Power data processing method and device, equipment and medium

Abstract

The invention provides an electric power data processing method and device, equipment and a medium. The method comprises the steps that an electric power data set used for training a target electric power large model is acquired; according to the power data set, training the initial power large model to obtain a target power large model; and processing the to-be-processed power data through the target power large model to obtain a processing result. The system can achieve the automatic inspection of the power system, can improve the inspection efficiency, improves the inspection quality, and improves the operation safety and stability of the power system.

Description

Technical Field

[0001] The embodiments of the present application relate to the field of data processing technology, and in particular, to a method, device, equipment and medium for processing power data. Background Art

[0002] As the basic energy supply system of modern equipment, the power system is widely used in all aspects of life. Therefore, it is very important to ensure the safe, stable and efficient operation of the power system. In related technologies, the power system is usually inspected, evaluated and maintained by manual inspection, so as to timely discover and eliminate the hidden faults of power equipment through inspection of the power system, and ensure the safe, stable and efficient operation of the power system. However, the manual inspection method of the power system has problems such as low inspection efficiency and difficulty in ensuring inspection quality. Summary of the invention

[0003] The embodiments of the present application provide a power data processing method, device, equipment and medium, which can realize automated inspection of the power system, improve inspection efficiency, enhance inspection quality, and thus improve the safety and stability of power system operation.

[0004] In a first aspect, an embodiment of the present application provides a method for processing power data, including:

[0005] Acquire a power data set for training a target power model;

[0006] According to the power data set, an initial power big model is trained to obtain a target power big model;

[0007] The target power big model is used to process the power data to be processed to obtain a processing result.

[0008] In a second aspect, an embodiment of the present application provides a power data processing device, including:

[0009] A data acquisition module, used to acquire a power data set for training a target power large model;

[0010] A model training module, used for training an initial large power model to obtain a target large power model according to the power data set;

[0011] The data processing module is used to process the power data to be processed through the target power model to obtain a processing result.

[0012] In a third aspect, an embodiment of the present application provides an electronic device, including:

[0013] A processor and a memory, wherein the memory is used to store a computer program, and the processor is used to call and run the computer program stored in the memory to execute the power data processing method as described in the first aspect embodiment and various implementation methods mentioned above.

[0014] In a fourth aspect, an embodiment of the present application provides a computer-readable storage medium for storing a computer program, wherein the computer program enables a computer to execute the power data processing method as described in the embodiment of the first aspect and various implementation methods.

[0015] In a fifth aspect, an embodiment of the present application provides a computer program product comprising program instructions. When the program instructions are executed on an electronic device, the electronic device executes the power data processing method as described in the embodiment of the first aspect and various implementation methods.

[0016] The technical solution disclosed in the embodiment of the present application obtains a power data set for training a target power big model, trains the initial power big model according to the power data set to obtain the target power big model, and then processes the power data to be processed by the target power big model to obtain a processing result, thereby realizing automated inspection of the power system, improving inspection efficiency, enhancing inspection quality, and thus improving the safety and stability of the power system operation. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0018] Figure 1 A flowchart of a power data processing method provided in an embodiment of the present application;

[0019] Figure 2 A flowchart for obtaining a power data set provided in an embodiment of the present application;

[0020] Figure 3 A flowchart of another power data processing method provided in an embodiment of the present application;

[0021] Figure 4 A schematic block diagram of a power data processing device provided in an embodiment of the present application;

[0022] Figure 5 A schematic block diagram of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0023] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. According to the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of this application.

[0024] It should be noted that the terms "first", "second", etc. in the specification and claims of the present application and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments of the present application described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions, for example, a process, method, system, product or server that includes a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0025] In the embodiments of the present application, words such as "exemplary" or "for example" are used to indicate examples, illustrations or descriptions. Any embodiment or scheme described as "exemplary" or "for example" in the embodiments of the present application should not be interpreted as being more preferred or more advantageous than other embodiments or schemes. Specifically, the use of words such as "exemplary" or "for example" is intended to present related concepts in a specific way.

[0026] In the description of the embodiments of the present application, unless otherwise specified, "multiple" means two or more than two, that is, at least two. "At least one" means one or more than one, that is, one or more. "Any" means any one or any several.

[0027] As mentioned above, in the related art, the power system is usually inspected, evaluated and maintained by manual inspection, so as to timely discover and eliminate the hidden faults of power equipment through power system inspection, and ensure the safe, stable and efficient operation of the power system. However, the above manual inspection method has problems such as low inspection efficiency and difficulty in ensuring inspection quality.

[0028] In response to the above technical problems, the present application provides a power data processing method, device, equipment and medium to solve the problems of low inspection efficiency and difficulty in ensuring inspection quality when inspecting, evaluating and maintaining the power system through manual inspection.

[0029] The technical solutions provided by the embodiments of the present application are described in detail below through some embodiments. The embodiments described below can be combined with each other, and the same or similar concepts or processes may not be repeated in some embodiments.

[0030] Figure 1 A flowchart of a power data processing method provided in an embodiment of the present application. The power data processing method in the embodiment of the present application can be executed by a power data processing device, which can be composed of hardware and / or software and can be integrated into an electronic device. The electronic device can be any hardware device with a data processing function. In the present application, the electronic device can be, but is not limited to: a personal computer, a tablet computer, a laptop computer, etc., and the present application does not specifically limit the type of electronic device.

