Task prediction model training method and device, maintenance task distinguishing method and device and storage medium

Abstract

The invention discloses a task prediction model training method and device, a maintenance task distinguishing method and device and a storage medium, and the method comprises the steps: collecting system operation data and maintenance task operation data of a power grid when the maintenance task is executed in the power grid, and executing a simulation maintenance task on the power grid, so as to carry out the expansion of the operation data under various system operation data, if expansion is completed, preprocessing system operation data and operation data, extracting feature data, labeling label data for the feature data according to the maintenance result of the simulation maintenance task, and training a preset deep learning model into a task prediction model according to the feature data under the supervision of the label data; and on the other hand, preprocessing and feature extraction are performed on the operation data of the power grid acquisition system and the operation data of the maintenance task, and the feature data are input into the trained task prediction model to obtain the execution result of the maintenance task, so that the efficiency of maintaining the power grid is improved.

Description

Technical Field

[0001] Embodiments of the present invention relate to the field of big data analysis technology, and in particular to task prediction model training, maintenance task identification methods, devices and storage media. Background Art

[0002] With the long-term operation of the power grid, electrical equipment such as transformers, transmission lines, switchgear, etc. will gradually age and there is a risk of failure.

[0003] At present, electrical equipment in the power grid mainly relies on technicians to formulate maintenance plans based on operating specifications and personal experience, conduct maintenance on electrical equipment to evaluate and maintain it, promptly discover and repair potential faults in electrical equipment, and reduce power outages caused by sudden failures of electrical equipment.

[0004] However, operating specifications and personal experience are highly subjective, and when faced with a complex and changing power grid environment, maintenance errors and omissions are prone to occur, resulting in low maintenance efficiency. Summary of the invention

[0005] The present invention provides a task prediction model training, a maintenance task identification method, a device and a storage medium, so as to improve the efficiency of maintenance of a power grid.

[0006] In a first aspect, an embodiment of the present invention provides a task prediction model training method, comprising:

[0007] When performing a maintenance task in a power grid, collecting system operation data of the power grid and operation data of the maintenance task;

[0008] simulating the maintenance task on the power grid to expand the operation data under various system operation data;

[0009] If the expansion is completed, preprocessing the system operation data and the operation data is performed;

[0010] If the preprocessing is completed, extracting feature data from the system operation data and the operation data;

[0011] Marking the characteristic data with label data according to the maintenance result of the simulation of the maintenance task;

[0012] Under the supervision of the label data, the preset deep learning model is trained as a task prediction model based on the feature data.

[0013] In a second aspect, an embodiment of the present invention further provides a maintenance task identification method, comprising:

[0014] Integrate power safety knowledge and power safety cases into power safety text information;

[0015] A feature extraction network loaded with the feature extraction network training method as described in the first aspect; the feature extraction network includes a transformer represented by a bidirectional encoder and a bidirectional long short-term memory network;

[0016] Inputting the power safety text information into the transformer represented by the bidirectional encoder to extract target semantic features;

[0017] Inputting the target semantic features into the bidirectional long short-term memory network for processing to obtain power safety text features;

[0018] The power safety cases are matched according to the power text features.

[0019] In a third aspect, an embodiment of the present invention further provides a task prediction model training device, comprising:

[0020] A data acquisition module, used for collecting system operation data of the power grid and operation data of the maintenance task when performing maintenance tasks in the power grid;

[0021] A data expansion module, used for simulating the maintenance task on the power grid to expand the operation data under various system operation data;

[0022] A preprocessing module, used for performing preprocessing on the system operation data and the operation data if the expansion is completed;

[0023] A feature extraction module, for extracting feature data from the system operation data and the operation data if the preprocessing is completed;

[0024] A labeling module, used for labeling the characteristic data according to the maintenance result of the simulation of the maintenance task;

[0025] The model training module is used to train a preset deep learning model into a task prediction model based on the feature data under the supervision of the label data.

[0026] In a fourth aspect, an embodiment of the present invention further provides a maintenance task identification device, comprising:

[0027] A power grid data acquisition module, used to collect system operation data of the power grid and operation data of the maintenance task when waiting to perform the maintenance task in the power grid;

[0028] A data preprocessing module, used for performing preprocessing on the system operation data and the operation data;

[0029] A system data feature extraction module, used for extracting feature data from the system operation data and the operation data if the preprocessing is completed;

[0030] A model loading module, used to load the task prediction model trained by the task prediction model training method provided in the first aspect;

[0031] The execution result acquisition module is used to input the characteristic data into the task prediction model to obtain the execution result of the maintenance task.

[0032] In a fifth aspect, an embodiment of the present invention further provides a computer device, including:

[0033] one or more processors;

[0034] A storage device for storing one or more programs;

[0035] When the one or more programs are executed by the one or more processors, the one or more processors implement the task prediction model training method provided in the first aspect of the present invention, or the maintenance task identification method provided in the second aspect of the present invention.

[0036] In a sixth aspect, an embodiment of the present invention further provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements a task prediction model training method as provided in the first aspect of the present invention, or a maintenance task identification method as provided in the second aspect of the present invention.

[0037] In the seventh aspect, an embodiment of the present invention further provides a computer program product, which includes a computer program, and when the computer program is executed by a processor, it implements the task prediction model training method provided in the first aspect of the present invention, or the maintenance task identification method provided in the second aspect of the present invention.

