Method, system, device and medium for analyzing operating status of electric power equipment

Abstract

The present invention relates to the technical field of power equipment operating status analysis, and discloses a power equipment operating status analysis method, system, device, and medium. The method comprises: obtaining a first parameter of a target power equipment, performing a first classification on the first parameter, performing a first preprocessing on the first parameter after the first classification to obtain a second parameter, establishing a first analysis model, training the first analysis model according to the second parameter to obtain a trained first analysis model, recorded as a second analysis model, and inputting the second analysis model into the second analysis model in response to real-time data from the target power equipment to obtain an analysis result of the power equipment operating status. The method fully considers the complexity and diversity of the power equipment operating status, and achieves accurate analysis and real-time monitoring of the power equipment operating status through sophisticated parameter classification and preprocessing, as well as an advanced four-layer network architecture analysis model.

Description

Technical Field

[0001] The present invention relates to the technical field of power equipment operating status analysis, and in particular to a power equipment operating status analysis method, system, equipment and medium. Background Art

[0002] As modern power systems continue to grow in size and complexity, the demand for accurate analysis and real-time monitoring of the operating status of power equipment is growing. Traditional power equipment monitoring methods rely primarily on scheduled maintenance and manual inspections, which are not only inefficient but also difficult to detect potential problems in a timely manner, potentially leading to equipment failures or power outages. Furthermore, the operating status of power equipment is affected by a variety of factors, including but not limited to internal parameters (such as temperature and voltage) and external environmental parameters (such as weather conditions and geographical factors). The complex interactions between these parameters make accurate assessment of power equipment status even more difficult.

[0003] In recent years, with the development of artificial intelligence and machine learning technologies, using data-driven methods to analyze the operating status of power equipment has become a research hotspot. By collecting and analyzing large amounts of historical data, effective models can be built to predict potential problems with power equipment, thereby enabling preventive maintenance. However, existing models often fail to fully consider the differences between dynamic and static parameters and their varying impacts on the operating status of power equipment, resulting in inaccurate predictions. Summary of the Invention

[0004] In view of the above existing problems, the present invention is proposed.

[0005] Therefore, the present invention provides a method, system, device and medium for analyzing the operating status of electric equipment, which can solve the problem of inaccurate analysis of the operating status of electric equipment in the prior art, especially the problem of failing to fully consider the difference between dynamic parameters and static parameters.

[0006] In order to solve the above technical problems, the present invention provides the following technical solutions:

[0007] In a first aspect, the present invention provides a method for analyzing the operating status of an electric power device, comprising:

[0008] Acquiring a first parameter of a target power device and performing a first classification on the first parameter;

[0009] The first classification is used to divide the first parameter into a first dynamic parameter and a first static parameter;

[0010] performing a first preprocessing on the first parameter after the first classification to obtain a second parameter, where the second parameter includes a second dynamic parameter and a second static parameter;

[0011] Establishing a first analysis model, the first analysis model including two input terminals, a four-layer network architecture, and an output terminal;

[0012] The four-layer network architecture includes a first feature judgment network architecture and a second feature judgment network architecture connected in parallel, a third state analysis network structure connected in series with the first feature judgment network architecture and the second feature judgment network architecture, and a fourth decision network structure connected in series with the third state analysis network structure;

[0013] Training the first analysis model according to the second parameter to obtain a trained first analysis model, which is recorded as a second analysis model;

[0014] In response to the real-time data of the target power equipment being input into the second analysis model, an analysis result of the power equipment operating status is obtained.

[0015] As a preferred solution of the method for analyzing the operating status of electric power equipment according to the present invention, the first classification of the first parameter includes:

[0016] The first parameter includes a first internal parameter and a second external parameter;

[0017] The first category includes:

[0018] Preset the first parameter classification library and the second parameter classification judgment logic;

[0019] Acquire classified parameters and unclassified parameters in the first internal parameter and the second external parameter according to the first parameter classification library;

[0020] Performing classification judgment on the unclassified parameters according to the second parameter classification judgment logic to obtain a classification judgment result;

[0021] A first dynamic parameter and a first static parameter are obtained according to the classified parameters and the classification judgment result.

[0022] As a preferred solution of the method for analyzing the operating status of electric power equipment according to the present invention, the first classification further includes:

[0023] The initial first parameter classification library is a blank library;

[0024] The initial first classification is performed only through the second parameter classification judgment logic, and the initial first classification is performed based on the second parameter classification judgment logic on the initial first parameter;

[0025] According to the parameters in the initial classification judgment result, a non-blank first parameter classification library is established.

[0026] As a preferred solution of the method for analyzing the operating status of electric power equipment according to the present invention, the first analysis model further includes:

[0027] The two input ends are input ends of the first feature judgment network architecture and the second feature judgment network architecture respectively;

[0028] One of the first feature judgment network architecture and the second feature judgment network architecture is used to process the second dynamic data in the second parameter, and the other is used to process the second static data in the second parameter;

[0029] The output of the first feature judgment network architecture is a first state analysis result;

[0030] The output of the second feature judgment network architecture is a second state analysis result;

[0031] The first state analysis result and the second state analysis result are inputs of the third state analysis network structure.

[0032] As a preferred solution of the method for analyzing the operating status of electric power equipment according to the present invention, the first analysis model further includes:

[0033] The third state analysis network structure is used to perform a third state analysis on the first state analysis result and the second state analysis result;

[0034] Presetting a third state analysis logic, wherein the third state analysis network structure is established according to the third state analysis logic;

[0035] The result of the third state analysis is the final analysis state of the target power equipment;

[0036] The final analysis state is input as input to the fourth decision network structure.

[0037] As a preferred solution of the method for analyzing the operating status of electric power equipment according to the present invention, the first analysis model further includes:

[0038] The fourth decision network structure is used to decide a state processing strategy according to the final analysis state;

[0039] The state processing strategy includes several processing strategies for different target power equipment operating states.

[0040] As a preferred solution of the method for analyzing the operating status of electric power equipment described in the present invention, the four-layer network architecture includes different network objective functions and network constraints.

[0041] In a second aspect, the present invention provides a system for analyzing the operating status of power equipment, comprising:

[0042] a data acquisition and classification module, configured to acquire a first parameter of a target power device and perform a first classification on the first parameter;

[0043] The first classification is used to divide the first parameter into a first dynamic parameter and a first static parameter;

[0044] a preprocessing module, configured to perform a first preprocessing on the first parameter after the first classification to obtain a second parameter, where the second parameter includes a second dynamic parameter and a second static parameter;

[0045] A model building module, configured to build a first analysis model, wherein the first analysis model includes two input terminals, a four-layer network architecture, and an output terminal;

[0046] The four-layer network architecture includes a first feature judgment network architecture and a second feature judgment network architecture connected in parallel, a third state analysis network structure connected in series with the first feature judgment network architecture and the second feature judgment network architecture, and a fourth decision network structure connected in series with the third state analysis network structure;

[0047] A model training module, configured to train the first analysis model according to the second parameter to obtain a trained first analysis model, which is recorded as a second analysis model;

[0048] The analysis module is used to obtain an analysis result of the operating status of the power equipment in response to real-time data of the target power equipment input into the second analysis model.