[0031] It should be noted that the electronic device of the present application may include a machine learning operation (MLOps) platform, and the power data processing method of the present application can be executed through the MLOps.

[0032] The above-mentioned MLOps platform is a practical method that applies DevOps (development and operation) principles to machine learning projects. It aims to improve the efficiency of production deployment and operational management of machine learning models through automation, monitoring and collaboration. The MLOps platform can include stages such as data processing, model training, model deployment, model monitoring and model update, and is mainly used to promote the rapid iteration of machine learning models in production and ensure the rapid deployment and efficient and stable operation of machine learning systems.

[0033] like Figure 1 As shown, the method may include the following steps:

[0034] S101, obtaining a power data set for training a target power model.

[0035] In the present application, the power data set includes multiple power data, and the type of the power data corresponds to the function implemented by the target power big model. For example, when the target power big model is a power defect detection big model, the power data in the power data set may be the operation picture data of the power equipment (grid equipment) during operation; for another example, when the target power big model is a power anomaly detection big model, the power data in the power data set may be the voltage value, current value, etc. during the operation of the power equipment. That is, the power data in the power data set obtained by the present application is determined based on the target power big model to be trained.

[0036] In some optional embodiments, the power data set can be obtained from an open source database storing power data; or, the power data set can be obtained by collecting data from power equipment in operation in the power system through a data acquisition device, etc. The present application does not impose any specific restrictions on the implementation method of obtaining the power data set.

[0037] The data acquisition equipment may be a camera on a drone, a fixed camera installed near each power grid device, other types of acquisition equipment installed on each power grid device, and a handheld acquisition equipment. Other types of acquisition equipment may be, but are not limited to, current sensors, voltage sensors, power sensors, and frequency sensors.

[0038] After obtaining the power data set, the present application can also add label information to each power data in the power data set to lay the foundation for subsequent training of the target power model. Optionally, when adding label information to each power data in the power data set, it can be achieved by manual addition and / or automatic addition, and the present application does not make specific restrictions on this.

[0039] S102, training the initial power big model according to the power data set to obtain the target power big model.

[0040] It should be understood that the above-mentioned initial power model refers to any model that has not been trained and is applied to the power system. For example, a fault detection model, a defect detection model, and a load prediction model. That is, the initial power model is a power model in an initial state.

[0041] In the present application, when the initial power big model is a big model that supports power fault detection, the target power big model obtained by training the initial power big model is a power fault detection big model; when the initial power big model is a big model that supports power defect detection, the target power big model obtained by training the initial power big model is a power defect detection big model; when the initial power big model is a big model that supports power anomaly detection, the target power big model obtained by training the initial power big model is a power anomaly detection big model, and so on.

[0042] That is, the target power big model in the present application may be at least one of a power fault detection big model, a power anomaly detection big model, a power defect detection big model, and a power load prediction big model.

[0043] In some optional embodiments, training the initial power model to obtain the target power model may include the following steps:

[0044] In the first step, according to the training requirements of the target power big model, the target loading method of the power data in the power data set is selected from the data loading component, the initial power big model corresponding to the power data set is obtained from the big model component, and the target tuning index for the target power big model is selected from the tuning index component.

[0045] It should be understood that when training target power big models with different functions, the target loading method selected from the data loading component, the initial power big model obtained from the big model component, and the target tuning index selected from the tuning index component are also different.

[0046] Among them, the data loading component can provide multiple training data loading methods, so that the MLOps platform can select the target loading method from the data loading component to load the power data set into the initial power big model according to the training requirements of the target power big model. The big model component can provide initial power big models with different functions, so the MLOps platform can select any initial power big model from the big model component according to the training requirements of the target power big model. The tuning indicator component can provide multiple optional tuning indicators, such as accuracy, generalization ability, and real-time performance, so the MLOps platform can select the required target tuning indicators from the tuning indicator component according to the training requirements of the target power big model.

[0047] The second step is to load the power data in the power data set into the initial power big model according to the target loading method, so as to analyze and process each power data through the initial power big model to obtain the prediction result of each power data.

[0048] The third step is to determine the loss value between the label information of each power data and the prediction result, and obtain the total loss value based on the loss value corresponding to each power data.

[0049] The total loss value may be obtained according to the loss value corresponding to each power data, by adding up the loss values ​​corresponding to all power data and taking the sum as the total loss value.

[0050] The fourth step is to train the initial power model according to the total loss value to obtain the target power model.

[0051] Optionally, the total loss value can be propagated back layer by layer to each network layer in the initial power model through a back-propagation algorithm to adjust the model parameters in the initial power model, such as adjusting the weight parameters of each network layer in the initial power model.

[0052] The total loss value is propagated back to each network layer in the initial power model layer by layer through the back propagation algorithm to adjust the model parameters in the initial power model. The gradient of the loss function to the model parameters can be calculated during the back propagation process, and then an optimization algorithm, such as a gradient descent algorithm, is used to update the model parameters in the initial power model according to the calculated gradient.