[0038] In an embodiment of the present invention, when a maintenance task is performed in a power grid, system operation data and operation data of the maintenance task are collected from the power grid; a simulated maintenance task is performed on the power grid to expand the operation data under various system operation data; if the expansion is completed, the system operation data and operation data are preprocessed; if the preprocessing is completed, feature data is extracted from the system operation data and operation data; label data is annotated on the feature data according to the maintenance result of the simulated maintenance task; under the supervision of the label data, the preset deep learning model is trained as a task prediction model according to the feature data. Collecting and expanding the operation data helps to more comprehensively understand the performance of the maintenance task of the power grid under different operating conditions, thereby improving the generalization ability of the model. Secondly, the extraction of feature data and the annotation of label data provide a high-quality training foundation for the deep learning model; on the other hand, when waiting for the maintenance task to be performed in the power grid, the system operation data and operation data of the maintenance task are collected from the power grid; preprocessing and feature data extraction are performed on the system operation data and operation data; the trained task prediction model is loaded; the feature data is input into the task prediction model to obtain the execution result of the maintenance task. By preprocessing and extracting features from the operating data of the power grid acquisition system and the operation data of maintenance tasks, a scientific basis can be provided for decision-making. The feature data can be input into the trained task prediction model and the execution results of the maintenance tasks can be obtained, thereby optimizing resource allocation, enhancing safety, reducing costs, and improving the efficiency of power grid maintenance. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] Figure 1 A flowchart of a task prediction model training method provided in Example 1 of the present invention;

[0040] Figure 2 A block diagram of collecting simulated power grid data provided in the first embodiment of the present invention;

[0041] Figure 3 A data block diagram of a simulated power grid maintenance result provided in the first embodiment of the present invention;

[0042] Figure 4 A flowchart of a maintenance task identification method provided in Embodiment 2 of the present invention;

[0043] Figure 5 A structural block diagram of a task prediction model training device provided in Embodiment 3 of the present invention;

[0044] Figure 6 A structural block diagram of a maintenance task identification device provided in Embodiment 4 of the present invention;

[0045] Figure 7 A schematic diagram of the structure of a computer device provided in Embodiment 5 of the present invention. DETAILED DESCRIPTION

[0046] In order to enable those skilled in the art to better understand the scheme of the present invention, the technical scheme in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present invention.

[0047] It should be noted that the terms "first", "second", etc. in the specification and claims of the present invention 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 invention described herein can cover sequential implementations other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0048] Embodiment 1

[0049] See also Figure 1 , shows a flowchart of a task prediction model training method provided by the first embodiment of the present invention, the method can be performed by a task prediction model training device, the task prediction model training device can be implemented in the form of hardware and / or software, and the task prediction model training device can be configured in a computer device. Figure 1 As shown, the method includes:

[0050] Step 101: When performing a maintenance task in a power grid, collect system operation data and operation data of the maintenance task from the power grid.

[0051] The power grid usually performs maintenance tasks in the case of regular maintenance, troubleshooting, equipment upgrades and safety inspections. These maintenance tasks are usually assigned to professional operation and maintenance teams or external contractors, who are responsible for specific maintenance and repair work to ensure the normal operation of power grid equipment and the reliability of the system.

[0052] In this embodiment, when the power grid performs maintenance tasks, the system operation data of the dispatching automation system and the distribution automation system in the power grid and the operation data of the maintenance tasks are collected. The system operation data includes the power system topology, equipment parameters, line parameters, load data, etc., which helps monitor the performance and health status of the power grid, detect abnormal conditions in time, and ensure the safety and stability of the system; and the operation data of the maintenance tasks include the maintenance plan time, power outage equipment, safety measures, power supply plan, reclosing plan, operation tasks and operation equipment, etc., which record the specific measures and results in the maintenance process. The collected system operation data and maintenance task operation data are historical real data, which are maintenance tasks that have been executed, and are mainly used for the training of deep learning models.

[0053] Step 102: Perform a simulated maintenance task on the power grid to expand the operation data under various system operation data.

[0054] Simulation refers to the use of computer technology to simulate the behavior and characteristics of a real system or process by building a model in order to analyze its operating status, predict results, and evaluate performance in different scenarios. The present invention aims to solve the problems of insufficient processing of various system operation data and imperfect methods for generating training data for deep learning models in the prior art. It uses simulation technology to simulate a variety of power grid operation scenarios, expand the operation data of power grid operation scenarios and the operation data under the operation data, so as to train subsequent deep learning models and improve the accuracy of deep learning models.

[0055] In this embodiment, when the power grid performs a simulated maintenance task, by simulating different system operation data and matching the system operation data with the operation data of the maintenance task, the performance of the power grid in various scenarios and the execution results of the maintenance task can be comprehensively evaluated. Potential problems can be identified in advance, the operation data of the maintenance task can be expanded, the risks in the actual maintenance task execution process can be reduced, and the safety and reliability of the system can be improved.

[0056] In one embodiment of the present invention, step 102 may include the following steps:

[0057] Step 1021: formulate simulation rules for the power grid based on the knowledge graph of the maintenance task.

[0058] Knowledge graph is a structured data model that organizes and represents information in the form of nodes and edges, and is used to describe entities and their relationships. In the maintenance tasks of power grids, knowledge graphs can integrate various relevant data, such as equipment characteristics, failure modes, maintenance history, and operating procedures, to form a comprehensive knowledge system.

[0059] In this embodiment, simulation rules are formulated for the power grid based on the knowledge graph of scheduling and maintenance tasks and actual business needs. Existing maintenance knowledge, experience, and data relationships are used to build a set of systematic rules to guide the simulation process and parameter settings, improve the accuracy and effectiveness of the simulation model, and enable it to truly reflect the performance of the power grid in different maintenance task scenarios. The simulation rules include conditions, data sources, and operating steps.

[0060] Step 1022: Call a preset rule engine to perform maintenance tasks on the power grid simulation according to the simulation rules, and obtain operation data under various system operation data.

[0061] A rule engine (such as Drools) is a software system that executes predefined rules and logic to handle complex data and decision-making processes.

[0062] In this embodiment, when the power grid simulation performs maintenance tasks, the rule engine generates various system operation data according to the simulation rules, and generates paired operation data for the various simulated system operation data by performing operation tasks on the power grid simulation. By applying these simulation rules, the rule engine can quickly identify problems and generate corresponding operation suggestions to ensure the safety and reliability of the power grid in different scenarios.