[0049] In a third aspect, the present invention provides a computer device comprising a memory and a processor, wherein the memory stores a computer program, and the processor implements the steps of the above-described method when executing the computer program.

[0050] In a fourth aspect, the present invention provides a computer-readable storage medium having a computer program stored thereon, which implements the steps of the method described above when the computer program is executed by a processor.

[0051] Compared with the prior art, the present invention has the following advantages: The present invention provides a method, system, device, and medium for analyzing the operating status of power equipment. The method comprises: obtaining a first parameter of a target power equipment and performing a first classification on the first parameter; performing a first preprocessing on the first parameter after the first classification to obtain a second parameter, which includes a second dynamic parameter and a second static parameter; establishing a first analysis model comprising two input terminals, a four-layer network architecture, and an output terminal; training the first analysis model based on the second parameter to obtain a trained first analysis model, referred to as a second analysis model; and inputting real-time data from the target power equipment into the second analysis model to obtain an analysis result of the power equipment's operating status. This method fully considers the complexity and diversity of power equipment operating status. Through sophisticated parameter classification and preprocessing, and an advanced four-layer network architecture analysis model, it achieves accurate analysis and real-time monitoring of the power equipment's operating status. Compared with traditional power equipment monitoring methods, the method of the present invention not only improves analysis efficiency but also enables timely detection of potential problems, avoiding economic losses and social impacts caused by equipment failures or power outages. Furthermore, the method dynamically adjusts the analysis model based on historical and real-time data from the power equipment, improving the accuracy and reliability of the analysis. Therefore, the present invention has broad application prospects and important practical significance in the technical field of power equipment operating status analysis.

[0052] Specifically, first, the first parameters of the target power equipment are obtained and classified, which helps to distinguish parameters of different properties and provides a basis for subsequent processing. Next, the classified parameters are preprocessed to further purify and normalize the data and improve data quality. Then, a first analysis model consisting of two input ports, a four-layer network architecture and an output port is established, which can fully utilize the dynamic and static characteristics of the data and improve the accuracy of the analysis. By training the first analysis model according to the second parameters to obtain the trained second analysis model, it can be ensured that the model can adapt to changes in the actual operating status of the power equipment. Finally, in response to the real-time data of the target power equipment, the second analysis model is input to obtain the analysis results of the operating status of the power equipment, which can realize real-time monitoring and accurate evaluation of the operating status of the power equipment. Each step provides key support for the final accurate analysis and real-time monitoring, ensuring the effectiveness and reliability of the entire method. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. 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 any creative labor.

[0054] Figure 1 A flowchart of a method for analyzing the operating status of electric equipment provided by one embodiment of the present invention.

[0055] Figure 2 A first classification logic diagram of a method for analyzing the operating status of electric equipment provided by an embodiment of the present invention.

[0056] Figure 3 A fourth decision network structure diagram of a method for analyzing the operating status of electric power equipment provided by an embodiment of the present invention.

[0057] Figure 4 This is a structural diagram of a first analysis model of a method for analyzing the operating status of electric equipment provided by one embodiment of the present invention.

[0058] Figure 5 This is a diagram of the internal structure of a computer device for a method for analyzing the operating status of electric power equipment provided by one embodiment of the present invention. DETAILED DESCRIPTION

[0059] To make the above-mentioned objects, features, and advantages of the present invention more clearly understood, the following detailed description of the specific embodiments of the present invention is given in conjunction with the accompanying drawings. It is obvious that the described embodiments are only part of the embodiments of the present invention, not all of them. Based on the embodiments of the present invention, all other embodiments obtained by ordinary persons in this field without creative work should fall within the scope of protection of the present invention.

[0060] Example 1, reference Figure 1-Figure 5 , which is the first embodiment of the present invention, provides a method for analyzing the operating status of an electric power device, comprising:

[0061] Before introducing the embodiments of the present application in detail, some related concepts are first explained for the sake of clarity.

[0062] A convolutional neural network (CNN) is a deep learning model that is particularly well-suited for processing data with a grid structure, such as images. It automatically extracts feature representations from input data through a series of layers:

[0063] Convolutional layer: Uses filters to scan the input data to identify patterns or features in local areas.

[0064] Activation function: Usually nonlinear functions such as ReLU are used to increase the expressive power of the model.

[0065] Pooling layer: reduces the spatial size of the data, reduces computational complexity, and improves invariance to input transformations.

[0066] Fully connected layer: Converts the feature map obtained previously into a specific output, such as a classification result.

[0067] CNN can effectively capture the spatial hierarchical structure of input data by sharing weights and local receptive fields.

[0068] Reinforcement Learning (RL) is a branch of machine learning that focuses on learning optimal behavior strategies through trial and error based on how behaviors interact with the environment. Its core concepts include:

[0069] Agent: A learner that performs actions and interacts with the environment.

[0070] Environment: The world in which the subject lives, which gives feedback based on the subject's actions.

[0071] Reward: The numerical feedback provided by the environment to the subject to guide the subject to learn the optimal strategy.

[0072] Policy: The method by which an agent chooses an action based on its current state.

[0073] Value Function: Evaluates the quality of taking a specific strategy in a certain state.

[0074] The goal of reinforcement learning is to find a strategy that maximizes the cumulative reward of the agent in the long run.

[0075] Network Architecture refers to the specific structural arrangements for building artificial neural networks or other types of computational models. This includes but is not limited to:

[0076] The number of layers and the type of each layer: such as input layer, hidden layer (which can be convolutional layer, recurrent layer, etc.), and output layer.

[0077] The number of neurons in each layer determines the width of the layer and affects the capacity and complexity of the model.

[0078] The connection method between layers: for example, full connection, convolutional connection, recursive connection, etc.

[0079] Choice of activation function: determines how each neuron converts its input into output.

[0080] Design of forward propagation and backpropagation algorithms: used to calculate prediction errors and adjust network weights.