[0053] That is, the present application performs reverse training on the initial power model according to the total loss value to obtain the target power model.

[0054] In some optional embodiments, the training of the initial power model according to the total loss value to obtain the target power model may include the following steps:

[0055] Step S1, determine whether the total loss value is greater than or equal to a preset threshold. If the total loss value is greater than or equal to the preset threshold, execute step S2, otherwise execute step S3.

[0056] Step S2, if the total loss value is greater than the preset threshold, it means that the processing result of the initial power model on the power data in the power data set is inaccurate, that is, the initial power model has not been trained yet. At this time, the initial power model can be reversely trained based on the total loss value, and it is determined again whether the new total loss value is greater than or equal to the preset threshold. If the new total loss value is still greater than or equal to the preset threshold, repeat step S2 until the training stop condition is met. If the latest total loss value is less than the preset threshold, it means that the latest trained power model can accurately process the power data in the power data set. At this time, it means that the training of the initial power model is completed, and the latest trained power model is determined as the target power model.

[0057] In the present application, the training stop condition may be that the total loss value is less than a preset threshold, the number of training times is greater than a preset number of times, or the total loss value tends to a certain stable value.

[0058] Among them, the preset threshold and the preset number of times are both adjustable parameters, which can be flexibly set according to the prediction accuracy of the large power model, and no restrictions are imposed on them here.

[0059] That is, when a processing result with higher accuracy is required, the present application may set the preset threshold relatively smaller or the preset number relatively larger; conversely, the preset threshold may be set relatively larger or the preset number relatively smaller. The specific setting method is flexibly set according to the image fusion model training requirements.

[0060] Step S3, if the total loss value is less than the preset threshold, it means that the initial power model has a more accurate processing result on the power data in the power data set. At this time, the training operation of the initial power model can be ended and the initial power model can be used as the target power model.

[0061] In some optional embodiments, before training the initial power model based on the power data set, the present application can determine the training parameter information for training the target power model from the training parameter definition component according to the training requirements of the target power model. That is, when training target power models with different functions, the determined training parameter information is also different.

[0062] Among them, the training parameter information can be understood as hyperparameter information, such as data source, cyclic training batch, computing resources, optimizer, and loss strategy. In this way, by defining hyperparameters before training the target power model, it can be ensured that the target power model finally trained can be effectively learned and optimized, while improving the training efficiency and model performance, and avoiding problems such as poor performance of the trained target power model, low training efficiency, overfitting or underfitting.

[0063] In some optional embodiments, the present application may configure a corresponding preprocessing method according to the task type of the target power model before training the initial power model according to the power data set. Then, according to the preprocessing method, the power data set is preprocessed to obtain a new power data set, which can improve the quality of the training data set used to train the target power model, enhance the consistency of the data, and adapt to the requirements of the large model, thereby helping to train a more accurate and stable target power model.

[0064] Exemplarily, according to the task type of the target power model, the corresponding preprocessing method can be configured as follows:

[0065] When the task type of the target power model is an image classification task, the configured preprocessing method can be to normalize the input image data according to the description prompt words and perform a one-hot encoding operation on the classification category.

[0066] When the task type of the target power model is a target detection task, the configured preprocessing method can be to normalize the input image data, adjust the proportion of the annotation box relative to the width and height of the original image according to a preset adjustment method, and store the category information of the annotation box.

[0067] The preset adjustment method may be to enlarge the annotation box according to a preset ratio, or to reduce the annotation box according to a preset ratio, and this application does not impose any specific limitation on this.

[0068] The above classification categories are pre-defined based on image classification tasks. For example, in object recognition tasks, the classification categories may include: animals, plants, buildings, and natural landscapes, etc. For another example, in autonomous driving recognition tasks, the classification categories may include: pedestrians, vehicles, and roads, etc.

[0069] The above unique hot encoding can be expressed as One-hot encoding. One-hot encoding is a method of converting categorical categories into binary vectors. In One-hot encoding, each category value is converted into a binary vector whose length is equal to the number of all possible categories, in which only one element is 1 and all other elements are 0. The position of the element with 1 represents the specific value of the categorical variable.

[0070] That is to say, before training the initial power big model according to the power data set, the present application can determine the training parameter information for training the target power big model and / or pre-process the power data set to obtain a new power data set. Furthermore, the initial power big model can be trained according to the determined training parameter information and / or the new power data set to obtain the target power big model, thereby ensuring that the big model can learn and predict efficiently and accurately, thereby improving the performance of the target power big model.

[0071] S103, processing the power data to be processed by the target power big model to obtain a processing result.

[0072] The power data to be processed may be the power data obtained by collecting data from any power equipment in the power system through a data collection device.

[0073] Optionally, after obtaining the target power model, the present application may use the target power model to process the power data to be processed sent by the data acquisition device to obtain a processing result output by the target power model for the data to be processed.

[0074] In the present application, the to-be-processed power data sent by each data acquisition device also carries the identification information of each data acquisition device. And because the power equipment corresponding to each data acquisition device is fixed, that is, each data acquisition device corresponds to one power equipment, the identity of the corresponding power equipment can be determined according to the identification information of each data acquisition device. Among them, the identification information of the data acquisition device can be understood as the unique information of the data acquisition device, such as the serial number of the data acquisition device or the name of the data acquisition device.

