Power system model reasoning analysis method and device based on containerization

Abstract

The invention provides a containerization-based power system model reasoning analysis method and device, and relates to the technical field of power grids. According to the method, the containers for bearing the power system models are established, so that the power system models run in the same system environment, and the compatibility of the power system models is realized by adopting a containerization technology. Standard features obtained through real-time operation data are input into all containers, all power system models are operated, and output results of a transformer diagnosis model, a circuit breaker fault diagnosis model and a load prediction model are obtained. Then, a health assessment model is adopted to perform operation state analysis on each output result, a health score representing the fault probability of each device is determined, a reasoning analysis report is generated, the overall condition of the power system is visually displayed to operation and maintenance personnel, the analysis difficulty of the operation and maintenance personnel is reduced, and the operation and maintenance efficiency of the power system is improved. And the intelligent operation and maintenance level of the power system is improved.

Description

Technical Field

[0001] The present invention relates to the field of electric power grid technology, and in particular to a container-based power system model reasoning and analysis method and device. Background Art

[0002] In the field of intelligent operation and maintenance of power systems, the application of artificial intelligence models has become an important means to improve the state monitoring, fault diagnosis and predictive maintenance capabilities of power equipment. However, the current model deployment method still faces many challenges, especially in cross-platform compatibility and multi-model management.

[0003] Traditional model deployment usually relies on specific hardware platforms or cloud servers, which greatly limits the flexibility of model application in different power scenarios. Especially on resource-constrained edge devices, due to hardware and software limitations, traditional deployment methods make it difficult to achieve real-time and stable reasoning of models, thus affecting the intelligent operation and maintenance level of power systems.

[0004] The complexity and diversity of the power system also exacerbates the difficulty of model deployment. Different types of power equipment and application scenarios require different models for monitoring and diagnosis, but existing power system models often lack compatibility. Each model can only monitor part of the operating data and cannot achieve overall real-time monitoring of the power system. This not only affects the operation and maintenance personnel's comprehensive control of the power system status, but also increases the difficulty and uncertainty of fault diagnosis.

[0005] The existing model management methods have significant deficiencies in multi-platform adaptation and efficient management. Each power system model usually relies on its own system environment, which makes it difficult to achieve unified management and coordination when multiple models are deployed in parallel. This not only increases the workload of operation and maintenance personnel, but also diversifies the output results and increases the difficulty of analysis, thereby reducing the efficiency of operation and inspection.

[0006] In summary, the traditional deployment and management methods of power system models have the problem of low operation and maintenance efficiency. Summary of the invention

[0007] The present invention provides a container-based power system model reasoning and analysis method and device, which can improve the operation and inspection efficiency of the power system and enhance the intelligent operation and maintenance level of the power system.

[0008] In a first aspect, the present invention provides a container-based power system model reasoning and analysis method, the method comprising: acquiring real-time operating data in the power system, the real-time operating data comprising image data and parameter data; based on a meta-information module, performing feature extraction on the real-time operating data to obtain standard features; inputting the standard features into the containers of each power system model to obtain output results of each container; the power system model comprises a transformer diagnosis model, a circuit breaker fault diagnosis model and a load prediction model; based on the output results of each container and a preset health assessment model, performing an operating status analysis to determine the health score of each device in the power system, the health score being used to assess the failure probability of the device; and generating a reasoning and analysis report based on the output results of each container and the health score of each device in the power system.

[0009] In a possible implementation, based on the meta-information module, feature extraction is performed on real-time operation data to obtain standard features, including: using the service script module of the meta-information module to extract key features in image data, the key features include surface temperature distribution features and current thermal distribution features; using the service script module of the meta-information module to extract key features in parameter data, the key features include current, current offset, current change rate, voltage, voltage offset, voltage change rate, load state and load fluctuation range; based on the equipment type of each equipment in the power system, the operating environment of the container and the model type of the power system model, feature conversion is performed on the key features of the image data and the parameter data to obtain conversion features; based on the conversion features and the data format of each container service interface, format conversion is performed to obtain standard features.

[0010] In a possible implementation, based on the device type of each device in the power system, the operating environment of the container and the model type of the power system model, feature conversion is performed on key features of image data and parameter data to obtain conversion features, including: determining feature requirements based on the device type of each device in the power system, the feature requirements are used to determine features related to the device type in the key features; based on the feature requirements, feature extraction is performed on the key features to obtain relevant features of each device; based on the operating environment of the container and the model type of the power system model, the data format and data dimension of the conversion features are determined; based on the data format and data dimension of the conversion features, as well as the relevant features of each device, feature conversion is performed to obtain conversion features.

[0011] In a possible implementation, standard features are input into the container of each power system model to obtain output results of each container, including: based on the input side service interface information of the container of each power system model, feature extraction is performed on the standard features to obtain input features corresponding to each container; based on the input features corresponding to each container and the power system model in each container, the model output results are obtained; based on the output side service interface information of each container, feature conversion is performed on the model output results to obtain the output results of each container.

[0012] In a possible implementation, an operation status analysis is performed based on the output results of each container and a preset health assessment model to determine the health score of each device in the power system, including: integrating features based on the output results of each container and predefined rules of the cloud information module to obtain integrated features; splitting the integrated features based on the identification information of each device in the power system to obtain evaluation features of each device; and performing an operation status analysis on the evaluation features of each device and a preset health assessment model to obtain a health score of each device in the power system.

[0013] In a possible implementation, a reasoning analysis report is generated based on the output results of each container and the health score of each device in the power system, including: based on the output results of each container, determining load forecast data and demand analysis results, the load forecast data including the load change trend within the forecast period, and the demand analysis results including the load demand within the forecast era; based on the health score of each device in the power system, determining the fault information of each device, the fault information including the possibility of fault, fault type prediction and remaining service life estimation; based on the load forecast data and demand analysis results, as well as the correlation between each load and each device, determining the risk level of each device; based on the framework of the reasoning analysis report, the load forecast data, the demand analysis results, the fault information and risk level of each device, generating a reasoning analysis report.