[0063] Specifically, Figure 2 As shown, the grid operation is simulated in the preset rule engine, and real-time data streams are integrated to ensure that corresponding operations are quickly executed when conditions such as load or temperature reach thresholds. System operation data is collected from devices on the line with the same timestamp. Figure 2 In the example, each row represents the system operation data of each device at this timestamp, including the device's current, power, temperature, fault type, and fault description; each simulation rule is loaded, and when the system operation data meets the conditions in the preset rule engine, the target data is read from the system operation data according to the data source, and the operation is performed according to the target data in the operation steps to obtain the operation data under various system operation data. Whenever the system operation data is updated, the rule engine is triggered to execute the simulation rules and automatically generate the operation data under the system operation data.

[0064] Exemplarily, based on historical load data and equipment status, simulation software (MATLAB / Simulink) is used to simulate operating data under high load and high temperature conditions. Taking the high load period of the power grid (such as the high temperature period in summer) as an example, the power grid equipment faces great pressure and requires effective load management measures to ensure the safety of the equipment and the stability of power supply. The following are predefined simulation rules, which are not limited by the present invention.

[0065] Rule 1 performs load splitting operations:

[0066] Condition: If the load on a line exceeds 80%, load shedding is required.

[0067] Data source: Real-time load monitoring data, recording the current and power of each line.

[0068] Operation steps: Monitor line load data. When the load exceeds 80%, the load diversion rule is triggered to transfer part of the load to the backup line to ensure that the load on the main line is reduced to a safe level.

[0069] Rule 2 performs equipment cooling operations:

[0070] Condition: If the device temperature exceeds 90°C, you need to perform device cooling operation.

[0071] Data source: device temperature sensor data.

[0072] Operation steps: Monitor the device temperature in real time. When the temperature exceeds 90°C, trigger the cooling operation. Start the cooling system to reduce the device temperature and continue monitoring until it returns to normal.

[0073] Step 1023: Perform a data enhancement operation on the operation data to expand the operation data.

[0074] In this embodiment, data enhancement operations are performed on the operation data obtained by simulation. Data enhancement techniques include adding noise, random scaling and cropping, rotation, flipping, color transformation (such as adjusting brightness and contrast), time offset, synthetic data (using a generative adversarial network), and simulation parameter perturbation. Using data enhancement techniques can increase the diversity of operation data and achieve the purpose of expanding the operation data.

[0075] Step 103: If the expansion is completed, pre-processing is performed on the system operation data and the operation data.

[0076] In this embodiment, after the operation data is expanded, corresponding pre-processing operations may be performed on the system operation data and the operation data respectively to improve the quality of the system operation data and the operation data.

[0077] Exemplarily, the preprocessing of system operation data may include the following steps:

[0078] Step 1031: Use linear interpolation to fill in missing values ​​in the system operation data.

[0079] In this embodiment, linear interpolation is a technique for filling missing values, which is applicable to time series data or continuous data. When processing system operation data, linear interpolation estimates the unrecorded system operation data by using the linear relationship between the system operation data recorded at a certain time, and the formula is as follows:

[0080]

[0081] In the formula, x i Indicates unrecorded system operation data, x i-1 and x i+1 Indicates the adjacent system operation data recorded at a certain time.

[0082] Step 1032: If the missing values ​​are filled, use the wavelet transform algorithm to remove noise from the system operation data.

[0083] In this embodiment, wavelet transform is an effective signal processing technology, which is suitable for analyzing signals with non-stationary characteristics, especially time series data. Because wavelet transform can decompose signals into different frequency components and wavelet transform is suitable for processing system operation data containing mutations or abnormal points, it is often used to remove short-term burst noise. After completing the missing value filling, the wavelet transform algorithm can be used to decompose the system operation data into components of different frequencies, thereby distinguishing useful signals from noise, which can effectively suppress noise while retaining the main characteristics of the signal. The formula used is as follows:

[0084]

[0085] In the formula, x ’ represents the signal of the system operation data after removing the noise, x represents the original signal of the system operation data, represents the wavelet basis function, and k represents the parameter of the wavelet basis function translation.

[0086] Step 1033: If the noise is removed, Gaussian filtering is performed on the system operation data.

[0087] In this embodiment, if the system operation data is de-noised, the system operation data is subjected to Gaussian filtering. Gaussian filtering is usually used to smooth signals and remove high-frequency noise, and is suitable for processing continuous and stable data. Gaussian filters are particularly useful for system operation data of power grid equipment (such as temperature, pressure and other sensor data), because Gaussian filters can smooth the measurement system operation data and reduce high-frequency noise caused by sensor jitter or small fluctuations. The formula used is as follows:

[0088] y'=y*g;

[0089] In the formula, y' represents the signal after the system operation data is filtered, y represents the original signal of the system operation data, * represents the convolution operation, and g represents the Gaussian kernel function.

[0090] Exemplarily, the preprocessing of the operation data may include the following steps:

[0091] Step 1034: Use linear interpolation to fill missing values ​​in the operating data.

[0092] In this embodiment, when processing the operation data, the linear interpolation method estimates the unrecorded operation data by utilizing the linear relationship between the operation data recorded at a certain time, and the formula is as follows:

[0093]

[0094] In the formula, x i Indicates unrecorded operation data, x i-1 and x i+1 Indicates adjacent operation data recorded at a certain time.

[0095] Step 1035: If the missing values ​​are filled, use the wavelet transform algorithm to remove noise from the operating data.

[0096] In this embodiment, after completing the missing value filling, the wavelet transform algorithm can be used to decompose the operation data into components of different frequencies, thereby distinguishing useful signals from noise, effectively suppressing noise, and retaining the main features of the signal. The formula used is as follows:

[0097]

[0098] In the formula, x ’ represents the signal of the operation data to remove noise, x represents the original signal of the operation data, represents the wavelet basis function, and k represents the parameter of the wavelet basis function translation.