[0075] Thus, when the target power model is a power fault detection model, the target power model outputs the processing result for the data to be processed as follows: the processing result of whether the power equipment corresponding to the data to be processed has a fault. When any power equipment has a fault, the processing result may be that the power equipment XX has a fault; when the task power equipment has no fault, the processing result may be that the power equipment YY is operating normally.

[0076] When the target power model is a power defect detection model, the target power model outputs the processing result for the data to be processed as follows: the processing result of whether the power equipment corresponding to the data to be processed has defects. When any power equipment has defects, the processing result may be that power equipment XX has defects; when the task power equipment has no defects, the processing result may be that power equipment YY is operating normally, and so on.

[0077] The technical solution disclosed in the embodiment of the present application obtains a power data set for training a target power big model, trains the initial power big model according to the power data set to obtain the target power big model, and then processes the power data to be processed by the target power big model to obtain a processing result, thereby realizing automated inspection of the power system, improving inspection efficiency, enhancing inspection quality, and thus improving the safety and stability of the power system operation.

[0078] Based on the above embodiments, considering that during the online use of the target power model, new data will be continuously generated, old data will be outdated, and data distribution will change, resulting in a decrease in the performance of the target power model. To this end, the present application optionally updates the target power model according to a preset time period to maintain or improve the performance of the target power model and improve the generalization ability of the target power model.

[0079] The preset time period can be flexibly adjusted according to the update requirements of the target power model, and this application does not impose specific restrictions on this. That is, when the update requirements for the target power model are high, the preset time period can be set to be smaller; when the update requirements for the target power model are low, the preset time period can be set to be larger.

[0080] In some optional embodiments, during the online use of the target power model, training data for updating the target power model can be continuously obtained, and when the model update time is reached, that is, when the preset time period is reached, the target power model is updated according to the training data to obtain an updated target power model. Then, the updated target power model is used to process the power data to be processed to obtain a processing result.

[0081] It should be understood that in the present application, the target power model is updated according to the training data. The training data can be used to fine-tune the target power model so that the target power model can quickly adapt to the new data distribution; or, the target power model can be partially modified for specific tasks, and most of the knowledge of the target power model can be retained, while the information of the specific task is updated; or, the target power model can be retrained according to the training data to perform large-scale updates on the target power model. The present application does not impose any specific restrictions on the update operation of the target power model.

[0082] Based on the above embodiments, the method disclosed in this application for obtaining a power data set for training a target power model is further explained. Figure 2 As shown, Figure 1 The S101 may include the following steps S101-1 and S101-2:

[0083] S101 - 1 , obtaining power data of at least one power device.

[0084] S101 - 2 , obtaining a power data set for training a target power model based on power data of at least one power device.

[0085] The power data of the above-mentioned power equipment may include the identification information of the power equipment, the data collection time and the power operation data. Among them, the identification information of the power equipment is the unique information of the power equipment, that is, the power equipment can be uniquely identified by the identification information. The data collection time can be understood as the time when the power data is obtained. The power operation data may include various electrical signals when the power equipment is running, such as current, voltage and operating power.

[0086] In some optional embodiments, when acquiring power data of at least one power device in the power system, it can be achieved through a data acquisition device. In order to avoid the leakage of power data in the power field, after acquiring the power data of at least one power device in the power system, the data acquisition device can transmit the power data to a data storage module through an encrypted network to achieve contactless data acquisition and data storage, thereby ensuring the privacy and security of the power data. Among them, the data storage module can be understood as a functional module located in an electronic device and used to store power data.

[0087] Afterwards, the power data set for training the target power model can be obtained according to the power data stored in the data storage module. In some optional embodiments, the power data set for training the target power model can be obtained by the following steps:

[0088] Step 11: Screening and processing the power data of at least one power device to obtain first candidate power data.

[0089] The above screening process includes one or more of removing abnormal power data, removing duplicate power data, removing low-quality power data, and removing unusable power data.

[0090] Optionally, the present application may screen and process each power data stored in the data storage module to separate abnormal, duplicate, low-quality and unavailable power data to obtain more accurate and reliable first candidate power data, thereby improving the quality and reliability of the power data.

[0091] Step 12: verify the first candidate power data to obtain second candidate power data, wherein the amount of the second candidate power data is greater than or equal to the first candidate operation data.

[0092] In some embodiments, the hash value of each first candidate power data can be calculated according to a hash algorithm, and the hash value can be stored in a data storage module. When there is new power data to be stored in the data storage module, the present application can use the above-mentioned hash algorithm to calculate the hash value of the new power data. Then compare the hash value of the new power data with the hash value of each first candidate power data in the data storage module. If the hash value of the new power data is consistent with the hash value of any first candidate power data, it means that the same power data has been stored in the data storage module. At this time, the new power data and the hash value of the new power data are discarded; if the hash value of the new power data is inconsistent with the hash value of each first candidate power data, it means that the new power data is not stored in the data storage module. At this time, the new power data and the hash value of the new power data are stored as the first candidate power data in the data storage module, thereby ensuring the uniqueness of the power data stored in the data storage module.