[0014] In a possible implementation, the method also includes: obtaining the system environment of the power system model to be deployed, and the model files of each power system model; based on the system environment, determining resource information, the resource information including CPU, memory and hard disk capacity; based on the resource information, determining the service interface information and planned occupied space of each power system model; based on the service interface information and planned occupied space of each power system model, and the model files of each power system model, performing knowledge distillation on each power system model to obtain the model files of each power system model after distillation; based on the system environment and each power system model, determining the container corresponding to each power system model; based on the container corresponding to each power system model, and the model files of each power system model after distillation, performing model deployment to obtain multiple model containers; based on the multiple model containers, determining the image file; sending the image file to the device to be deployed to implement the power system model deployment.

[0015] In a possible implementation, the method also includes: obtaining real-time operating data of multiple devices with known health information in a historical period, and health information of the multiple devices; the health information includes whether there is a fault and the type of fault; based on the real-time operating data of the multiple devices with known health information, feature extraction is performed to obtain standard features of the known health information; based on the standard features of the known health information and the containers of each power system model, the output results of the known health information are determined; based on the output results of the known health information, feature integration and splitting are performed to obtain evaluation features of multiple devices with known health information; using the evaluation features of multiple devices with known health information as input and the health information of multiple devices with known health information as output, a plurality of training samples are determined; based on the training samples, neural network training is performed to obtain a health assessment model; based on the health assessment model and the container corresponding to the health assessment model, a model container of the health assessment model is generated; based on the model container of the health assessment model, an image file of the model container of the health assessment model is generated; the image file of the model container of the health assessment model is sent to the device to be deployed to implement the deployment of the health assessment model.

[0016] In the second aspect, an embodiment of the present invention provides a container-based power system model reasoning and analysis device, including: a communication module, used to obtain real-time operation data in the power system, the real-time operation data including image data and parameter data; a processing module, used to extract features of the real-time operation data based on the metadata module to obtain standard features; the standard features are input into the containers of each power system model to obtain the output results of each container; the power system model includes a transformer diagnosis model, a circuit breaker fault diagnosis model and a load prediction model; based on the output results of each container and a preset health assessment model, an operation status analysis is performed to determine the health score of each device in the power system, and the health score is used to assess the failure probability of the equipment; based on the output results of each container and the health score of each device in the power system, an reasoning analysis report is generated.

[0017] In a possible implementation, the processing module is specifically used to use the service script module of the meta-information module to extract key features in the image data, the key features including surface temperature distribution features and current thermal distribution features; use the service script module of the meta-information module to extract key features in the parameter data, the key features including current, current offset, current change rate, voltage, voltage offset, voltage change rate, load state and load fluctuation range; based on the equipment type of each equipment in the power system, the operating environment of the container and the model type of the power system model, the key features of the image data and the parameter data are converted to obtain conversion features; based on the conversion features and the data format of each container service interface, format conversion is performed to obtain standard features.

[0018] In a third aspect, an embodiment of the present invention provides an electronic device, comprising a memory and a processor, wherein the memory stores a computer program, and the processor is used to call and run the computer program stored in the memory to execute the method described in the first aspect and any possible implementation method of the first aspect.

[0019] In a fourth aspect, an embodiment of the present invention provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and wherein when the computer program is executed by a processor, the steps of the method described in the first aspect and any possible implementation method of the first aspect are implemented.

[0020] The present invention provides a method and device for reasoning and analyzing power system models based on containerization. The present invention establishes containers that carry various power system models, enables various power system models to run in the same system environment, and uses containerization technology to achieve compatibility of multiple power system models. The present invention inputs the standard features obtained from real-time operation data into each container, runs each power system model, and obtains the output results of the transformer diagnosis model, the circuit breaker fault diagnosis model, and the load prediction model. After that, a health assessment model is used to analyze the operating status of each output result, determine the health score that characterizes the probability of failure of each device, and then generate a reasoning analysis report to intuitively display the overall situation of the power system to the operation and maintenance personnel, reduce the analysis difficulty of the operation and maintenance personnel, improve the operation and inspection efficiency of the power system, and enhance the intelligent operation and maintenance level of the power system. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0022] Figure 1 It is a flowchart of a container-based power system model reasoning and analysis method provided by an embodiment of the present invention;

[0023] Figure 2 It is a structural schematic diagram of a container-based power system model reasoning and analysis device provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0024] In the following description, specific details such as specific system structures, technologies, etc. are provided for the purpose of illustration rather than limitation, so as to provide a thorough understanding of the embodiments of the present invention. However, it should be clear to those skilled in the art that the present invention may be implemented in other embodiments without these specific details. In other cases, detailed descriptions of well-known systems, devices, circuits, and methods are omitted to prevent unnecessary details from obstructing the description of the present invention.

[0025] In the description of the present invention, unless otherwise specified, " / " means "or", for example, A / B can mean A or B. "And / or" in this article is only a description of the association relationship of associated objects, indicating that there can be three relationships. For example, A and / or B can mean: A exists alone, A and B exist at the same time, and B exists alone. In addition, "at least one" and "plurality" refer to two or more. The words "first", "second", etc. do not limit the quantity and execution order, and the words "first", "second", etc. do not limit them to be different.

[0026] 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 design 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 designs. Specifically, the use of words such as "exemplary" or "for example" is intended to present related concepts in a concrete way for easy understanding.