[0081] Existing technologies present several challenges. For example, traditional power equipment monitoring methods often rely on manual inspections and empirical judgment, which is not only inefficient but also difficult to accurately detect potential equipment failures. Furthermore, while some existing automated monitoring methods can collect equipment data in real time, they lack sufficient processing and analysis, resulting in inaccurate analysis results and a lack of effective support for equipment maintenance and management.

[0082] This application provides a method that can effectively solve the above-mentioned problems. Next, we will describe in detail how to implement the power equipment operating status analysis method in combination with multiple embodiments.

[0083] Figure 1 A method flow chart of a method for analyzing the operating status of an electric power device is shown, comprising:

[0084] S101, obtaining a first parameter of a target power device, and performing a first classification on the first parameter;

[0085] It should be noted that the target power equipment in this application may include transformers, generators, circuit breakers, and the like. During operation, these devices generate a large amount of data, including parameters such as current, voltage, and temperature. This data is an important basis for analyzing the operating status of these devices. The purpose of the first classification is to categorize these parameters according to their properties and characteristics, providing a foundation for subsequent processing and analysis.

[0086] In the embodiment of the present application, the first classification is used to divide the first parameter into a first dynamic parameter and a first static parameter.

[0087] It should be noted that the complexity of the current power system is often difficult to fully reflect through a single dynamic or static parameter. Therefore, this application subdivides the first parameter into a first dynamic parameter and a first static parameter through a first classification to more accurately capture the operating status of the power equipment.

[0088] It should also be noted that traditionally, dynamic parameters generally refer to parameters that change over time, such as instantaneous current and voltage values, which can reflect the real-time operating status of power equipment. Static parameters, on the other hand, refer to relatively stable parameters, such as device temperature and resistance, which more closely reflect the long-term operating status and performance of power equipment. However, this traditional classification is not suitable for most power equipment. Therefore, this application proposes a method that can provide detailed and accurate classification.

[0089] In the embodiment of the present application, performing a first classification on the first parameter includes:

[0090] The first parameter includes a first internal parameter and a second external parameter;

[0091] Specifically, when the power equipment is a transformer, the first internal parameter includes at least one of the following:

[0092] Oil temperature, winding temperature, oil level, winding current, winding voltage, power, frequency deviation, rated voltage, rated power and rated capacity, etc.;

[0093] The second external parameter includes at least one of the following: ambient temperature, humidity, load change, operating environment, altitude, and noise level.

[0094] When the electric power equipment is a generator, the first internal parameter includes at least one of the following: stator current, stator voltage, rotor current, rotor voltage, power factor, frequency, active power, reactive power, temperature, and vibration;

[0095] The second external parameter includes at least one of the following: ambient temperature, humidity, cooling medium flow rate, cooling medium temperature, altitude, load change, etc.

[0096] When the electrical equipment is a circuit breaker, the first internal parameter includes at least one of the following: number of operations, opening and closing time, contact wear, insulation resistance, leakage current, and gas pressure;

[0097] The second external parameter includes at least one of the following: ambient temperature, humidity, installation location, operating frequency, and external electric field or magnetic field interference.

[0098] It should be noted that different first internal parameters and second external parameters can be set for different types of power equipment. The selection and setting of specific parameters depends on the type, structure, and operating characteristics of the power equipment. By classifying the first parameters in detail, a deeper understanding of the operating status of the power equipment can be obtained, providing more accurate and comprehensive data support for subsequent analysis and processing. This application does not limit this.

[0099] It is important to note that some static parameters in actual operation need to be calculated based on real-time data or directly obtained through sensors. These are considered preliminary operation steps in this application. For example, noise level can be collected and calculated in real time using sensors such as microphones, ambient temperature and humidity can be obtained using temperature and humidity sensors, and load changes can be monitored and calculated using current and voltage sensors. Real-time acquisition and monitoring of these static parameters helps to more comprehensively reflect the operating status of power equipment and improve the accuracy of analysis.

[0100] It should be noted that the benefit of dividing the data into first internal parameters and second external parameters is that it provides a more detailed reflection of the operating status of power equipment, taking into account various influencing factors both internal and external to the equipment. First internal parameters primarily reflect the operating status of the power equipment itself, such as oil temperature and winding temperature, which are crucial for determining the health of the equipment. Second external parameters, on the other hand, take into account the environmental conditions of the equipment, such as ambient temperature and humidity. While these parameters do not directly reflect the internal status of the equipment, they can also affect its operation. By comprehensively considering these two types of parameters, the operating status of power equipment can be more comprehensively assessed, improving the accuracy of the analysis.

[0101] It should also be noted that, in order to avoid being too long, the subsequent content of this application only describes the "transformer" as an exemplary electronic device.

[0102] like Figure 2 As shown, in the embodiment of the present application, the first category includes:

[0103] Preset the first parameter classification library and the second parameter classification judgment logic;

[0104] Acquire classified parameters and unclassified parameters in the first internal parameter and the second external parameter according to the first parameter classification library;

[0105] Performing classification judgment on the unclassified parameters according to the second parameter classification judgment logic to obtain a classification judgment result;

[0106] A first dynamic parameter and a first static parameter are obtained according to the classified parameters and the classification judgment result.

[0107] For example, when the power equipment is a transformer, and the dynamic parameters in the first classification library are "power, frequency offset", and the static parameters are "rated voltage, rated power, and rated capacity", then the classified parameters in the first internal parameters and the second external parameters are several types of "power, frequency offset, rated voltage, rated power, and rated capacity"; the unclassified parameters are the parameters in the first internal parameters and the second external parameters excluding "power, frequency offset, rated voltage, rated power, and rated capacity";

[0108] Furthermore, the unclassified parameters are judged by the second parameter classification judgment logic (i.e., whether there is a judgment logic for the unclassified parameters), the parameters that meet the judgment logic are classified, and the parameters that do not meet the judgment logic are judged by importance;

[0109] If the importance is insufficient, the unsatisfied parameter will not be considered;

[0110] If the importance is sufficient, then regenerate the second parameter classification judgment logic in combination with the characteristics of the unsatisfied parameter;

[0111] For example, if the unsatisfied parameter is the oil temperature trend, this is highly significant because it reflects the transformer's heat dissipation and internal temperature changes, making it crucial for determining whether the transformer is overheating. Therefore, based on the characteristics of the oil temperature trend, the second parameter classification and judgment logic can be regenerated, classifying it as a dynamic parameter. Separate criteria for determining the oil temperature trend are generated and incorporated into the second parameter classification and judgment logic to complete the update.

[0112] It should be noted that the importance judgment standard can be obtained by conducting experiments on historical data to obtain thresholds and judged by thresholds, or it can be judged by establishing an importance judgment library based on expert experience. There are many other ways to make judgments, which are not limited in this application.