[0093] That is, the present application performs verification processing on the first candidate power data, and when it is necessary to store new power data in the data storage module, the first candidate power data stored in the data storage module is updated according to the hash value of each first candidate power data and the hash value of the new power data. When the hash value of the new power data is consistent with the hash value of any first candidate power data, it means that the same power data has been stored in the data storage module, then there is no need to store the new power data and the hash value of the new power data in the data storage module, that is, there is no need to update the first candidate power data stored in the data storage module. At this time, the second candidate power data is the first candidate power data stored in the data storage module. When the hash value of the new power data is inconsistent with the hash value of each first candidate power data, it means that the new power data is not stored in the data storage module, then the new power data and the hash value of the new power data need to be stored in the data storage module, that is, the first candidate power data stored in the data storage module needs to be updated. At this time, the second candidate power data is the first candidate power data and the new power data stored in the data storage module.

[0094] In the present application, the above hash algorithm may be but is not limited to MD5 algorithm, SHA-1 and SHA-256, and no limitation is imposed on it here.

[0095] Step 13: label the second candidate power data to obtain a power data set for training the target power model.

[0096] Considering that when training the target power big model, it is necessary to determine the difference between the label information of the power data and the prediction results of the big model output for the power data. Then, the initial power big model is trained according to the difference to obtain the target power big model. Therefore, the present application can perform labeling processing on each second candidate power data in the data storage module according to the target power big model training task to add corresponding label information to each second candidate power data, so that when the initial power big model is trained, the initial power big model can understand and utilize the data, thereby performing model training and prediction tasks.

[0097] Exemplarily, when the training task of the target power model is to train a power defect detection model, each second candidate power data in the data storage module can be annotated with defect content, that is, label information of whether there is a defect can be added to each second candidate power data.

[0098] In some optional embodiments, each second candidate power data in the data storage module is labeled, and each second candidate power data can be labeled manually to add label information to each second candidate power data. Alternatively, each second candidate power data can be labeled automatically to add label information to each second candidate power data.

[0099] The above-mentioned automatic labeling method can be to use an algorithm or model that supports the automatic labeling function to label each second candidate power data, and the present application does not impose any specific restrictions on this.

[0100] In some optional embodiments, considering that each second candidate power data is labeled by automatic labeling, there may be problems with inaccurate labeling. Therefore, when the present application labels each second candidate power data in the data storage module, it can optionally be labeled by automatic labeling first, and then each second candidate power data after automatic labeling can be manually checked. When it is found that the label information of any second candidate power data is incorrectly labeled or omitted, the label information of the second candidate power data is modified or label information is added to the second candidate power data. This can improve the labeling quality of the power data and ensure the accuracy of the labeling of the power data, thereby helping to improve the recognition accuracy and generalization ability of the training target power model.

[0101] After labeling each second candidate power data in the data storage module, each second candidate power data can be used as power data for training the target power model. Thus, a power data set can be obtained from all power data.

[0102] The embodiment of the present application can improve the performance of training the target power model and enhance the robustness of the target power model by screening, verifying and labeling the power data of at least one power device, thereby improving the actual application effect of the target power model.

[0103] Combine the following Figure 3 , further explain the power data processing method disclosed in this application. Figure 3 As shown, the power data processing method may include the following steps:

[0104] S201, obtaining a power data set for training a target power model.

[0105] S202: Based on the power data set, the initial power model is trained to obtain the target power model.

[0106] S203, performing model evaluation on the target power large model to obtain evaluation results.

[0107] After training the target power model, the present application may optionally first evaluate the target power model to determine the performance of the target power model on unknown data, so as to understand the generalization ability, robustness and applicability of the target power model.

[0108] Optionally, a test data set for evaluating the target power model can be obtained, and the test data set can be input into the target power model, so that the target power model analyzes and processes each test power data in the test data set, and outputs the prediction result of each test power data. The test data set can be a sub-data set divided from the power data set obtained above, or a re-acquired power data set, and the power data is used as the test data set, and label information is added to each test power data in the test data set.

[0109] Then, the evaluation result of the target power model is determined according to the label information of each test power data and the corresponding prediction result. For example, the evaluation index such as the accuracy of the power model is determined.

[0110] Considering that the present application can update the target power model, each time the target power model is updated, the updated target power model can be evaluated, so that evaluation results of multiple versions of the target power model can be obtained.

[0111] Next, the target power model, the task description information of the target power model, the training information of the target power model, the evaluation results of the target power model, and the information to be evaluated for the target power model can be input into the model evaluation module to obtain the evaluation results for the target power model output by the model evaluation module.

[0112] The information to be evaluated for the target power model can be understood as outputting the automated analysis results, improvement suggestions and other information that needs to be known for the target power model based on the input information through the model evaluation module.

[0113] For example, comment on the training process of the target power model, compare and comment on the current and historical evaluation results, and provide one or more improvement suggestions and foreseeable effects based on the current evaluation results.

[0114] In addition, the above-mentioned model evaluation module can be any large language model that supports natural speech processing, or any content generation module that supports automatic generation of new content through artificial intelligence technology, such as an AIGC module, etc., which is not specifically limited here. In addition, the model evaluation module has knowledge in the power field.