[0099] Step 1036: If the noise is removed, Gaussian filtering is performed on the operation data.

[0100] In this embodiment, if the noise of the operation data is removed, the operation data is subjected to Gaussian filtering, which can smooth the measurement operation data and reduce high-frequency noise caused by sensor jitter or small fluctuations. The formula used is as follows:

[0101] y'=y*g;

[0102] Where y' represents the signal after filtering of the operation data, y represents the original signal of the operation data, * represents the convolution operation, and g represents the Gaussian kernel function.

[0103] Step 104: If the preprocessing is completed, feature data is extracted from the system operation data and the operation data.

[0104] In this embodiment, after completing data preprocessing, feature data can be extracted from the system operation data and operation data to improve the effect of deep learning model training. Through feature extraction, the dimensions of the system operation data and operation data can be reduced, the training effect of the deep learning model can be improved, and the interpretability of the system operation data and operation data can be enhanced, and the impact of noise can be reduced.

[0105] For example, using the minimum-maximum scaling method to perform normalization operations on the system operation data and the operation data respectively can effectively eliminate the dimensional influence between different features, making the deep learning model training more stable and efficient. In addition, by handling special cases, such as when the maximum value is equal to the minimum value, the correctness of the system operation data and the operation data processing can be ensured.

[0106] Use the minimum and maximum scaling method to normalize the system operation data. The formula is as follows:

[0107]

[0108] Wherein, x represents the system operation data, x' represents the normalized system operation data, max(x) represents the maximum value of the system operation data, and min(x) represents the minimum value of the system operation data.

[0109] The operation data is normalized using the minimum and maximum scaling method, and the formula is as follows:

[0110]

[0111] In the formula, x represents the operation data, x' represents the normalized operation data, max(x) represents the maximum value of the operation data, and min(x) represents the minimum value of the operation data.

[0112] In this embodiment, if the system operation data and the operation data have completed the normalization operation, the least absolute shrinkage and selection algorithm (LASSO) is used to extract feature data from the system operation data and the operation data respectively, and finally the extracted feature data are spliced.

[0113] The least absolute shrinkage and selection algorithm (LASSO) is used to extract features from the system operation data. The formula is as follows:

[0114]

[0115] In the formula, n represents the number of system operation data, y i represents the actual output value of the ith system operation data, x i represents the feature vector of the i-th system operation data, w represents the weight vector, and λ represents the regularization parameter.

[0116] The least absolute shrinkage and selection algorithm (LASSO) is used to extract features from the operation data. The formula is as follows:

[0117]

[0118] In the formula, n represents the number of operation data, y i represents the actual output value of the i-th operation data, x i represents the feature vector of the i-th operation data, w represents the weight vector, and λ represents the regularization parameter.

[0119] The least absolute shrinkage and selection algorithm (LASSO) can perform feature selection and parameter estimation simultaneously. By introducing λ regularization, LASSO can effectively reduce the complexity of deep learning models and prevent overfitting. At the same time, the normalization operation helps to improve the performance of LASSO because it ensures that each feature is on the same scale, which helps the algorithm to more accurately identify important features, thereby improving the interpretability and predictive ability of deep learning models. Therefore, when processing system operation data and operation data, valuable feature data can be more effectively extracted by applying LASSO after normalization.

[0120] Step 105: annotate the feature data with label data according to the maintenance result of the simulated maintenance task.

[0121] Label data refers to the output information corresponding to feature data in machine learning and data analysis, which is used to represent the category or attribute of the sample. In supervised learning, label data provides target values ​​for deep learning models, enabling them to learn the mapping relationship from input features to output results, providing clear goals for model training and optimizing model training.

[0122] In this embodiment, if Figure 3 As shown in the figure, each row represents the system operation data collected when simulating a device on a certain line at a certain timestamp, recording the current, power, temperature, fault type, fault description and maintenance results of each device. The maintenance results simulated at each timestamp are used as label data for deep learning model training.

[0123] In this embodiment, the feature data is labeled with label data according to the maintenance results of the simulated maintenance task, which can provide clear target information for subsequent analysis and model training. This labeling enables the machine learning algorithm to effectively learn the relationship between the input feature data, model output results and label data, thereby improving the training effect of the model.

[0124] Step 106: Under the supervision of the label data, the preset deep learning model is trained as a task prediction model based on the feature data.

[0125] In this embodiment, by utilizing the annotated label data, the model can learn the complex relationship between the input feature data, the model output results and the label data, thereby training the preset deep learning model into a task prediction model.

[0126] In one embodiment of the present invention, step 106 may include the following steps:

[0127] Step 1061: Divide the feature data into a training set and a test set.

[0128] In this embodiment, the feature data is divided into a training set and a test set in a ratio of 7 to 3. The training set is used to train the deep learning model to help it learn the relationship between the training set and the label data; and the test set is used to evaluate the performance of the deep learning model on unseen data, thereby detecting its predictive ability and robustness.

[0129] Step 1062: Iteratively train the preset deep learning model according to the training set until the deep learning model meets the preset training conditions.

[0130] Iterative training is a method of optimizing deep learning model parameters by repeating the training process multiple times in machine learning and deep learning. Each iteration includes steps such as forward propagation to calculate prediction results, loss calculation, and back propagation to update parameters. The process is repeated until the performance of the deep learning model reaches the preset training conditions.

[0131] In this embodiment, the training set is input into a preset deep learning model for iterative training. Through multiple iterations, the model can gradually learn the complex patterns and features in the training set and label data. When the deep learning model meets the preset training conditions, the training stops.

[0132] The deep learning model may be a long short-term memory network (LSTM), a transformer, etc., and the embodiment of the present invention does not limit this.