[0113] In this embodiment of the present application, the first category also includes:

[0114] The initial first parameter classification library is a blank library;

[0115] The initial first classification is performed only through the second parameter classification judgment logic, and the initial first classification is performed based on the second parameter classification judgment logic on the initial first parameter;

[0116] According to the parameters in the initial classification judgment result, a non-blank first parameter classification library is established.

[0117] Exemplarily, assuming an initial state, there is a target power device (such as a transformer) in the example, whose first parameter set P={p1, p2, ..., pn} includes n different parameters (such as voltage, current, temperature, etc.).

[0118] Initially, the dynamic parameter set D={} in the first parameter classification library is an empty set, and the static parameter set S={} is an empty set.

[0119] The second parameter classification judgment logic evaluates and classifies each parameter pi in P according to a preset time variation pattern or other characteristics.

[0120] Assume that after the initial classification, the results are as follows:

[0121] The dynamic parameter set D'={p1, p3, p5} satisfies the dynamic parameter judgment in the second parameter classification judgment logic.

[0122] The static parameter set S'={p2, p4, p6} satisfies the static parameter judgment in the second parameter classification judgment logic.

[0123] Based on the results of the above initial classification, a non-empty first parameter classification library can be established in the example:

[0124] The updated dynamic parameter set D=D'.

[0125] The updated static parameter set S=S'.

[0126] Therefore, after the initial classification is complete, the first parameter classification library is no longer blank, but instead contains a specific set of dynamic parameters D and static parameters S. This classification library can be further updated and refined based on subsequent data and analysis to accommodate different types or models of power equipment. For power equipment of the same type, the generated classification library can be directly applied; however, for equipment of different models, the corresponding classification library must be regenerated.

[0127] In an optional embodiment, the second parameter classification judgment logic includes:

[0128] Presetting a first period threshold and a second fluctuation threshold;

[0129] If there is a parameter in the unclassified parameters whose fluctuation (i.e., data change) within the first period threshold is greater than the second fluctuation threshold, then the parameter is recorded as a dynamic sampling parameter, otherwise it is a static parameter;

[0130] For example, assume that a transformer's oil temperature is being evaluated as an unclassified parameter, with the first period threshold set to one day (24 hours) and the second fluctuation threshold set to 5 degrees Celsius. If the oil temperature fluctuates by more than 5 degrees Celsius over a 24-hour period, the above logic classifies the oil temperature as a dynamic parameter. If the fluctuation threshold is not exceeded, the oil temperature is considered a static parameter.

[0131] It should be noted that the first cycle threshold is a value that can be adjusted according to the needs of relevant technical personnel. The second fluctuation threshold is initially an empirical value, which is obtained based on previous historical data. After the initial judgment, the second fluctuation threshold is updated according to the cycle in the set first cycle threshold, that is, after entering the next cycle, the value of the second fluctuation threshold is obtained again based on the data obtained in this cycle and the previous historical data.

[0132] It should also be noted that the second fluctuation threshold includes several different branch thresholds. The number of branch thresholds is determined by the number of dynamic parameters or static parameters obtained from historical data. For example, if there are 5 dynamic parameters and 3 static parameters obtained from historical data, the number of branch thresholds is 5 + 3 = 8. These branch thresholds correspond to different types of parameters, allowing for more precise control of the classification process. In practical applications, when the fluctuation of an unclassified parameter exceeds its corresponding branch threshold, the parameter is classified as a dynamic parameter; otherwise, it is considered a static parameter.

[0133] In this way, the accuracy and flexibility of parameter classification can be further improved, thereby better supporting the analysis and monitoring of the operating status of power equipment and avoiding the one-sidedness caused by the traditional selection of dynamic data or static data.

[0134] The dynamic parameters in the above-mentioned classified parameters and the combination of the dynamic parameters in the classification judgment result are recorded as the first dynamic parameters, and the static parameters in the above-mentioned classified parameters and the combination of the static parameters in the classification judgment result are recorded as the second static parameters.

[0135] It should be noted that as power equipment is continuously updated and upgraded, the data types and characteristics it generates will also change, and traditional classification methods may not be able to adapt to these changes. The first parameter classification library and the second parameter classification judgment logic of this application can be updated. The advantage of this is that it can adapt to different types of power equipment and different operating environments, improving the accuracy and flexibility of classification.

[0136] It should also be noted that obtaining the first parameter of the target power equipment and performing a first classification on the first parameter can provide more accurate and comprehensive data support for subsequent analysis and processing. By performing a detailed and accurate classification of the first parameter, the present application can gain a deeper understanding of the operating status of the power equipment and capture the dynamic and static characteristics of the equipment, thereby providing a scientific basis for equipment health monitoring, fault prediction, and maintenance decisions. In addition, this classification method also helps to improve the efficiency of data analysis, reduce unnecessary data redundancy, and make the analysis results more intuitive and easy to understand. In the next step, the present application will further construct an analysis model for the operating status of power equipment based on these classified parameters to achieve real-time monitoring and accurate analysis of the operating status of power equipment.

[0137] S102, performing a first preprocessing on the first parameter after the first classification to obtain a second parameter, where the second parameter includes a second dynamic parameter and a second static parameter;

[0138] In an optional embodiment, the first preprocessing may include, but is not limited to, data cleaning, data normalization, and data smoothing. Data cleaning aims to remove outliers, missing values, or duplicate data to ensure the accuracy and reliability of the data used for analysis. Data normalization scales the data to a small, specific interval, such as between 0 and 1, so that parameters of different dimensions or magnitudes can be compared and analyzed under the same standard. Data smoothing is used to reduce noise in the data, improve the smoothness and continuity of the data, and thus more realistically reflect the operating status of the power equipment.

[0139] It should be noted that directly using traditional preprocessing steps cannot meet the accuracy and timeliness requirements for analyzing the operating status of power equipment. Traditional preprocessing steps often only process a single type of data, failing to consider the complexity and diversity of power equipment operating status data. Therefore, in the embodiments of this application, a first preprocessing step is proposed that customizes the first parameter after the first classification.

[0140] In the embodiment of the present application, different preprocessing methods and strategies are selected for the first dynamic parameter and the first static parameter according to their characteristics and analysis requirements. For example, for dynamic parameters that change rapidly over time, a more sophisticated time series analysis method may be required for preprocessing. For relatively stable static parameters, more emphasis may be placed on the integrity and accuracy of the data, and the preprocessing process may be relatively simple. Through such a preprocessing step, a higher quality data foundation can be provided for subsequent analysis and modeling.