[0115] In some optional embodiments, the target power big model, task description information of the target power big model, training information of the target power big model, and information to be evaluated for the target power big model are input into the model evaluation module, so that the model evaluation module processes the above input information and outputs an evaluation result for the target power big model based on the information to be evaluated for the target power big model.

[0116] For example, assume that the information output to the model evaluation module includes: the target power model is the target power defect detection model; the task description information of the target power model is that this training is a target defect detection task for the power distribution network line, and 18 targets such as tower cracks, insulator contamination, etc. in the tower will be identified; the training information of the target power model is that there are 30,000 training data, 5,000 verification data, and 10,000 evaluation data. The input image size is 1280. The training adopts the Stochastic Gradient Descent (SGD) algorithm for optimization, the batch size is 8, and the optimization target is focal loss + minimum absolute error (L1 loss) is minimized, and the optimal model selects the maximum AP value; the evaluation results for the target power model are as follows: The target indicator of version v1.2.1 is..., the target indicator of version v1.2.2 is..., and the target indicator of this time is...; the evaluation information for the target power model is Now you are a professional power industry expert and artificial intelligence expert. You need to integrate the above information and give the output information I want, including comments on the training process, comparative comments on the current and historical evaluation results, and based on the results of this evaluation, give improvement suggestions and foreseeable effects, as well as other information you want me to know.

[0117] Furthermore, the model evaluation module integrates the above information and outputs the review information on the training process of the target power defect detection model, the comparison and review information of the current and historical evaluation results, and provides improvement suggestions and foreseeable effects and other information that needs to be known.

[0118] S204: When the evaluation result meets the preset conditions, the target power model is deployed in the target terminal device.

[0119] S205: When the evaluation result does not meet the preset conditions, the target power model is optimized.

[0120] S206, processing the power data to be processed by the target power model or the optimized target power model to obtain a processing result.

[0121] The preset conditions can be flexibly set according to the evaluation requirements of the target power model, and no specific restrictions are imposed on them here. For example, if the evaluation results of the target power model are good, it can be put into use online.

[0122] The target terminal device may be a power device in the power system or a monitoring platform in an electronic device that supports power system inspection.

[0123] Optionally, after obtaining the evaluation results output by the model evaluation module, the present application may analyze the evaluation content for the target power model in the evaluation results to determine whether the target power model meets the preset conditions. If it is determined that the target power model meets the preset conditions, such as the evaluation results of the target power model are good and can be used online, the target power model can be deployed to the target terminal device for online use. If it is determined that the target power model does not meet the preset conditions, such as the interpretability of the target power model evaluation results is poor, the target power model is optimized according to the improvement suggestions given in the evaluation results, that is, the target power model is adjusted in a targeted manner. In addition, after the target power model is adjusted, the optimized target power model is evaluated again through the model evaluation module, and the above optimization and evaluation operations are repeated until the latest optimized target power model meets the preset conditions, and then the latest optimized target power model is deployed to the target terminal device to process the power data to be processed.

[0124] In this application, the above-mentioned deployment of the target power model to the target terminal device may include the following situations:

[0125] In the first case, when the target terminal device is located in the power system, the target power big model can be deployed to the target terminal device through MLOps.

[0126] In the second case, when the target terminal device is not located in the power system, the deployment files and deployment environment required for the target power model deployment can be exported in the form of an installation package, so that the staff can deploy the installation package to the target terminal device through manual deployment.

[0127] In the third case, when the target terminal device is not located in the power system, the deployment files and deployment environment required for the target power large model deployment can be sent to the target terminal device via wireless transmission, so that the target terminal device can be automatically deployed according to the deployment files and deployment environment required for the target power large model deployment.

[0128] In some embodiments, before deploying the target power big model to the target terminal device, the deployment hyperparameters required for the deployment of the target power big model are optionally selected, such as online threshold, visualization type and other parameters, so that the target power big model is deployed according to the deployment hyperparameter information to ensure that the target power big model can run efficiently and accurately in the actual environment.

[0129] The technical solution disclosed in the embodiment of the present application obtains an electric power data set for training the target electric power big model, trains the initial electric power big model according to the electric power data set to obtain the target electric power big model, and then processes the electric power data to be processed by the target electric power big model to obtain the processing result, so as to realize the automatic inspection of the electric power system, improve the inspection efficiency, improve the inspection quality, and thus improve the safety and stability of the operation of the electric power system. In addition, before using the target electric power big model online, by evaluating the target electric power big model, the performance of the target electric power big model used online in the actual environment can achieve the expected effect, reduce the occurrence of service errors caused by model problems, and ensure that the target big model can maintain stable performance in the face of different electric power data and scenarios, thereby improving the user experience.

[0130] Please refer to the attached Figure 4 , a power data processing device proposed in an embodiment of the present application is described. Figure 4 As shown, the power data processing device 300 includes: a data acquisition module 310 , a model training module 320 and a data processing module 330 .

[0131] The data acquisition module 310 is used to acquire a power data set for training a target power model;

[0132] A model training module 320 is used to train the initial power model according to the power data set to obtain a target power model;

[0133] The data processing module 330 is used to process the power data to be processed through the target power model to obtain a processing result.

