invention provides a computer-readable storage medium. Computer program instructions are stored in the computer-readable storage medium. When the computer program instructions run on a computer, the computer is enabled to execute the above-mentioned substation fault identification method based on image detection.

[0086] In an embodiment of the present invention, the processor may be an integrated circuit chip with signal processing capabilities. The processor may be a general-purpose processor, a digital signal processor (DSP for short), an application specific integrated circuit (ASIC for short), a field programmable gate array (FPGA for short), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components.

[0087] It can implement or execute the various methods, steps, and logic block diagrams disclosed in the embodiments of the present invention. The general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc. The steps of the method disclosed in combination with the embodiments of the present invention may be directly embodied as being executed by a hardware decoding processor, or executed by a combination of hardware and software modules in the decoding processor. The software module may be located in a mature storage medium in the art such as random access memory, flash memory, read-only memory, programmable read-only memory, or electrically erasable programmable memory, register, etc. The processor reads the information in the storage medium and combines its hardware to complete the steps of the above method.

[0088] The storage medium may be a memory, for example, it may be a volatile memory or a non-volatile memory, or may include both a volatile memory and a non-volatile memory.

[0089] Among them, the non-volatile memory can be a Read-Only Memory (ROM), a Programmable ROM (PROM), an Erasable PROM (EPROM), an Electrically Erasable PROM (EEPROM), or a flash memory.

[0090] The volatile memory can be a Random Access Memory (RAM), which is used as an external cache. By way of example but not limitation, many forms of RAM are available, such as Static RAM (SRAM), Dynamic RAM (DRAM), Synchronous DRAM (SDRAM), Double Data Rate SDRAM (DDR SDRAM), Enhanced SDRAM (ESDRAM), Synchlink DRAM (SLDRAM), and Direct Rambus RAM (DRRAM).

[0091] The storage medium described in the embodiments of the present invention is intended to include but not limited to these and any other suitable types of memories.

[0092] Those skilled in the art should be able to realize that in one or more of the above examples, the functions described in the present invention can be implemented by a combination of hardware and software. When applying software, the corresponding functions can be stored in a computer-readable medium or transmitted as one or more instructions or codes on a computer-readable medium. The computer-readable medium includes computer storage media and communication media, where the communication media includes any medium that facilitates the transfer of a computer program from one place to another. The storage media can be any available medium accessible by a general-purpose or special-purpose computer.

[0093] The specific embodiments described above further elaborate on the purpose, technical solutions, and beneficial effects of the present invention. It should be understood that the above description is only the specific embodiments of the present invention and is not used to limit the protection scope of the present invention. Any modifications, equivalent replacements, improvements, etc. made on the basis of the technical solutions of the present invention should be included in the protection scope of the present invention.

Claims

1. A substation fault identification method based on image detection, characterized in that: include: Step 1: Collect historical imaging data in the substation and generate the original data set by combining it with the weather data at the time of imaging acquisition; Step 2: Calculate the global impact coefficient of weather data on imaging data based on the original data set; Step 3: Collect the historical diagnostic data of equipment failures in the substation and generate a training data set by combining it with the original data set; Step 4: Use the global influence coefficient and training data set to establish a fault identification model; Step 5: Obtain the imaging data collected by the imaging device in real time, obtain the fault identification result through the fault identification model, and forward the fault identification result to the operation and maintenance personnel in the substation.

2. A method for identifying substation faults based on image detection according to claim 1, characterized in that: The global influence coefficient of weather data on imaging data is calculated based on the original data set, which is divided into the following sub-steps: Extract the color features of each image in the original data set to replace the imaging data in the original data pair; The replaced data is deduplicated and sorted according to the values ​​of each weather data; The global influence coefficient of weather data on imaging data is calculated based on the deviations between adjacent data pairs.

3. A method for identifying substation faults based on image detection according to claim 1, characterized in that: Collect the historical diagnostic data of equipment faults in the substation and generate a training data set based on the original data set. The specific steps are as follows: Vectorize the equipment fault history diagnosis data into text; The vectorized diagnostic data is used as output, and the color features of the historical imaging used for diagnosis and the weather data at that time are used as input to form input-output pairs. These input-output pairs are organized into a data set to form a training data set.

4. A method for identifying substation faults based on image detection according to claim 1, characterized in that: The global influence coefficient and the training data set are used to establish a fault identification model, which is divided into the following sub-steps: Create a fault identification model based on the global impact coefficient; Use the training data set to train and tune the fault identification model; Verify the accuracy of the fault identification model and put it into production once it reaches the preset standard.

5. A substation fault identification system based on image detection, characterized in that: include: Raw data sorting module, weather impact quantification module, training data preparation module, fault identification model building module, fault identification module; The raw data sorting module is used to collect historical imaging data in the substation and generate the raw data set in combination with the weather data at the time of imaging acquisition; The weather impact quantification module is used to calculate the global impact coefficient of weather data on imaging data based on the original data set; The training data preparation module is used to collect the equipment fault history diagnosis data in the substation and generate a training data set by combining the original data set; A fault identification model building module is used to build a fault identification model using a global influence coefficient and a training data set; The fault identification module is used to obtain the imaging data collected by the imaging equipment in real time, obtain the fault identification result through the trained fault identification model, and forward the fault identification result to the operation and maintenance personnel in the substation.

6. A substation fault identification system based on image detection according to claim 5, characterized in that: The weather impact quantification module specifically includes: a raw data processing submodule and a global impact coefficient calculation submodule; The raw data processing submodule is used to extract the color features of each image in the original data set to replace the image data in the original data pair, and to remove duplicates and sort the replaced data pairs according to the values ​​of various weather data; The global influence coefficient calculation submodule is used to calculate the global influence coefficient of weather data on imaging data based on the deviations generated between adjacent data pairs.

7. A substation fault identification system based on image detection according to claim 5, characterized in that: The fault identification model building module specifically includes: a model creation submodule, a model training submodule and a model verification submodule; A model creation submodule is used to create a fault identification model using global influence coefficients; The model training submodule is used to train and tune the fault identification model using the training data set; The model verification submodule is used to verify the accuracy of the fault identification model and put it into the production environment after reaching the preset standard.

8. A computer storage medium, characterized in that: include: at least one memory and at least one processor; A memory for storing one or more program instructions; A processor, used to run one or more program instructions to execute a substation fault identification method based on image detection as described in any one of claims 1 to 4.

Images