[0141] Specifically, the preprocessing of dynamic parameters includes:

[0142] For the first dynamic parameters that change rapidly with time (such as power, frequency offset, etc.), their main characteristic is that they have significant time dependence. Therefore, special attention should be paid to the time series characteristics of the data during the preprocessing process.

[0143] Outlier detection and processing: Use statistical methods or machine learning algorithms to identify and process outliers, such as eliminating extreme values ​​based on the 3σ principle.

[0144] Missing value filling: Use interpolation methods (linear interpolation, spline interpolation) or model-based prediction methods to fill missing data points.

[0145] Smoothing: Apply moving average, exponential smoothing and other techniques to reduce noise interference and highlight trend characteristics.

[0146] Decomposition analysis: Use seasonal decomposition (such as STL decomposition) to separate long-term trends, cyclical components and random fluctuations.

[0147] Feature extraction: Extract useful features from time series, such as maximum, minimum, mean, variance, slope change rate, etc.; autocorrelation coefficients and cross-correlation coefficients can also be calculated to capture the relationship between parameters.

[0148] Standardization / normalization: Convert data of different dimensions to the same scale to facilitate subsequent modeling.

[0149] Update the classification library: Based on the latest data analysis results, adjust the classification logic of dynamic parameters and update them to the parameter classification library.

[0150] The preprocessing of static parameters includes:

[0151] For relatively stable static parameters (such as rated voltage, rated power, etc.), the focus is on ensuring the integrity and accuracy of the data, and the preprocessing process may be relatively simple.

[0152] Consistency check: Ensure the consistency of all records to avoid duplicate entries or incorrect input.

[0153] Verify data integrity: Check whether there is any missing key information and complete the data through additional investigation or other means if necessary.

[0154] Unified format: Ensure that all data is stored in a consistent format to facilitate subsequent processing.

[0155] Unit conversion: If there are data in different units, they need to be converted into standard units.

[0156] Accuracy verification: Verify the accuracy of data by comparing it with data from other reliable sources.

[0157] Reliability assessment: Evaluate the reliability of the data source to determine whether additional data support or correction is needed.

[0158] Archiving management: Properly preserve the cleaned and converted static parameters, and establish a reliable database or archiving system to facilitate long-term management and query.

[0159] For example, when analyzing the operating status of a power transformer, the dynamic parameter is the oil temperature:

[0160] For oil temperature, the 3σ principle (data points outside the mean ±3 standard deviations are considered outliers) is used to identify and remove outliers. Missing oil temperature data points are filled using linear interpolation. A moving average filter (e.g., with a window size of 5 hours) is applied to reduce the impact of short-term fluctuations and better demonstrate long-term trends. The STL decomposition method is used to separate the oil temperature data into trend, seasonal, and residual components to facilitate subsequent feature extraction. The hourly oil temperature change rate (i.e., the current oil temperature minus the previous hour's oil temperature) and the daily difference between the maximum and minimum temperatures are calculated. All oil temperature data is scaled to the [0, 1] interval to ensure that data of different dimensions can be effectively processed by the model. Based on the latest data analysis results, the classification logic of the dynamic parameters is adjusted and updated in the parameter classification library. For example, if the oil temperature change rate is found to exhibit significant fluctuations within a specific time period, it is redefined as a key dynamic feature and added to the classification library. (This step facilitates the update of the second parameter classification judgment logic described above. The parameters used in the second parameter classification judgment logic can directly access the data in this parameter classification library.)

[0161] It should be noted that performing a first preprocessing on the first parameter after the first classification to obtain the second parameter can significantly improve the accuracy and efficiency of subsequent analysis and modeling. The second parameter after preprocessing, whether it is the second dynamic parameter or the second static parameter, is closer to the actual operating state of the power equipment, reducing the interference of noise and outliers. Such a data foundation provides a strong guarantee for building an accurate analysis model. In addition, the preprocessing step also takes into account the different characteristics of dynamic parameters and static parameters, and adopts a targeted processing strategy, which makes the results of data preprocessing more in line with the actual needs of power equipment operating status analysis. Therefore, in the subsequent steps, whether it is feature extraction, model training or state prediction, you can rely more on the preprocessed data, thereby improving the reliability and practicality of the entire analysis process.

[0162] S103, establishing a first analysis model, wherein the first analysis model includes two input terminals, a four-layer network architecture, and one output terminal;

[0163] In an embodiment of the present application, the four-layer network architecture includes a first feature judgment network architecture and a second feature judgment network architecture in parallel, a third state analysis network structure connected in series with the first feature judgment network architecture and the second feature judgment network architecture, and a fourth decision network structure connected in series with the third state analysis network structure, such as Figure 4 shown.

[0164] In this embodiment of the present application, the first analysis model further includes:

[0165] The two input ends are input ends of the first feature judgment network architecture and the second feature judgment network architecture respectively;

[0166] One of the first feature judgment network architecture and the second feature judgment network architecture is used to process the second dynamic data in the second parameter, and the other is used to process the second static data in the second parameter;

[0167] The output of the first feature judgment network architecture is a first state analysis result;

[0168] The output of the second feature judgment network architecture is a second state analysis result;

[0169] The first state analysis result and the second state analysis result are inputs of the third state analysis network structure.

[0170] In this embodiment of the present application, the first analysis model further includes:

[0171] The third state analysis network structure is used to perform a third state analysis on the first state analysis result and the second state analysis result;

[0172] Presetting a third state analysis logic, wherein the third state analysis network structure is established according to the third state analysis logic;

[0173] The result of the third state analysis is the final analysis state of the target power equipment;

[0174] The final analysis state is input as input to the fourth decision network structure.

[0175] In this embodiment of the present application, the first analysis model further includes:

[0176] The fourth decision network structure is used to decide a state processing strategy according to the final analysis state;

[0177] The state processing strategy includes several processing strategies for different target power equipment operating states.

[0178] In the embodiment of the present application, the four-layer network architecture includes different network objective functions and network constraints.

[0179] In an optional embodiment, the first analysis model can employ a deep learning network architecture, such as a convolutional neural network (CNN) or a recurrent neural network (RNN) and its variants, such as a long short-term memory (LSTM) or a gated recurrent unit (GRU), to accommodate power equipment operating status analysis requirements of varying complexity. CNNs excel at processing spatially structured data, such as images, extracting local features and performing efficient classification. RNNs and their variants excel at processing time series data, capturing the temporal trends and cyclical characteristics of parameters, which is particularly important for dynamic parameter analysis. Selecting an appropriate deep learning network architecture based on the characteristics of the target power equipment and the analysis requirements can better mine the potential information in the data and improve the accuracy and efficiency of the analysis.