[0134] In an optional implementation of an embodiment of the present application, the target power big model is at least one of a power fault detection big model, a power anomaly detection big model, a power defect detection big model and a power load prediction big model.

[0135] In an optional implementation of the embodiment of the present application, the data acquisition module 310 includes:

[0136] A first acquisition unit, configured to acquire power data of at least one power device;

[0137] The second acquisition unit is used to obtain the power data set for training the target power model according to the power data of the at least one power device.

[0138] An optional implementation method of an embodiment of the present application, the second acquisition unit is specifically used to: screen the power data of the at least one power device to obtain first candidate power data; verify the first candidate power data to obtain second candidate power data, wherein the number of the second candidate power data is greater than or equal to the first candidate power data; and label the second candidate power data to obtain the power data set used to train the target power large model.

[0139] In an optional implementation of the embodiment of the present application, the power data processing device 300 further includes:

[0140] An information determination module, which determines training parameter information for training a target power model;

[0141] And / or, a preprocessing module, used to preprocess the power data set to obtain a new power data set.

[0142] In an optional implementation of the embodiment of the present application, the power data processing device 300 further includes:

[0143] An evaluation module, used to perform model evaluation on the target power model to obtain an evaluation result;

[0144] The control module is used to deploy the target power model in the target terminal device when the evaluation result meets the preset conditions; when the evaluation result does not meet the preset conditions, optimize the target power model.

[0145] An optional implementation method of an embodiment of the present application is an evaluation module, which is specifically used to input the target power big model, the task description information of the target power big model, the training information of the target power big model, the evaluation results of the target power big model, and the information to be evaluated for the target power big model into the model evaluation module to obtain the evaluation result of the target power big model output by the model evaluation module.

[0146] In an optional implementation of the embodiment of the present application, the power data processing device 500 further includes:

[0147] The updating module is used to update the target power model according to a preset time period.

[0148] It should be understood that the device embodiment and the aforementioned method embodiment may correspond to each other, and similar descriptions may refer to the method embodiment. To avoid repetition, they will not be described here. Specifically, Figure 4 The device 300 shown can perform Figure 1 The corresponding method embodiments, and the aforementioned and other operations and / or functions of each module in the device 300 are respectively to implement Figure 1 For the sake of brevity, the corresponding processes in each method are not repeated here.

[0149] In the above, the device 300 of the embodiment of the present application is described from the perspective of the functional module in conjunction with the accompanying drawings. It should be understood that the functional module can be implemented in hardware form, can be implemented by instructions in software form, and can also be implemented by a combination of hardware and software modules. Specifically, the steps of the first aspect method embodiment in the embodiment of the present application can be completed by the hardware integrated logic circuit and / or software form instructions in the processor, and the steps of the first aspect method disclosed in conjunction with the embodiment of the present application can be directly embodied as a hardware decoding processor to perform, or a combination of hardware and software modules in the decoding processor to perform. Optionally, the software module can be located in a mature storage medium in the field such as a random access memory, a flash memory, a read-only memory, a programmable read-only memory, an electrically erasable programmable memory, a register, etc. The storage medium is located in a memory, and the processor reads the information in the memory, and completes the steps in the above-mentioned first aspect method embodiment in conjunction with its hardware.

[0150] Figure 5 A schematic block diagram of an electronic device provided in an embodiment of the present application. Figure 5 As shown, the electronic device 400 may include: a memory 410 and a processor 420, wherein the memory 410 is used to store a computer program and transmit the program code to the processor 420. In other words, the processor 420 may call and run the computer program from the memory 410 to implement the power data processing method described in the first aspect of the embodiment of the present application.

[0151] For example, the processor 420 may be configured to execute the above-mentioned power data processing method according to the instructions in the computer program.

[0152] In some embodiments of the present application, the processor 420 may include but is not limited to:

[0153] General-purpose processor, digital signal processor (DSP), application-specific integrated circuit (ASIC), field programmable gate array (FPGA) or other programmable logic device, discrete gate or transistor logic device, discrete hardware components, etc.

[0154] In some embodiments of the present application, the memory 410 includes but is not limited to:

[0155] Volatile memory and / or non-volatile memory. Among them, the non-volatile memory can be read-only memory (ROM), programmable ROM (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM) or flash memory. The volatile memory can be random access memory (RAM), which is used as an external cache. By way of example and not limitation, many forms of RAM are available, such as static random access memory (SRAM), dynamic random access memory (DRAM), synchronous dynamic random access memory (SDRAM), double data rate synchronous dynamic random access memory (DDR SDRAM), enhanced synchronous dynamic random access memory (ESDRAM), synchronous link DRAM (SLDRAM), and direct RAM bus random access memory (DR RAM).

[0156] In some embodiments of the present application, the computer program may be divided into one or more modules, which are stored in the memory 410 and executed by the processor 420 to complete the power data processing method provided by the present application. The one or more modules may be a series of computer program instruction segments capable of completing specific functions, and the instruction segments are used to describe the execution process of the computer program in the electronic device.