[0027] In addition, the terms "including" and "having" and any variations thereof mentioned in the description of this application are intended to cover non-exclusive inclusions. For example, a process, method, system, product or device that includes a series of steps or modules is not limited to the listed steps or modules, but may optionally include other steps or modules that are not listed, or may optionally include other steps or modules that are inherent to these processes, methods, products or devices.

[0028] In order to make the purpose, technical solutions and advantages of the present invention more clear, the following will be described through specific embodiments in conjunction with the accompanying drawings of the present invention.

[0029] like Figure 1 As shown, an embodiment of the present invention provides a container-based power system model reasoning and analysis method, which includes steps S101-S105.

[0030] S101. Acquire real-time operation data in the power system.

[0031] In the embodiment of the present application, the real-time operation data includes image data and parameter data.

[0032] In some embodiments, image data includes inspection image data of power equipment (such as transformers, distribution cabinets, line connectors, etc.), such as infrared thermal imaging, visible light images, etc. These data are collected in real time by cameras or sensors installed on the power equipment.

[0033] In some embodiments, parameter data includes real-time operating parameters of power equipment, such as temperature, voltage, current, load status, etc. These data are collected in real time by monitoring equipment (such as temperature sensors, voltage transformers, current transformers, etc.) of the power system.

[0034] S102: Based on the meta-information module, feature extraction is performed on the real-time operation data to obtain standard features.

[0035] It should be noted that the meta-information module, as the data coordination core of the service image, stores the detailed rules required for data preprocessing. For image data, the meta-information module drives the service script module to perform image preprocessing, such as denoising, enhancement, resizing, etc., and then uses the feature generator to extract key features in the image, such as temperature distribution on the surface of the device, abnormal hot spots, etc. For parameter data, the meta-information module calls the corresponding algorithm module for numerical processing according to predefined rules, such as missing value filling, normalization, sliding window segmentation, etc., to eliminate noise and outliers in the data and extract key features, such as the average, maximum, and rate of change of temperature. The meta-information module is also responsible for integrating the preprocessing results of different types of data, and specifies the flow of processed data through mapping rules to ensure that the data can be directly input into the model container.

[0036] As a possible implementation manner, step S102 may be specifically implemented as steps S1021 - S1024 .

[0037] S1021. Utilize the service script module of the meta-information module to extract key features from image data.

[0038] In some embodiments, key features in image data include surface temperature distribution features and current thermal distribution features.

[0039] Exemplarily, an embodiment of the present invention may receive raw inspection image data from a power system, such as images of transformers or switchgear, infrared thermal imaging, etc. The image data is preprocessed using the image processing algorithm in the service script module, including denoising, cropping, and size normalization. Key features in the image are extracted, such as the surface temperature distribution features of the equipment, the thermal distribution features of the current, etc. These features can reflect the operating status of the equipment, such as whether there are abnormal conditions such as overheating and arcing. The image processing algorithm may include Gaussian filtering, edge detection, morphological processing, etc. The feature extraction process may require the use of deep learning models, such as convolutional neural networks (CNNs) to identify key information in the image.

[0040] S1022. Utilize the service script module of the meta-information module to extract key features from the parameter data.

[0041] In some embodiments, key features in the parameter data include current, current offset, current change rate, voltage, voltage offset, voltage change rate, load state, and load fluctuation range.

[0042] Exemplarily, an embodiment of the present invention can receive real-time parameter data such as current, voltage, load status, etc. from a power system. The parameter data is preprocessed using the data processing algorithm in the service script module, including missing value filling, outlier detection and processing, etc. Key features in the parameter data are extracted, such as current, current offset, current change rate, voltage, voltage offset, voltage change rate, load status, and load fluctuation range, etc. The data preprocessing process may include smoothing, normalization, etc. The feature extraction process may need to be based on statistical analysis and machine learning algorithms to identify patterns and trends in the parameter data.

[0043] S1023. Based on the equipment type of each equipment in the power system, the operating environment of the container and the model type of the power system model, key features of the image data and parameter data are converted to obtain conversion features.

[0044] It should be noted that the embodiments of the present invention can determine the rules for feature conversion according to the equipment type (such as transformer, circuit breaker) and the operating environment of the container (such as the cloud, edge). According to the model type of the power system model (such as transformer diagnosis model, circuit breaker fault diagnosis model, load forecasting model), adjust the rules for feature conversion to ensure the compatibility of the features with the target model. The extracted key features are converted to obtain conversion features. Feature conversion may include data scaling, feature selection, feature combination, etc. The conversion process needs to take into account the particularities of different equipment types and operating environments, as well as the different requirements of different models for features.

[0045] Exemplarily, step S1023 may be specifically implemented as steps A1-A4.

[0046] A1. Determine the characteristic requirements based on the equipment type of each equipment in the power system.

[0047] In some embodiments, feature requirements are used to determine features of key features that are relevant to the device type.

[0048] Exemplarily, the embodiments of the present invention determine the required key features according to the specific types of equipment in the power system (such as transformers, circuit breakers, lines, etc.) These feature requirements are usually closely related to the operating principles, failure modes and monitoring requirements of the equipment.

[0049] For transformers, we need to pay attention to parameters such as oil temperature, winding temperature, oil level, current and voltage, as well as image data such as oil chromatography analysis results. These features can reflect the operating status of the transformer and potential failure risks.

[0050] For circuit breakers, we need to pay attention to parameters such as opening and closing time, opening and closing speed, contact wear, number of operations, and images of the circuit breaker's appearance. These features help evaluate the performance and remaining life of the circuit breaker.

[0051] For the lines, it is necessary to pay attention to line parameters such as current, voltage, power factor, load conditions, line temperature, and monitoring images of the line corridors, in order to determine the overload condition, short circuit risk, and insulation status of the lines.