[0180] In an optional embodiment, the first analysis model may also utilize ensemble learning methods, such as random forests, gradient boosted tree (GBDT), or XGBoost. By combining the prediction results of multiple weak learners, ensemble learning methods can significantly improve the model's generalization and robustness. These methods excel at processing high-dimensional data, nonlinear relationships, and complex feature interactions, and are suitable for comprehensive multi-parameter and multi-feature assessments in power equipment operating status analysis. By training multiple base learners and integrating their prediction results, ensemble learning models can more comprehensively capture the operating status characteristics of power equipment, improving the accuracy and reliability of the analysis.

[0181] It should be noted that the model established by the above method cannot be directly associated with subsequent processing operations after analyzing the power equipment status, which limits the application effect of the model. To solve this problem, this application proposes an innovative strategy, namely, pre-embedding the decision logic associated with subsequent processing operations when designing the first analysis model.

[0182] It should also be noted that this application uses a convolutional neural network design to establish the first analysis model, combined with reinforcement learning for improvement. This design enables the model to have stronger feature extraction and status analysis capabilities when processing power equipment operating status data. Convolutional neural networks (CNNs) have achieved remarkable results in processing image data due to their powerful feature learning capabilities. Applying them to the analysis of power equipment operating status data can automatically extract key features from the data, avoiding the tedious process of manually designing features in traditional methods. At the same time, combined with improvements in reinforcement learning (RL), the model can optimize its decision-making ability during the continuous learning process, thereby more accurately formulating corresponding processing strategies based on the operating status of the power equipment.

[0183] Specifically, the specific network objective functions and network constraints of the four networks in this application can be as follows (the input of the first feature judgment network architecture is selected as the second dynamic parameter):

[0184] The first feature judgment network architecture is improved and designed through a convolutional neural network. Its goal is to judge the state of the target power equipment through dynamic parameters and obtain the first state analysis result;

[0185] The second feature judgment network architecture is improved and designed through convolutional neural networks. Its goal is to judge the state of the target power equipment through static parameters and obtain the second state analysis result;

[0186] The third state analysis network structure is improved and designed using a convolutional neural network. Its goal is to analyze the operating state of the final target device based on the first and second state analysis results, obtain the final operating state type, and provide an accurate input for the subsequent fourth decision network structure;

[0187] The fourth decision network structure is improved and designed through reinforcement learning, and its goal is to determine which strategy to implement based on the final operating state type.

[0188] For example, the specific network objective function and network constraints used in this application may be as follows:

[0189] The network objective function of the first feature judgment network (dynamic parameter processing) is as follows. The goal of this network is to extract the dynamic characteristics of the equipment operation status through the second dynamic parameters (such as voltage, current, temperature and other parameters that change with time and have undergone the first preprocessing) and output the first state analysis results.

[0190] Objective function (loss function):

[0191] ,

[0192] Where: LCE represents the cross entropy loss function, which measures the predicted state and the real state difference. is a regularization term (such as L2 regularization) to prevent overfitting. is the regularization coefficient. is the network weight parameter. N1 represents the number of training samples in the second dynamic parameter.

[0193] The network constraints are as follows:

[0194] Input constraint: Input is a dynamic parameter tensor , where T is the time step and D1 is the dynamic parameter feature dimension (such as the number of parameters such as voltage and current in this application).

[0195] Output constraint: Output is state probability distribution , where C is the number of status categories (such as normal, overload, fault, etc. selected in this application).

[0196] Activation function constraints: The convolutional layer uses the ReLU activation function, and the fully connected layer uses the Softmax activation function.

[0197] Gradient constraint: Use gradient clipping (e.g., in this application, the gradient norm is limited to ) to prevent gradient explosion.

[0198] It should be noted that the first feature judgment network is used to process dynamic parameters (such as time series data) and adopts a 1D convolutional neural network (CNN) structure combined with time series feature extraction.

[0199] The objective function of the second feature judgment network (static parameter processing) is as follows. The goal of this network is to extract the static characteristics of the equipment through static parameters (such as rated power, ambient temperature, and other parameters that do not change rapidly over time) and output the second state analysis results.

[0200] Objective function (loss function):

[0201] ,

[0202] Among them, LCE is the cross entropy loss function, which measures the predicted state and the real state difference. is a regularization term (such as L1 regularization) to prevent overfitting. is the regularization coefficient. is the network weight parameter, and N2 represents the number of training samples in the second static parameter.

[0203] The network constraints are as follows:

[0204] Input constraints: Input is a static parameter vector , where D2 is the static parameter feature dimension (such as the number of parameters such as rated voltage and ambient humidity).

[0205] Output constraint: Output is state probability distribution , where C is the number of status categories (such as normal, overload, fault, etc. selected in this application).

[0206] Activation function constraints: The fully connected layer uses the ReLU activation function, and the output layer uses Softmax.

[0207] Data normalization constraint: The input data must be standardized (for example, the mean is 0 and the variance is 1).

[0208] It should be noted that the second feature judgment network processes static parameters (such as scalar features) and adopts a fully connected neural network (FCNN) without convolutional layers.

[0209] The network objective function of the third state analysis network is as follows: the goal of this network is to fuse the analysis results of dynamic and static features and output the final device state.

[0210] Objective function (loss function):

[0211] ,

[0212] in, is the true final state label, To predict the final state probability distribution, is a regularization term (such as L2 regularization) to prevent overfitting. is the regularization coefficient. is the network weight parameter, and N3 represents the number of training samples in the output first state analysis results and the second state analysis results.

[0213] The network constraints are as follows:

[0214] Input constraint: The input is the concatenation of the outputs of the first two networks .

[0215] Fusion constraints: Use weighted average or fully connected layers to fuse dynamic and static features:

[0216] ,

[0217] in, are learnable parameters.

[0218] Output constraint: Output final state .

[0219] It should be noted that the third state analysis network fuses the outputs of the first two networks and must include a fully connected layer and an attention mechanism.

[0220] The network objective function in the fourth decision network structure is to determine which strategy to implement based on the final operating state type, which can be converted into the agent's maximum expected discounted cumulative reward, as follows:

[0221] The goal of the agent is to maximize the expected discounted cumulative reward:

[0222] ,

[0223] in, Represents the trajectory (the trajectory is determined by the strategy generate). represents the discount factor (e.g. ), controls the weight of the long-term reward. T represents the termination time of the trajectory (such as equipment failure or maintenance completion).

[0224] Among them, the immediate reward r at each time step t is t It is composed of the following items

[0225] ,

[0226] in, Indicates the reward for completing the task. For the penalty cost, For exploration rewards, To restrain punishment.