[0133] Long short-term memory network (LSTM) is a special recurrent neural network (RNN) that solves the gradient vanishing and exploding problems faced by traditional RNN when processing long sequence data. LSTM effectively memorizes long-term dependent information by introducing a gating mechanism to control the flow of information.

[0134] Transformer is a deep learning architecture based on the self-attention mechanism. It can effectively capture the relationship between elements in the sequence by processing the input sequence in parallel. Transformer mainly consists of an encoder and a decoder. The encoder is responsible for extracting input features, and the decoder generates output.

[0135] In one embodiment of the present invention, step 1062 may further include the following steps:

[0136] Step 110: Input the training set into the deep learning model to predict the execution results of future maintenance tasks.

[0137] In this embodiment, the training set is input into an untrained deep learning model to predict the execution results of future maintenance tasks. This execution result is repeatedly compared with the label data to optimize the parameters of the deep learning model and assist in the training of the deep learning model.

[0138] Step 120: Generate a loss value based on the predicted execution result and the label data.

[0139] In this embodiment, a loss value is generated based on the predicted execution result and the label data to evaluate the gap between the execution result predicted by the deep learning model and the actual label data. Through the loss value, the deep learning model can identify the accuracy of the current prediction, so as to adjust the parameters during the training process to optimize the performance. The smaller the loss value, the more accurate the prediction of the deep learning model, so the loss value is an important indicator to guide the learning and improvement of the deep learning model.

[0140] Step 130: Update the deep learning model according to the loss value.

[0141] In this embodiment, by calculating the loss value, the deep learning model can identify which parameters need to be adjusted to reduce the gap between the prediction and the true value, and update the deep learning model based on the loss value using a back-propagation algorithm such as stochastic gradient descent (SGD) or adaptive moment estimation (Adam).

[0142] The SGD algorithm is an optimization algorithm used to update the parameters of the model during the training process. SGD uses only a small portion of the training data (one or a few samples) to calculate the gradient each time, thereby speeding up the training and improving the adaptability to large data sets, helping to jump out of the local optimal solution and find a better global solution.

[0143] Adaptive Moment Estimation (Adam) is an optimization algorithm that improves the efficiency of gradient descent by combining momentum method and adaptive learning rate. It has strong adaptability and is particularly suitable for processing large-scale data and high-dimensional space.

[0144] Step 140, determine whether the deep learning model meets the preset training conditions; if so, execute step 150; if not, return to execute step 110.

[0145] In this embodiment, after completing a parameter update, the system will check whether the deep learning model has reached the set training criteria. These training criteria may include whether the reduction in loss value is within an acceptable range, whether the performance of the deep learning model has been significantly improved, or whether the number of training times has reached a predetermined upper limit. If it is detected that these criteria are met, the system will confirm that the training process has been completed.

[0146] On the contrary, if these preset training conditions are not met, it means that the deep learning model still has not achieved the ideal learning effect. At this time, the system will return to execute and input the training set into the deep learning model to predict the execution results of future maintenance tasks. This cyclic design ensures that the deep learning model is continuously optimized and iterated, so that the results output by the deep learning model are close to the label data, thereby improving the effectiveness and reliability of the deep learning model in practical applications.

[0147] Step 150: Determine whether the deep learning model has completed training.

[0148] In this embodiment, at this time, the deep learning model meets the preset training conditions in multiple iterations of training, and it is determined that the deep learning model has completed training, which can be used for subsequent applications in predicting the execution results of maintenance tasks.

[0149] Step 1063: Use the test set to test the deep learning model to generate multiple performance indicators.

[0150] In this embodiment, the trained deep learning model is tested by the divided test set to generate multiple performance indicators. The performance indicators refer to indicators such as accuracy, precision, and recall rate of the calculated deep learning model.

[0151] Step 1064: If multiple performance indicators meet the preset performance standards, the deep learning model is deployed as a task prediction model.

[0152] In this embodiment, if the multiple performance indicators generated can meet the preset performance standards, it means that the training of the deep learning model has reached the preset standards. The deep learning model is deployed as a task prediction model to predict the execution results of the power grid maintenance tasks, so as to adjust the maintenance tasks and improve the power grid maintenance efficiency.

[0153] In an embodiment of the present invention, when a maintenance task is performed in a power grid, system operation data and operation data of the maintenance task are collected from the power grid; a simulated maintenance task is performed on the power grid to expand the operation data under various system operation data; if the expansion is completed, preprocessing is performed on the system operation data and the operation data; if the preprocessing is completed, feature data is extracted from the system operation data and the operation data; label data is annotated on the feature data according to the maintenance results of the simulated maintenance task; under the supervision of the label data, a preset deep learning model is trained as a task prediction model according to the feature data. Collecting and expanding the operation data helps to more comprehensively understand the performance of the maintenance task of the power grid under different operating conditions, thereby improving the generalization ability of the model. Secondly, the extraction of feature data and the annotation of label data provide a high-quality training foundation for the deep learning model, so that the model can more accurately predict the execution results of future maintenance tasks and improve the efficiency of power grid maintenance.

[0154] Embodiment 2

[0155] See also Figure 4 , shows a flowchart of a maintenance task identification method provided by the second embodiment of the present invention, the method can be executed by a maintenance task identification device, the maintenance task identification device can be implemented in the form of hardware and / or software, and the maintenance task identification device can be configured in a computer device. Figure 4 As shown, the method includes:

[0156] Step 401: While waiting for a maintenance task to be executed in a power grid, collect system operation data and operation data of the maintenance task from the power grid.

[0157] In this embodiment, while waiting for the maintenance task to be executed in the power grid, the system operation data and the operation data of the maintenance task are collected in real time. The system operation data and operation data collected in real time are not yet maintained, and the results of the maintenance are not yet known.