[0052] A2. Based on feature requirements, key features are extracted to obtain relevant features of each device.

[0053] Exemplarily, an embodiment of the present invention extracts key features related to the device type from real-time operation data. This usually involves preprocessing work such as cleaning, denoising, missing value processing, and outlier detection of the original data, and extracting features closely related to the device operation status from the preprocessed data using a feature generator. For parameter data, we can use statistical methods (such as mean, variance, maximum value, minimum value, etc.) and time series analysis methods (such as sliding windows, autoregressive models, etc.) to extract features. For image data, we can use image processing techniques (such as image enhancement, edge detection, feature point extraction, etc.) and machine learning algorithms (such as convolutional neural networks, support vector machines, etc.) to extract features.

[0054] A3. Determine the data format and data dimension of the conversion feature based on the operating environment of the container and the model type of the power system model.

[0055] Exemplarily, the embodiment of the present invention determines the data format and data dimension of the conversion feature according to the operating environment of the container (such as the cloud, edge or terminal) and the model type of the power system model (such as a deep learning model, a machine learning model or a traditional physical model). The data format may include JSON, CSV, NumPy array, etc., depending on the input requirements of the model. The data dimension depends on the number of input features required by the model, as well as the correlation and independence between these features.

[0056] A4. Based on the data format and data dimension of the conversion feature and the relevant features of each device, perform feature conversion to obtain the conversion feature.

[0057] Exemplarily, the embodiments of the present invention will convert the relevant features of each device according to the data format and data dimension requirements of the conversion features. This usually involves operations such as scaling, normalization, encoding (such as one-hot encoding, label encoding, etc.) and mapping of the data to ensure that the converted features meet the input requirements of the model. For numerical features, we can use min-max normalization, z-score standardization and other methods for scaling and normalization. For categorical features, we can use one-hot encoding, label encoding and other methods for encoding. For time series features, we can use sliding windows, time differences and other methods for feature extraction and conversion.

[0058] The embodiment of the present invention converts key features in real-time operation data into conversion features that meet model input requirements, thereby providing reliable data support for subsequent power system model reasoning and health assessment.

[0059] S1024: Based on the conversion feature and the data format of each container service interface, format conversion is performed to obtain a standard feature.

[0060] Exemplarily, an embodiment of the present invention may determine the rules for format conversion according to the data format requirements of each container service interface. The conversion features are format converted, including data type conversion, data structure adjustment, etc., to ensure that the features meet the requirements of the service interface. The standard features are obtained and input into the container of each power system model. The format conversion process may require the use of data exchange formats such as JSON and XML. The conversion process needs to consider the data compatibility and consistency between different service interfaces.

[0061] In this way, the embodiment of the present invention can obtain standard features that meet the requirements of each container service interface by performing feature extraction, feature conversion and format conversion on image data and parameter data, thereby realizing efficient processing and utilization of real-time operation data in the power system.

[0062] S103, inputting the standard features into the containers of each power system model to obtain output results of each container.

[0063] In an embodiment of the present application, the power system model includes a transformer diagnosis model, a circuit breaker fault diagnosis model and a load prediction model.

[0064] It should be noted that the transformer diagnosis model: receives standard features related to the transformer (such as temperature distribution, vibration frequency, etc.), diagnoses the health status of the transformer, and outputs the health score and fault warning information of the transformer. The circuit breaker fault diagnosis model: receives standard features related to the circuit breaker (such as current waveform, contact temperature, etc.), diagnoses the fault of the circuit breaker, and outputs the health score and fault type of the circuit breaker. The load forecasting model: receives the load data and other relevant features of the power system (such as historical load data, weather forecast, etc.), forecasts the load of the power system, and outputs the future load forecast value.

[0065] As a possible implementation manner, step S103 may be specifically implemented as steps S1031 - S1033 .

[0066] S1031. Based on the input-side service interface information of the container of each power system model, feature extraction is performed on the standard features to obtain input features corresponding to each container.

[0067] Exemplarily, an embodiment of the present invention can further extract or screen the standard features previously obtained based on the input-side service interface information of each power system model container to ensure that they match the input requirements of the model in the container. Interface information parsing requires parsing the input-side service interface information of each container to understand the input feature type, quantity, format and other requirements required by the model. Feature matching and extraction, matching the standard features with the input requirements in the interface information, extracting the features that meet the requirements as the input features of each container. If the standard features contain some unnecessary or redundant information, we also need to perform feature screening to remove this information.

[0068] S1032. Obtain a model output result based on the input characteristics corresponding to each container and the power system model in each container.

[0069] Exemplarily, an embodiment of the present invention can input the extracted input features into the power system model in each container to perform model reasoning or calculation to obtain the output results of the model. Model loading and initialization: First, we need to load and initialize the power system model in each container to ensure that they are in a usable state. Input the input features into the model for reasoning or calculation. This usually involves processes such as data preprocessing, model calculation, and result generation. Obtain output results from the model, which may include key information such as the health status of the equipment, fault type, and remaining life.

[0070] S1033. Based on the output service interface information of each container, perform feature conversion on the model output result to obtain the output result of each container.

[0071] Exemplarily, the embodiment of the present invention can perform necessary feature conversion on the model output results according to the output side service interface information of each container to ensure that they meet the interface requirements and can be used by subsequent systems or applications. Parse the output side service interface information of each container to understand the format, type and other requirements of the model output results. Perform feature conversion on the model output results, which may include operations such as formatting, encoding, and mapping of data to ensure that they meet the interface requirements. Output the converted output results to the corresponding container for use by subsequent systems or applications.

[0072] In this way, standard features can be input into the containers of each power system model and output results that meet the requirements can be obtained, thus providing strong support for tasks such as monitoring, diagnosis, and prediction of the power system.