[0227] In the reinforcement learning structure of this application, the environment is the power equipment operation system, which provides states and rewards. The intelligent agent is a policy network based on PPO / SAC (PPO (Proximal Policy Optimization): suitable for discrete action space, stable training; SAC (Soft Actor-Critic): suitable for continuous action space, automatic balance between exploration and utilization), which selects maintenance actions. The state is a vector that integrates dynamic and static features and contains historical dependencies. The action is a discrete or continuous maintenance strategy (such as shutdown, alarm). The reward is to quantify the effect of the action and drive the intelligent agent to learn the optimal strategy. Specifically, Figure 3 As shown in the figure, the environment: provides the current state to the agent, and gives rewards based on the results of the action after the agent takes action. The agent: receives state information from the environment, decides what action to take through its internal algorithm (such as PPO or SAC), and feeds this action back to the environment. State: data obtained from the environment, containing information about the operating conditions of power equipment, which is input into the agent to help decision-making. Action: an operation selected by the agent based on the current state (such as shutdown, alarm, etc.), which acts on the environment. Reward: The environment gives corresponding rewards or penalties based on the actions taken by the agent to guide the agent to learn the optimal strategy.

[0228] The data flow is as follows: the environment provides the state to the agent (state flow), the agent decides and performs actions based on the received state (action flow), and after the action is performed, the environment gives the agent a reward based on the effect of the action (reward flow). The agent uses the reward to update its strategy model so that it can make better decisions in the future.

[0229] Specifically, in this application Used to reward the agent for completing key tasks (such as avoiding failures).

[0230] .

[0231] In this application Used to penalize unnecessary maintenance actions (such as frequent downtime leading to increased costs).

[0232] .

[0233] In this application Used to encourage the agent to explore new actions (such as trying strategies that have not been executed before).

[0234] ,

[0235] in, represents the exploration coefficient (such as 0.1), N a Indicates the historical execution times of action a.

[0236] In this application Used to ensure that actions comply with physical constraints (such as the load cannot exceed the rated power).

[0237] ,

[0238] in, is the constraint penalty coefficient (taken as 50 in this application).

[0239] It should be noted that the reward function in this application can be further updated to account for the effects of sparsity, providing high rewards for critical events (failures, maintenance), thus avoiding the learning difficulties caused by sparse rewards. It is also possible to increase the weight of exploration rewards and intensive rewards in the early stages (exploration phase) to encourage the agent to try a variety of actions, and then strengthen task rewards in the later stages (convergence phase) and sparsify non-critical rewards.

[0240] It should also be noted that the fourth decision network structure in this application maximizes the discounted cumulative reward , the intelligent agent will learn to adopt the optimal maintenance strategy (such as shutdown, alarm) under the equipment status (such as overload, fault precursor), while avoiding the cost loss caused by excessive maintenance.

[0241] It should be noted that establishing the first analysis model fully leverages the advantages of deep learning technology to conduct a comprehensive and accurate analysis of the operating status of power equipment. By selecting appropriate deep learning network architectures, such as RNNs and their variants, convolutional neural networks, and improved designs incorporating reinforcement learning, not only is the accuracy and efficiency of the analysis improved, but the model's generalization and robustness are also enhanced. Furthermore, by pre-embedding decision logic associated with subsequent processing operations during model design, the model can directly guide practical applications, avoiding the problem of limited model application effectiveness. This innovative power equipment operating status analysis system will provide strong support for the safe and stable operation of the power system.

[0242] S104, training the first analysis model according to the second parameter to obtain a trained first analysis model, which is recorded as a second analysis model;

[0243] It should be noted that training the established first analysis model based on the second parameter can further optimize the model's performance, making it more suitable for actual power equipment operating status analysis needs. The second parameter is divided into a training set, a validation set, and a test set. The training set is used for model training, the validation set is used to adjust hyperparameters and prevent overfitting, and the test set is used to evaluate the final model performance.

[0244] In this embodiment of the present application, the first analysis model is trained as a whole based on the objective function and constraints established above, combined with the second parameter. Once the model training is complete (referred to as the second analysis model), it is evaluated using an independent test set to check its generalization ability on unseen data. Once the second analysis model passes the test, it can be deployed in practical applications for real-time analysis of the operating status of target power equipment. This process not only improves the model's accuracy and robustness but also ensures its adaptability to different power equipment and operating environments, thereby providing strong support for the safe and stable operation of the power system.

[0245] S105 , in response to the real-time data of the target power equipment being input into the second analysis model, obtaining an analysis result of the power equipment operating status.

[0246] It should be noted that real-time data collection is required from the target power equipment. This data includes, but is not limited to, dynamic parameters such as voltage, current, power, and temperature, as well as static parameters such as rated power and rated voltage. The corresponding data undergoes a first classification and preprocessing phase, and is then input into a second analysis model. Based on this data, the second analysis model conducts a comprehensive analysis of the target power equipment's operating status through its internal first feature judgment network architecture, second feature judgment network architecture, third state analysis network architecture, and fourth decision network architecture. Specifically, the first feature judgment network architecture processes dynamic parameters to extract dynamic features of the equipment's operating status; the second feature judgment network architecture processes static parameters to extract static features; the third state analysis network architecture integrates the analysis results of dynamic and static features to output the final equipment status; and the fourth decision network architecture determines the appropriate strategy based on the final operating status. This process occurs in real time, ensuring rapid and accurate reflection of the target power equipment's actual operating status, providing strong support for the safe and stable operation of the power system.

[0247] In summary, the present invention proposes a method for analyzing the operating status of power equipment. The method comprises obtaining first parameters of a target power equipment and performing a first classification on the first parameters; performing a first preprocessing on the first parameters after the first classification to obtain second parameters, which include second dynamic parameters and second static parameters; establishing a first analysis model comprising two input terminals, a four-layer network architecture, and an output terminal; training the first analysis model based on the second parameters to obtain a trained first analysis model, referred to as a second analysis model; and inputting real-time data from the target power equipment into the second analysis model to obtain an analysis result of the power equipment's operating status. This method fully considers the complexity and diversity of power equipment operating status. Through sophisticated parameter classification and preprocessing, and an advanced four-layer network architecture analysis model, it achieves accurate analysis and real-time monitoring of the power equipment's operating status. Compared with traditional power equipment monitoring methods, the method of the present invention not only improves analysis efficiency but also enables timely detection of potential problems, avoiding economic losses and social impacts caused by equipment failures or power outages. Furthermore, the method dynamically adjusts the analysis model based on historical and real-time data from the power equipment, improving the accuracy and reliability of the analysis. Therefore, the present invention has broad application prospects and important practical significance in the technical field of power equipment operating status analysis.