[0158] Step 402: Preprocess the system operation data and operation data.

[0159] In one embodiment of the present invention, linear interpolation is used to fill missing values ​​in system operation data and operation data; if the missing values ​​are filled, a wavelet transform algorithm is used to remove noise from the system operation data and operation data; if the noise is removed, Gaussian filtering is performed on the system operation data and operation data.

[0160] Step 403: If the preprocessing is completed, feature data is extracted from the system operation data and the operation data.

[0161] In one embodiment of the present invention, a minimum maximum scaling method is used to perform a normalization operation on the system operation data and the operation data; if the normalization operation is completed, a minimum absolute shrinkage and selection algorithm is used to extract feature data from the system operation data and the operation data.

[0162] Step 404: Load the task prediction model.

[0163] In this embodiment, the task prediction model trained in the first embodiment is loaded, the collected system operation data and operation data are analyzed, and the execution result of the maintenance task is predicted.

[0164] Step 405: Input the characteristic data into the task prediction model to obtain the execution result of the maintenance task.

[0165] In this embodiment, feature data is extracted from the collected system operation data and operation data, and then the extracted feature data is input into the trained task prediction model. The task prediction model outputs the execution result of this maintenance task, provides a basis for decision-making, optimizes resource allocation, and improves the efficiency of power grid maintenance.

[0166] In an embodiment of the present invention, when waiting for the execution of a maintenance task in the power grid, the operation data of the power grid collection system and the operation data of the maintenance task are collected; the system operation data and the operation data are preprocessed; if the preprocessing is completed, feature data is extracted from the system operation data and the operation data; the task prediction model is loaded; the feature data is input into the task prediction model to obtain the execution result of the maintenance task. By preprocessing and extracting features from the operation data of the power grid collection system and the operation data of the maintenance task, a scientific basis can be provided for decision-making, and the feature data is input into the trained task prediction model to obtain the execution result of the maintenance task, thereby optimizing resource allocation, enhancing safety and reducing costs, and also improving the efficiency of power grid maintenance.

[0167] Embodiment 3

[0168] Figure 5 A schematic diagram of the structure of a task prediction model training device provided in Embodiment 3 of the present invention is shown in FIG. Figure 5 As shown, the device comprises:

[0169] The data acquisition module 501 is used to collect system operation data of the power grid and operation data of the maintenance task when performing the maintenance task in the power grid;

[0170] A data expansion module 502, configured to perform a simulation of the maintenance task on the power grid to expand the operation data under various system operation data;

[0171] A preprocessing module 503, configured to perform preprocessing on the system operation data and the operation data if the expansion is completed;

[0172] A feature extraction module 504 is used to extract feature data from the system operation data and the operation data if the preprocessing is completed;

[0173] A labeling module 505 is used to label the characteristic data according to the maintenance result of the simulation of the maintenance task;

[0174] The model training module 506 is used to train the preset deep learning model into a task prediction model based on the feature data under the supervision of the label data.

[0175] In one embodiment of the present invention, the data expansion module 502 includes:

[0176] A rule formulation module, used to formulate simulation rules for the power grid according to the knowledge graph of the maintenance task; wherein the simulation rules include conditions, data sources and operation steps;

[0177] A simulation execution module, used to call a preset rule engine to simulate and perform maintenance tasks on the power grid according to the simulation rules, and obtain operation data under various system operation data;

[0178] The data enhancement module is used to perform a data enhancement operation on the operation data to expand the operation data.

[0179] In one embodiment of the present invention, the simulation execution module includes:

[0180] A power grid simulation module, used to simulate the operation of the power grid in a preset rule engine, obtain system operation data, and load each of the simulation rules;

[0181] The operation data acquisition module is used to read the target data from the system operation data according to the data source when it is detected in the preset rule engine that the system operation data meets the conditions, and to perform the operation according to the target data in the operation steps to obtain the operation data under various system operation data.

[0182] In one embodiment of the present invention, the preprocessing module 503 includes:

[0183] A missing value filling module, used for filling missing values ​​of the system operation data and the operation data using a linear interpolation method;

[0184] A noise removal module, for removing noise from the system operation data and the operation data using a wavelet transform algorithm if the missing value filling is completed;

[0185] The data filtering module is used to perform Gaussian filtering on the system operation data and the operation data if noise removal is completed.

[0186] In one embodiment of the present invention, the feature extraction module 504 includes:

[0187] A data normalization operation module, used for performing a normalization operation on the system operation data and the operation data using a minimum-maximum scaling method;

[0188] The feature data extraction module is used to extract feature data from the system operation data and the operation data using a minimum absolute shrinkage and selection algorithm if the normalization operation is completed.

[0189] In one embodiment of the present invention, the model training module 506 includes:

[0190] A data partitioning module, used for partitioning the feature data into a training set and a test set;

[0191] A deep learning model training module, used to iteratively train a preset deep learning model according to the training set until the deep learning model meets preset training conditions;

[0192] A model testing module, used to test the deep learning model using the test set to generate multiple performance indicators;

[0193] The performance standard compliance module is used to deploy the deep learning model as a task prediction model if multiple performance indicators meet the preset performance standards.

[0194] In one embodiment of the present invention, the deep learning model training module includes:

[0195] A data input module, used to input the training set into the deep learning model to predict the execution results of future maintenance tasks;

[0196] A loss value generating module, used to generate a loss value according to the predicted execution result and the label data;

[0197] A model updating module, used to update the deep learning model according to the loss value;

[0198] A training condition judgment module, used to judge whether the deep learning model meets the preset training conditions; if so, execute the method of the model training completion module; if not, return to execute the method of the data input module;

[0199] The model training completion module is used to determine whether the deep learning model has completed training.

[0200] The task prediction model training device provided in the embodiment of the present invention can execute the task prediction model training method provided in any embodiment of the present invention, and has the corresponding functional modules and beneficial effects of executing the task prediction model training method.