[0073] S104: Based on the output results of each container and a preset health assessment model, an operation status analysis is performed to determine the health score of each device in the power system.

[0074] In the embodiment of the present application, the health score is used to evaluate the failure probability of the equipment.

[0075] It should be noted that the health assessment model is a comprehensive assessment model. It comprehensively assesses the health status of each device in the power system based on the output results of each container (such as the health score of the transformer, the health score of the circuit breaker, the load forecast value, etc.), as well as the equipment's operating history data, maintenance records and other information. The health assessment model outputs the health score of the equipment, which is used to assess the failure probability of the equipment. The higher the score, the better the health status of the equipment and the lower the failure probability; the lower the score, the worse the health status of the equipment and the higher the failure probability.

[0076] As a possible implementation manner, step S104 may be specifically implemented as steps S1041 - S1043 .

[0077] S1041. Based on the output results of each container and the predefined rules of the cloud information module, feature integration is performed to obtain integrated features.

[0078] Exemplarily, an embodiment of the present invention can integrate the output results from different containers to form a complete feature set that can reflect the operating status of the power system. Collect the output results from each container, which may include real-time monitoring data of the equipment, fault diagnosis results, prediction information, etc. Integrate these output results using predefined rules stored in the cloud information module. These rules may include data merging, data conversion, data filtering, etc. to ensure that the integrated feature set is consistent and accurate. According to the predefined rules, the output results from different containers are integrated into a unified feature set, i.e., integrated features. These features will be used for subsequent evaluation and analysis.

[0079] S1042. Based on the identification information of each device in the power system, the integrated features are split to obtain evaluation features of each device.

[0080] Exemplarily, the embodiment of the present invention can split the integrated features according to the identification information of each device in the power system to obtain the evaluation features of each device. Extract the identification information of each device from the power system, such as device ID, device type, device location, etc. According to the device identification information, the integrated features are split into feature sets corresponding to each device, namely evaluation features. These features will be used for subsequent health assessments.

[0081] S1043. Analyze the operating status of each device using the evaluation characteristics of each device and a preset health evaluation model to obtain a health score for each device in the power system.

[0082] Exemplarily, an embodiment of the present invention can input the evaluation characteristics of each device into a preset health assessment model to perform an operating status analysis to obtain a health score for the device. Load a preset health assessment model, which may be built based on algorithms such as machine learning and deep learning, and is used to evaluate the operating status of the device. Input the evaluation characteristics of each device into the model to perform an operating status analysis. This includes processes such as data preprocessing and model calculation. Based on the output results of the model, calculate the health score of each device. This score may be a numerical value used to represent key information such as the health status or remaining life of the device.

[0083] Based on the output results of each container and the preset health assessment model, the operating status of each device in the power system is analyzed and their health scores are determined. This helps to timely detect potential failures or abnormal conditions of the equipment and provide strong protection for the safe operation of the power system.

[0084] S105. Generate a reasoning analysis report based on the output results of each container and the health score of each device in the power system.

[0085] It should be noted that the reasoning analysis report is a comprehensive report, including: the output results of each container: including the output of the transformer diagnosis model, the output of the circuit breaker fault diagnosis model, the output of the load forecasting model, etc. The health score of each device: the equipment health score calculated according to the health assessment model. Equipment failure warning information: Based on the health score and the preset threshold, the equipment with failure risk is warned. Maintenance suggestions: Based on the health status of the equipment and the fault warning information, corresponding maintenance suggestions are made, such as replacing parts, adjusting operating parameters, etc. Other relevant information: such as basic information of the equipment, operation history data, maintenance records, etc.

[0086] As a possible implementation manner, step S105 can be specifically implemented as steps S1051-S1053.

[0087] S1051. Determine load forecast data and demand analysis results based on the output results of each container.

[0088] In some embodiments, the load forecasting data includes load variation trends within a forecast period, and the demand analysis results include load demands within the forecast era.

[0089] For example, the embodiment of the present invention can collect historical load data, weather data, holiday data and other factors that may affect load changes from each container. The collected data is processed using a load forecasting model (such as time series analysis, machine learning model, etc.) to predict the load change trend in the future (forecast period). Key information such as the load change trend chart, load peak value, load valley value, etc. within the forecast period is output.

[0090] For example, the embodiment of the present invention can classify loads according to different user types (such as residential, industrial, commercial, etc.) in the power system. Based on historical load data and user behavior patterns, the demand for various types of loads in the future period is predicted. Key information such as the demand amount and demand growth rate of various types of loads in the forecast period is output.

[0091] S1052. Determine the fault information of each device based on the health score of each device in the power system.

[0092] In some embodiments, the fault information includes a probability of failure, a prediction of a fault type, and a remaining useful life estimate.

[0093] For example, the embodiments of the present invention can evaluate the overall health status of the equipment according to the health score of each equipment. Combine the health score of the equipment and the historical fault data to predict the possibility of equipment failure using statistical methods or machine learning models. Based on the equipment's operating data, historical fault records, and fault tree analysis methods, predict the possible types of failures. Estimate the remaining service life of the equipment based on the equipment's health score, fault history, and service life data provided by the manufacturer.

[0094] S1053. Determine the risk level of each device based on the load forecast data and demand analysis results, as well as the correlation between each load and each device.

[0095] For example, the embodiment of the present invention can use correlation analysis, cluster analysis and other methods to determine the degree of association between each load and each device. Combine the load forecast data, demand analysis results and equipment failure information to calculate the risk level of each device in the forecast period. The risk level may include multiple dimensions such as failure probability and failure impact. Output the risk level report of each device, including a list of high-risk devices, risk level classification and other information.