[0157] like Figure 5 As shown, the electronic device 400 may further include:

[0158] The transceiver 430 may be connected to the processor 420 or the memory 410 .

[0159] The processor 420 may control the transceiver 430 to communicate with other devices, specifically, to send information or data to other devices, or to receive information or data sent by other devices. The transceiver 430 may include a transmitter and a receiver. The transceiver 430 may further include an antenna, and the number of antennas may be one or more.

[0160] It should be understood that the various components in the electronic device are connected via a bus system, wherein the bus system includes not only a data bus but also a power bus, a control bus and a status signal bus.

[0161] The present application also provides a computer storage medium on which a computer program is stored. When the computer program is executed by a computer, the computer is enabled to execute the power data processing method described in the first aspect above.

[0162] The present application also provides a computer program product including program instructions. When the program instructions are executed on an electronic device, the electronic device executes the power data processing method described in the first aspect above.

[0163] When software is used for implementation, it can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, the process or function described in the embodiment of the present application is generated in whole or in part. The computer may be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions may be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the computer instructions may be transmitted from a website site, a computer, a server, or a data center by wired (e.g., coaxial cable, optical fiber, digital subscriber line (digital subscriber line, DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) mode to another website site, computer, server, or data center. The computer-readable storage medium may be any available medium that a computer can access or a data storage device such as a server or a data center that includes one or more available media integrations. The available medium may be a magnetic medium (e.g., a floppy disk, a hard disk, a tape), an optical medium (e.g., a digital video disc (digital video disc, DVD)), or a semiconductor medium (e.g., a solid state disk (solid state disk, SSD)), etc.

[0164] Those of ordinary skill in the art will appreciate that the modules and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0165] In the several embodiments provided in the present application, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of the module is only a logical function division. There may be other division methods in actual implementation, such as multiple modules or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or modules, which can be electrical, mechanical or other forms.

[0166] The modules described as separate components may or may not be physically separated, and the components displayed as modules may or may not be physical modules, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. For example, each functional module in each embodiment of the present application may be integrated into a processing module, or each module may exist physically separately, or two or more modules may be integrated into one module.

[0167] In the embodiments of the present application, the term "module" or "unit" refers to a computer program or a part of a computer program with a predetermined function, and works together with other related parts to achieve a predetermined goal, and can be implemented in whole or in part by using software, hardware (such as processing circuits or memories) or a combination thereof. Similarly, a processor (or multiple processors or memories) can be used to implement one or more modules or units. In addition, each module or unit can be part of an overall module or unit that includes the function of the module or unit.

[0168] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art who is familiar with the present technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

Claims

1. A method for processing power data, characterized in that: include: Acquire a power data set for training a target power model; According to the power data set, an initial power big model is trained to obtain a target power big model; The target power big model is used to process the power data to be processed to obtain a processing result.

2. The method according to claim 1, characterized in that The target power big model is at least one of a power fault detection big model, a power anomaly detection big model, a power defect detection big model and a power load prediction big model.

3. The method according to claim 1, characterized in that The step of obtaining a power data set for training a target power model includes: Acquiring power data of at least one power device; The power data set for training the target power model is obtained according to the power data of the at least one power device.

4. The method according to claim 3, characterized in that The step of obtaining the power data set for training the target power model according to the power data of the at least one power device includes: Screening the power data of the at least one power device to obtain first candidate power data; Performing verification processing on the first candidate power data to obtain second candidate power data, wherein the number of the second candidate power data is greater than or equal to the number of the first candidate power data; The second candidate power data is labeled to obtain the power data set used to train the target power large model.

5. The method according to claim 1, characterized in that The method further comprises: Determining training parameter information for training a target power model; And / or, preprocessing the power data set to obtain a new power data set.

6. The method according to claim 1, characterized in that The method further comprises: Performing model evaluation on the target large power model to obtain an evaluation result; When the evaluation result meets the preset conditions, deploying the target power model in the target terminal device; When the evaluation result does not meet the preset conditions, the target power model is optimized.

7. The method according to claim 6, characterized in that The step of performing a model evaluation on the target power model to obtain an evaluation result includes: The target power big model, the task description information of the target power big model, the training information of the target power big model, the evaluation results of the target power big model and the information to be evaluated for the target power big model are input into the model evaluation module to obtain the evaluation result of the target power big model output by the model evaluation module.

8. The method according to claim 1, characterized in that: The method further comprises: The target power model is updated according to a preset time period.

9. A power data processing device, characterized in that: include: A data acquisition module, used to acquire a power data set for training a target power large model; A model training module, used for training an initial large power model to obtain a target large power model according to the power data set; The data processing module is used to process the power data to be processed through the target power model to obtain a processing result.

10. An electronic device, characterized in that: include: A processor and a memory, the memory being used to store a computer program, and the processor being used to call and run the computer program stored in the memory to execute the method according to any one of claims 1 to 8.

11. A computer-readable storage medium, characterized in that: Used to store a computer program, wherein the computer program causes a computer to execute the method according to any one of claims 1 to 8.

12. A computer program product comprising program instructions, characterized in that When the program instructions are executed on an electronic device, the electronic device is caused to execute the method according to any one of claims 1 to 8.

Images