[0201] Embodiment 4

[0202] Figure 6 A schematic diagram of the structure of a maintenance task identification device provided in the fourth embodiment of the present invention is shown in FIG. Figure 6 As shown, the device comprises:

[0203] The power grid data collection module 601 is used to collect system operation data and operation data of the maintenance task for the power grid when waiting for the maintenance task to be performed in the power grid;

[0204] A data preprocessing module 602 is used to perform preprocessing on the system operation data and the operation data;

[0205] A system data extraction feature module 603 is used to extract feature data from the system operation data and the operation data if the preprocessing is completed;

[0206] A model loading module 604 is used to load the task prediction model trained by the task prediction model training method provided in the first embodiment of the present invention;

[0207] The execution result acquisition module 605 is used to input the characteristic data into the task prediction model to obtain the execution result of the maintenance task.

[0208] In one embodiment of the present invention, the data preprocessing module 602 includes:

[0209] A linear interpolation module, used for filling missing values ​​of the system operation data and the operation data using a linear interpolation method;

[0210] A wavelet transform module, for removing noise from the system operation data and the operation data using a wavelet transform algorithm if the missing values ​​are filled;

[0211] The Gaussian filtering module is used to perform Gaussian filtering on the system operation data and the operation data if noise removal is completed.

[0212] In one embodiment of the present invention, the system data extraction feature module 603 includes:

[0213] A normalization module, used for performing a normalization operation on the system operation data and the operation data using a minimum-maximum scaling method;

[0214] The extraction module is used to extract feature data from the system operation data and the operation data using a minimum absolute shrinkage and selection algorithm if the normalization operation is completed.

[0215] The maintenance task identification device provided in the embodiment of the present invention can execute the maintenance task identification method provided in any embodiment of the present invention, and has the corresponding functional modules and beneficial effects of executing the maintenance task identification method.

[0216] Embodiment 5

[0217] See also Figure 7 , showing a schematic diagram of the structure of a computer device provided by an embodiment of the present invention. The computer device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workbenches, personal digital assistants, blade servers, mainframe computers, and other suitable computers. The computer device may also represent various forms of mobile devices, such as personal digital processing, cellular phones, smart phones, wearable devices (such as helmets, glasses, watches, etc.), and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the present invention described and / or required herein.

[0218] like Figure 7 As shown, the computer device 10 includes at least one processor 11, and a memory connected to the at least one processor 11, such as a read-only memory (ROM) 12, a random access memory (RAM) 13, etc., wherein the memory stores a computer program that can be executed by at least one processor, and the processor 11 can perform various appropriate actions and processes according to the computer program stored in the read-only memory (ROM) 12 or the computer program loaded from the storage unit 18 to the random access memory (RAM) 13. In the RAM 13, various programs and data required for the operation of the computer device 10 can also be stored. The processor 11, ROM 12 and RAM 13 are connected to each other through a bus 14. An input / output (I / O) interface 15 is also connected to the bus 14.

[0219] A number of components in the computer device 10 are connected to the I / O interface 15, including: an input unit 16, such as a keyboard, a mouse, etc.; an output unit 17, such as various types of displays, speakers, etc.; a storage unit 18, such as a disk, an optical disk, etc.; and a communication unit 19, such as a network card, a modem, a wireless communication transceiver, etc. The communication unit 19 allows the computer device 10 to exchange information / data with other devices through a computer network such as the Internet and / or various telecommunication networks.

[0220] The processor 11 may be a variety of general and / or special processing components with processing and computing capabilities. Some examples of the processor 11 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various dedicated artificial intelligence (AI) computing chips, various processors running machine learning model algorithms, digital signal processors (DSPs), and any appropriate processors, controllers, microcontrollers, etc. The processor 11 executes the various methods and processes described above, such as the task prediction model training method, or the maintenance task discrimination method.

[0221] In some embodiments, the task prediction model training method, or the maintenance task identification method, may be implemented as a computer program, which is tangibly contained in a computer-readable storage medium, such as a storage unit 18. In some embodiments, part or all of the computer program may be loaded and / or installed on the computer device 10 via the ROM 12 and / or the communication unit 19. When the computer program is loaded into the RAM 13 and executed by the processor 11, one or more steps of the task prediction model training method, or the maintenance task identification method described above may be executed. Alternatively, in other embodiments, the processor 11 may be configured to execute the task prediction model training method, or the maintenance task identification method, in any other appropriate manner (for example, by means of firmware).

[0222] Various implementations of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), systems on chips (SOCs), load programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various implementations can include: being implemented in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which can be a special purpose or general purpose programmable processor that can receive data and instructions from a storage system, at least one input device, and at least one output device, and transmit data and instructions to the storage system, the at least one input device, and the at least one output device.

[0223] Computer programs for implementing the methods of the present invention may be written in any combination of one or more programming languages. These computer programs may be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing device, so that when the computer program is executed by the processor, the functions / operations specified in the flow chart and / or block diagram are implemented. The computer program may be executed entirely on the machine, partially on the machine, partially on the machine and partially on a remote machine as a stand-alone software package, or entirely on a remote machine or server.

[0224] In the context of the present invention, a computer-readable storage medium may be a tangible medium that may contain or store a computer program for use by or in combination with an instruction execution system, device or equipment. A computer-readable storage medium may include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, devices or equipment, or any suitable combination of the foregoing. Alternatively, a computer-readable storage medium may be a machine-readable signal medium. A more specific example of a machine-readable storage medium may include an electrical connection based on one or more lines, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0225] To provide interaction with a user, the systems and techniques described herein can be implemented on a computer device having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and a pointing device (e.g., a mouse or trackball) through which the user can provide input to the computer device. Other types of devices can also be used to provide interaction with the user; for example, the feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including acoustic input, voice input, or tactile input).