[0096] S1054. Generate a reasoning analysis report based on the framework of the reasoning analysis report, load forecast data, demand analysis results, fault information of each device and risk level.

[0097] Exemplarily, the embodiments of the present invention can design the framework of the reasoning analysis report according to the needs and purposes of the report, including an introduction, method introduction, data analysis results, conclusions and suggestions. Integrate key information such as load forecasting data, demand analysis results, equipment failure information and risk levels into the report. Use charts, graphs and other tools to intuitively display data analysis results, such as load change trend charts, equipment failure distribution charts, risk level radar charts, etc. Based on the data analysis results, put forward targeted conclusions and suggestions, such as strengthening equipment maintenance, optimizing load distribution, and formulating emergency plans. The integrated data, charts, conclusions and suggestions are compiled into a complete reasoning analysis report, and output in PDF, Word and other formats for relevant personnel to review.

[0098] In this way, the present invention systematically generates a reasoning analysis report containing multi-dimensional contents such as load forecasting, demand analysis, equipment failure information and risk level, providing a scientific basis for the operation and management of the power system.

[0099] The present invention provides a container-based power system model reasoning and analysis method. By establishing a container that carries each power system model, various power system models are operated in the same system environment, and containerization technology is used to achieve compatibility of multiple power system models. The present invention inputs the standard features obtained from real-time operation data into each container, operates each power system model, and obtains the output results of the transformer diagnosis model, the circuit breaker fault diagnosis model, and the load prediction model. After that, a health assessment model is used to analyze the operating status of each output result, determine the health score that characterizes the probability of failure of each device, and then generate a reasoning analysis report to intuitively display the overall situation of the power system to the operation and maintenance personnel, reduce the analysis difficulty of the operation and maintenance personnel, improve the operation and inspection efficiency of the power system, and enhance the intelligent operation and maintenance level of the power system.

[0100] Optionally, the containerized power system model reasoning and analysis method provided in an embodiment of the present invention further includes steps S201-S208.

[0101] S201. Obtain the system environment of the power system model to be deployed and the model files of each power system model.

[0102] Exemplarily, an embodiment of the present invention can communicate with the operation and maintenance team of the power system to be deployed to understand the system's hardware configuration (such as server model, number and model of CPUs, memory capacity, hard disk capacity, etc.), operating system type and version, network architecture, etc. Confirm whether the system supports containerization technology (such as Docker, Kubernetes, etc.) and whether there are specific security or performance requirements. Obtain the original model files of each power system model from the R&D team or model warehouse. These files may include trained neural network models, algorithm scripts, configuration files, etc. Ensure the integrity and correctness of the model files, and perform verification tests when necessary.

[0103] S202: Determine resource information based on the system environment.

[0104] In some embodiments, the resource information includes CPU, memory, and hard disk capacity.

[0105] For example, the embodiment of the present invention can record the number and model of CPUs, the total capacity of memory, the capacity and partition of hard disks in detail according to the description of the system environment, and consider whether it is necessary to reserve additional resources for the power system model to cope with peak demand.

[0106] S203: Determine the service interface information and planned occupied space of each power system model based on the resource information.

[0107] Exemplarily, the embodiments of the present invention can analyze the functional requirements of each power system model, determine the service interface (such as RESTful API, WebSocket, etc.) and its parameters and return value format that each model needs to provide. Determine the data interaction method between models, such as direct call, message queue, database sharing, etc. Plan reasonable storage space for each model based on the size of the model file, memory usage at runtime, disk I / O requirements, etc. Consider the additional space that may be required when the model is upgraded or updated.

[0108] S204. Based on the service interface information and planned occupied space of each power system model and the model files of each power system model, knowledge distillation is performed on each power system model to obtain the model files of each power system model after distillation.

[0109] Exemplarily, knowledge distillation is performed on a complex power system model. The embodiment of the present invention can train a smaller model (student model) to imitate the behavior of the original large model (teacher model) while maintaining high prediction accuracy. During the distillation process, the architecture, hyperparameters, etc. of the student model can be adjusted to optimize its performance and reduce resource usage. The model file after distillation should contain the weights, architecture information, configuration files, etc. of the student model.

[0110] S205. Based on the system environment and each power system model, determine the container corresponding to each power system model.

[0111] For example, the embodiment of the present invention can select a suitable container platform (such as Docker) according to the containerization technology supported by the system environment, and create an independent container for each power system model to ensure isolation and security between containers.

[0112] S206: Based on the containers corresponding to the power system models and the model files of the power system models after distillation, model deployment is performed to obtain multiple model containers.

[0113] Exemplarily, the embodiment of the present invention can deploy the distilled model file to the corresponding container, including copying the model file to a specified directory in the container, configuring environment variables, starting services, etc. Ensure that the container can correctly load and run the model and provide the expected service interface.

[0114] S207: Determine an image file based on multiple model containers.

[0115] Exemplarily, the embodiment of the present invention can use the tools provided by the container platform (such as Docker's dockerbuild command) to package the container containing the model file into an image file. The image file should contain all the configuration information, dependent libraries, model files, etc. of the container so that it can be quickly started and run on the target device.

[0116] S208. Send the image file to the device to be deployed to implement the power system model deployment.

[0117] Exemplarily, the embodiments of the present invention can use a secure file transfer protocol (such as SSH, SCP, FTP, etc.) to transfer the image file from the development environment to the device to be deployed. Ensure the integrity and security of the file during the transmission process. On the device to be deployed, use the command provided by the container platform (such as the Docker run command of Docker) to start the image file, create and run the model container. Perform functional and performance tests on the model container to ensure that it can provide services correctly and meet performance requirements.

[0118] In this way, the embodiment of the present invention can systematically complete the deployment of the power system model and ensure that the model can run efficiently and stably on the target device.

[0119] Optionally, the containerized power system model reasoning and analysis method provided in an embodiment of the present invention further includes steps S301-S309.

[0120] S301. Acquire real-time operation data of multiple devices with known health information within a historical period, and health information of multiple devices.

[0121] In some embodiments, the health information includes whether a fault occurs and the type of fault.

[0122] For example, the embodiments of the present invention can obtain real-time operation data of multiple devices in a historical period from the monitoring system of the power system, including sensor readings of voltage, current, power, temperature, humidity, etc. The health information of the corresponding device can be obtained from the maintenance record or fault log of the device, including whether it is faulty (yes / no) and the type of fault (such as overheating, short circuit, overload, etc.).

[0123] S302: Perform feature extraction based on the real-time operation data of multiple devices with known health information to obtain standard features of the known health information.

[0124] Exemplarily, the embodiments of the present invention can pre-process the collected real-time operation data, such as denoising, normalization, etc., extract features that have an important impact on the health status of the equipment, such as voltage fluctuation rate, current peak, power factor, temperature change trend, etc. Integrate the extracted features into a standard feature set of known health information, and each device corresponds to a feature vector.

[0125] S303. Based on the standard features of the known health information and the containers of each power system model, determine the output results of the known health information.

[0126] Exemplarily, embodiments of the present invention can utilize containers of existing power system models (such as load forecasting models, fault diagnosis models, etc.) to process standard features of known health information. These models may output information such as equipment status assessment and fault warning, but this information is mainly used in this step to verify the accuracy of subsequent health assessment models. Record these output results for comparison with subsequent health assessment model results.

[0127] S304: Based on the output results of the known health information, feature integration and splitting are performed to obtain evaluation features of multiple devices of the known health information.

[0128] S305: Determine multiple training samples using evaluation features of multiple devices with known health information as input and health information of multiple devices with known health information as output.

[0129] S306: Based on the training samples, neural network training is performed to obtain a health assessment model.

[0130] Exemplarily, the embodiments of the present invention can select a suitable neural network architecture (such as a multilayer perceptron, a convolutional neural network, a recurrent neural network, etc.) and configure corresponding hyperparameters (such as a learning rate, number of iterations, batch size, etc.). The neural network is trained using a training sample data set, with the goal of enabling the model to accurately predict the health status of the device. During the training process, cross-validation, early stopping and other techniques are used to prevent overfitting and monitor the performance of the model.

[0131] S307: Generate a model container of the health assessment model based on the health assessment model and the container corresponding to the health assessment model.

[0132] Exemplarily, the present invention encapsulates the trained health assessment model into a container so that it can be deployed and run in different environments. The container should contain all dependencies, configuration files, and runtime environments required by the model.

[0133] S308. Generate an image file of the model container of the health assessment model based on the model container of the health assessment model.

[0134] Exemplarily, the present invention uses container technology (such as Docker) to package the model container into an image file. The image file should contain all layers and configuration information of the container so that it can be quickly started and run on the target device.

[0135] S309: Send the image file of the model container of the health assessment model to the device to be deployed to implement the deployment of the health assessment model.

[0136] Exemplarily, the present invention enables the generated image file to be transferred to the device to be deployed. The image file is loaded and run on the device to be deployed using container technology (such as Docker) to create a model container of the health assessment model. The model container is configured and tested as necessary to ensure that it can run correctly and output the health assessment results.

[0137] In this way, the embodiment of the present invention can systematically complete the construction, training and deployment of the health assessment model, and provide strong support for equipment health monitoring of the power system.

[0138] It should be understood that the order of execution of the steps in the above embodiment does not necessarily mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiment of the present invention.

[0139] The following is an embodiment of the device of the present invention. For details not described in detail therein, reference may be made to the corresponding method embodiment described above.

[0140] Figure 2 The schematic diagram of the structure of a container-based power system model reasoning and analysis device provided by an embodiment of the present invention is shown. The reasoning and analysis device 400 includes a communication module 401 and a processing module 402 .

[0141] The communication module 401 is used to obtain real-time operation data in the power system, and the real-time operation data includes image data and parameter data.

[0142] The processing module 402 is used to extract features of the real-time operation data based on the meta-information module to obtain standard features; input the standard features into the containers of each power system model to obtain the output results of each container; the power system model includes a transformer diagnosis model, a circuit breaker fault diagnosis model and a load prediction model; based on the output results of each container and a preset health assessment model, an operation status analysis is performed to determine the health score of each device in the power system, and the health score is used to assess the failure probability of the equipment; based on the output results of each container and the health score of each device in the power system, an inference analysis report is generated.

[0143] The embodiments described above are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features may be replaced by equivalents. Such modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included in the protection scope of the present invention.

Claims

1. A container-based power system model reasoning and analysis method, characterized in that: include: Acquiring real-time operation data in the power system, wherein the real-time operation data includes image data and parameter data; Based on the meta-information module, feature extraction is performed on the real-time operation data to obtain standard features; Inputting the standard features into the containers of each power system model to obtain output results of each container; the power system model includes a transformer diagnosis model, a circuit breaker fault diagnosis model and a load prediction model; Based on the output results of each container and a preset health assessment model, an operation status analysis is performed to determine the health score of each device in the power system, where the health score is used to assess the failure probability of the device; Based on the output results of each container and the health score of each device in the power system, a reasoning analysis report is generated.