[0226] The systems and techniques described herein may be implemented in a computing system that includes backend components (e.g., as a data server), or a computing system that includes middleware components (e.g., an application server), or a computing system that includes frontend components (e.g., a user computer with a graphical user interface or a web browser through which a user can interact with implementations of the systems and techniques described herein), or a computing system that includes any combination of such backend components, middleware components, or frontend components. The components of the system may be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include: a local area network (LAN), a wide area network (WAN), a blockchain network, and the Internet.

[0227] A computing system may include a client and a server. The client and the server are generally remote from each other and usually interact through a communication network. The client and server relationship is generated by computer programs running on the corresponding computers and having a client-server relationship with each other. The server may be a cloud server, also known as a cloud computing server or cloud host, which is a host product in the cloud computing service system to solve the defects of difficult management and weak business scalability in traditional physical hosts and VPS services.

[0228] Embodiment 6

[0229] An embodiment of the present invention further provides a computer program product, which includes a computer program. When the computer program is executed by a processor, it implements the task prediction model training method or the maintenance task identification method provided in any embodiment of the present invention.

[0230] In the process of implementation, the computer program product can be written in one or more programming languages ​​or a combination thereof to perform the computer program code of the present invention, including object-oriented programming languages, such as Java, Smalltalk, C++, and conventional procedural programming languages, such as "C" language or similar programming languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as a separate software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In the case of a remote computer, the remote computer can be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or can be connected to an external computer (for example, using an Internet service provider to connect through the Internet).

[0231] It should be understood that the various forms of processes shown above can be used to reorder, add or delete steps. For example, the steps described in the present invention can be executed in parallel, sequentially or in different orders, as long as the desired results of the technical solution of the present invention can be achieved, and this document does not limit this.

[0232] The above specific implementations do not constitute a limitation on the protection scope of the present invention. It should be understood by those skilled in the art that various modifications, combinations, sub-combinations and substitutions can be made according to design requirements and other factors. Any modification, equivalent substitution and improvement made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A task prediction model training method, characterized in that: include: When performing a maintenance task in a power grid, collecting system operation data of the power grid and operation data of the maintenance task; simulating the maintenance task on the power grid to expand the operation data under various system operation data; If the expansion is completed, preprocessing the system operation data and the operation data is performed; If the preprocessing is completed, extracting feature data from the system operation data and the operation data; Marking the characteristic data with label data according to the maintenance result of the simulation of the maintenance task; Under the supervision of the label data, the preset deep learning model is trained as a task prediction model based on the feature data.

2. The method according to claim 1, characterized in that The performing of simulation on the power grid to perform the maintenance task to expand the operation data under various system operation data includes: Formulate simulation rules for the power grid according to the knowledge graph of the maintenance task; wherein the simulation rules include conditions, data sources and operation steps; Calling a preset rule engine to perform maintenance tasks on the power grid simulation according to the simulation rules, and obtaining operation data under various system operation data; A data enhancement operation is performed on the operation data to expand the operation data.

3. The method according to claim 2, characterized in that The calling of the preset rule engine performs maintenance tasks on the power grid simulation according to the simulation rules to obtain operation data under various system operation data, including: Simulating the operation of the power grid in a preset rule engine to obtain system operation data, and loading each of the simulation rules; When it is detected in the preset rule engine that the system operation data meets the conditions, the target data is read from the system operation data according to the data source, and the operation is performed according to the target data substituted into the operation steps to obtain the operation data under various system operation data.

4. The method according to claim 1, characterized in that: The preprocessing of the system operation data and the operation data comprises: Filling missing values ​​of the system operation data and the operation data using a linear interpolation method; If the missing values ​​are filled, a wavelet transform algorithm is used to remove noise from the system operation data and the operation data; If the noise is removed, Gaussian filtering is performed on the system operation data and the operation data.

5. The method according to claim 1, characterized in that The extracting characteristic data from the system operation data and the operation data comprises: Performing a normalization operation on the system operation data and the operation data using a minimum-maximum scaling method; If the normalization operation is completed, feature data is extracted from the system operation data and the operation data using a least absolute shrinkage and selection algorithm.

6. The method according to any one of claims 1 to 5, characterized in that: The step of training a preset deep learning model into a task prediction model according to the feature data under the supervision of the label data includes: Dividing the feature data into a training set and a test set; Iteratively training a preset deep learning model according to the training set until the deep learning model meets preset training conditions; Testing the deep learning model using the test set to generate multiple performance indicators; If the multiple performance indicators meet the preset performance standards, the deep learning model is deployed as a task prediction model.

7. The method according to claim 6, characterized in that The iterative training of a preset deep learning model according to the training set until the deep learning model meets a preset training condition includes: Inputting the training set into the deep learning model to predict the execution results of future maintenance tasks; Generate a loss value based on the predicted execution result and the label data; Updating the deep learning model according to the loss value; Determine whether the deep learning model meets the preset training conditions; if so, determine that the deep learning model has completed training; if not, return to execute the input of the training set into the deep learning model to obtain the execution result of the future maintenance task.

8. A maintenance task identification method, characterized in that: include: While waiting for a maintenance task to be performed in a power grid, collecting system operation data of the power grid and operation data of the maintenance task; performing preprocessing on the system operation data and the operation data; If the preprocessing is completed, extracting feature data from the system operation data and the operation data; Loading a task prediction model trained according to the method described in any one of claims 1 to 7; The characteristic data is input into the task prediction model to obtain the execution result of the maintenance task.

9. A computer device, characterized in that: include: one or more processors; A storage device for storing one or more programs; When the one or more programs are executed by the one or more processors, the one or more processors implement the task prediction model training method as described in any one of claims 1 to 7, or the maintenance task identification method as described in claim 8.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, it implements the task prediction model training method as described in any one of claims 1 to 7, or the maintenance task identification method as described in claim 8.

Images