2. The container-based power system model reasoning and analysis method according to claim 1 is characterized in that: The feature extraction of the real-time operation data based on the meta-information module to obtain standard features includes: Utilizing the service script module of the meta-information module, extracting key features from the image data, wherein the key features include surface temperature distribution features and current heat distribution features; Utilizing the service script module of the meta-information module, extracting key features in the parameter data, wherein the key features include current, current offset, current change rate, voltage, voltage offset, voltage change rate, load state and load fluctuation range; Based on the equipment type of each equipment in the power system, the operating environment of the container and the model type of the power system model, the key features of the image data and the parameter data are converted to obtain conversion features; Based on the conversion feature and the data format of each container service interface, format conversion is performed to obtain the standard feature.

3. The container-based power system model reasoning and analysis method according to claim 1 is characterized in that: The method of performing feature conversion on key features of the image data and parameter data based on the device type of each device in the power system, the operating environment of the container, and the model type of the power system model to obtain conversion features includes: Determining feature requirements based on the device type of each device in the power system, wherein the feature requirements are used to determine features related to the device type among the key features; Based on the feature requirements, feature extraction is performed on key features to obtain relevant features of each device; Determining a data format and a data dimension of a conversion feature based on an operating environment of the container and a model type of the power system model; Based on the data format and data dimension of the conversion feature and the relevant features of each device, feature conversion is performed to obtain the conversion feature.

4. The container-based power system model reasoning and analysis method according to claim 1 is characterized in that: The step of inputting the standard features into the containers of each power system model to obtain output results of each container includes: Based on the input side service interface information of the container of each power system model, feature extraction is performed on the standard feature to obtain the input feature corresponding to each container; Based on the input features corresponding to each container and the power system model in each container, a model output result is obtained; Based on the output service interface information of each container, feature conversion is performed on the model output result to obtain the output result of each container.

5. The container-based power system model reasoning and analysis method according to claim 1 is characterized in that: The operation status analysis is performed based on the output results of each container and the preset health assessment model to determine the health score of each device in the power system, including: Based on the output results of each container and the predefined rules of the cloud information module, feature integration is performed to obtain integrated features; Based on the identification information of each device in the power system, the integrated features are split to obtain evaluation features of each device; The evaluation characteristics of each device and the preset health assessment model are used to analyze the operating status to obtain the health score of each device in the power system.

6. The container-based power system model reasoning and analysis method according to claim 1 is characterized in that: The inference analysis report is generated based on the output results of each container and the health score of each device in the power system, including: Determine load forecast data and demand analysis results based on the output results of each container, wherein the load forecast data includes the load change trend within the forecast period, and the demand analysis results include the load demand within the forecast period; Determine fault information of each device based on the health score of each device in the power system, wherein the fault information includes fault probability, fault type prediction and remaining service life estimation; Determine the risk level of each device based on load forecast data and demand analysis results, as well as the correlation between each load and each device; The reasoning analysis report is generated based on the framework of the reasoning analysis report, load forecast data, demand analysis results, fault information of each device and risk level.

7. The container-based power system model reasoning and analysis method according to claim 1 is characterized in that: The method further comprises: Obtain the system environment of the power system model to be deployed and the model files of each power system model; Based on the system environment, determining resource information, the resource information including CPU, memory and hard disk capacity; Based on the resource information, determining service interface information and planned occupied space of each power system model; Based on the service interface information and planned occupied space of each power system model and the model files of each power system model, knowledge distillation is performed on each power system model to obtain the model files of each power system model after distillation; Based on the system environment and each power system model, determining a container corresponding to each power system model; Based on the containers corresponding to each power system model and the model files of each power system model after distillation, the model is deployed to obtain multiple model containers; Determine the image file based on multiple model containers; The image file is sent to the device to be deployed to implement the power system model deployment.

8. The container-based power system model reasoning and analysis method according to claim 1 is characterized in that: The method further comprises: Acquire real-time operation data of multiple devices with known health information in a historical period, as well as health information of the multiple devices; the health information includes whether there is a fault and the type of fault; Based on the real-time operation data of multiple devices with known health information, feature extraction is performed to obtain standard features of the known health information; Determine output results of the known health information based on standard features of the known health information and containers of each power system model; Based on the output results of known health information, feature integration and splitting are performed to obtain evaluation features of multiple devices with known health information; Taking evaluation features of multiple devices with known health information as input and health information of multiple devices with known health information as output, determining multiple training samples; Based on the training samples, neural network training is performed to obtain a health assessment model; Based on the health assessment model and the container corresponding to the health assessment model, generate a model container of the health assessment model; Based on the model container of the health assessment model, generate an image file of the model container of the health assessment model; The image file of the model container of the health assessment model is sent to the device to be deployed to implement the deployment of the health assessment model.

9. A container-based power system model reasoning and analysis device, characterized in that: include: A communication module, used to obtain real-time operation data in the power system, wherein the real-time operation data includes image data and parameter data; A processing module, used for extracting features from the real-time operation data based on the meta-information module to obtain standard features; Input the standard features into the containers of each power system model to obtain the output results of each container; the power system model includes a transformer diagnosis model, a circuit breaker fault diagnosis model and a load prediction model; based on the output results of each container and a preset health assessment model, perform an operation status analysis to determine the health score of each device in the power system, and the health score is used to assess the failure probability of the device; Based on the output results of each container and the health score of each device in the power system, a reasoning analysis report is generated.

10. The containerized power system model reasoning and analysis device according to claim 9, characterized in that: The processing module is specifically used to extract key features in the image data using the service script module of the meta information module, wherein the key features include surface temperature distribution features and current heat distribution features; Utilizing the service script module of the meta-information module, extracting key features in the parameter data, the key features including current, current offset, current change rate, voltage, voltage offset, voltage change rate, load state and load fluctuation range; performing feature conversion on the key features of the image data and parameter data based on the device type of each device in the power system, the operating environment of the container and the model type of the power system model to obtain conversion features; Based on the conversion feature and the data format of each container service interface, format conversion is performed to obtain the standard feature.

Images