[0248] Embodiment 2: This embodiment further provides a system for analyzing the operating status of electric power equipment, including:

[0249] a data acquisition and classification module, configured to acquire a first parameter of a target power device and perform a first classification on the first parameter;

[0250] The first classification is used to divide the first parameter into a first dynamic parameter and a first static parameter;

[0251] a preprocessing module, configured to perform a first preprocessing on the first parameter after the first classification to obtain a second parameter, where the second parameter includes a second dynamic parameter and a second static parameter;

[0252] A model building module, configured to build a first analysis model, wherein the first analysis model includes two input terminals, a four-layer network architecture, and an output terminal;

[0253] The four-layer network architecture includes a first feature judgment network architecture and a second feature judgment network architecture connected in parallel, a third state analysis network structure connected in series with the first feature judgment network architecture and the second feature judgment network architecture, and a fourth decision network structure connected in series with the third state analysis network structure;

[0254] A model training module, configured to train the first analysis model according to the second parameter to obtain a trained first analysis model, which is recorded as a second analysis model;

[0255] The analysis module is used to obtain an analysis result of the operating status of the power equipment in response to real-time data of the target power equipment input into the second analysis model.

[0256] The above-mentioned unit modules can be embedded in or independent of the processor in the computer device in the form of hardware, or can be stored in the memory of the computer device in the form of software, so that the processor can call and execute the corresponding operations of the above-mentioned modules.

[0257] This embodiment also provides a computer device, which may be a terminal, and its internal structure diagram may be as shown in FIG. Figure 5 As shown. The computer device includes a processor, a memory, a communication interface, a display screen and an input device connected via a system bus. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The communication interface of the computer device is used to communicate with an external terminal in a wired or wireless manner, and the wireless manner can be achieved through WIFI, an operator network, NFC (near field communication) or other technologies. When the computer program is executed by the processor, a method for analyzing the operating status of an electric power device is implemented. The display screen of the computer device can be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer device can be a touch layer covering the display screen, or a button, trackball or touchpad provided on the computer device housing, or an external keyboard, touchpad or mouse.

[0258] This embodiment further provides a computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, the following steps are implemented:

[0259] Acquiring a first parameter of a target power device and performing a first classification on the first parameter;

[0260] The first classification is used to divide the first parameter into a first dynamic parameter and a first static parameter;

[0261] performing a first preprocessing on the first parameter after the first classification to obtain a second parameter, where the second parameter includes a second dynamic parameter and a second static parameter;

[0262] Establishing a first analysis model, the first analysis model including two input terminals, a four-layer network architecture, and an output terminal;

[0263] The four-layer network architecture includes a first feature judgment network architecture and a second feature judgment network architecture connected in parallel, a third state analysis network structure connected in series with the first feature judgment network architecture and the second feature judgment network architecture, and a fourth decision network structure connected in series with the third state analysis network structure;

[0264] Training the first analysis model according to the second parameter to obtain a trained first analysis model, which is recorded as a second analysis model;

[0265] In response to the real-time data of the target power equipment being input into the second analysis model, an analysis result of the power equipment operating status is obtained.

[0266] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

[0267] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may take the form of a fully hardware embodiment, a fully software embodiment, or an embodiment combining software and hardware. Furthermore, the present application may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code. The solutions in the embodiments of the present application may be implemented using various computer languages.

[0268] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 a process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0269] These computer program instructions may also be stored in a computer readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 a process or multiple processes and / or boxes Figure 1 The function specified in one or more boxes.

[0270] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operational steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing the instructions executed on the computer or other programmable device for implementing the process. Figure 1 a process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0271] Although the preferred embodiments of the present application have been described, those skilled in the art may make additional changes and modifications to these embodiments once they have learned the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications that fall within the scope of the present application.

[0272] Obviously, those skilled in the art may make various changes and modifications to this application without departing from the spirit and scope of this application. Thus, if these modifications and variations of this application fall within the scope of the claims of this application and their equivalents, this application is intended to include these modifications and variations.

Claims

1. A method for analyzing the operating status of power equipment, characterized in that: include: Acquiring a first parameter of a target power device and performing a first classification on the first parameter; The first classification of the first parameter includes: The first parameter includes a first internal parameter and a second external parameter; The first category includes: Preset the first parameter classification library and the second parameter classification judgment logic; Acquire classified parameters and unclassified parameters in the first internal parameter and the second external parameter according to the first parameter classification library; Performing classification judgment on the unclassified parameters according to the second parameter classification judgment logic, and obtaining a classification judgment result if judgment logic exists for the unclassified parameters; If it does not exist, make a materiality judgment; Preset importance judgment criteria, if the importance judgment criteria are met, then regenerate the second parameter classification judgment logic; If the importance judgment criteria are not met, the corresponding unclassified parameters are not considered; The second parameter classification judgment logic includes: Presetting a first period threshold and a second fluctuation threshold; If there is a parameter among the unclassified parameters whose data change within the first period threshold is greater than the second fluctuation threshold, then the parameter is a dynamic parameter, otherwise it is a static parameter; The first cycle threshold is a value adjusted according to the needs of relevant technical personnel. The second fluctuation threshold is initially an empirical value. After the initial judgment, the second fluctuation threshold is updated according to the period set in the first cycle threshold; The second fluctuation threshold includes several different branch thresholds, and the number of branch thresholds is determined according to the number of dynamic parameters or static parameters obtained from historical data; Obtaining a first dynamic parameter and a first static parameter according to the classified parameters and the classification judgment result; The first category also includes: The initial first parameter classification library is a blank library; The initial first classification is performed only through the second parameter classification judgment logic, and the initial first classification is performed based on the second parameter classification judgment logic on the initial first parameter; Establishing a non-blank first parameter classification library based on the parameters in the initial classification judgment result; The first classification is used to divide the first parameter into a first dynamic parameter and a first static parameter; performing a first preprocessing on the first parameter after the first classification to obtain a second parameter, where the second parameter includes a second dynamic parameter and a second static parameter; Establishing a first analysis model, the first analysis model including two input terminals, a four-layer network architecture, and an output terminal; The four-layer network architecture includes a first feature judgment network architecture and a second feature judgment network architecture connected in parallel, a third state analysis network structure connected in series with the first feature judgment network architecture and the second feature judgment network architecture, and a fourth decision network structure connected in series with the third state analysis network structure; The four-layer network architecture includes different network objective functions and network constraints. The different network objective functions and network constraints of the four-layer network architecture are as follows: The first feature judgment network is used to process dynamic parameters. The objective function of the first feature judgment network is to extract the dynamic characteristics of the equipment operation status through the second dynamic parameters and output the first state analysis results. It uses a 1D convolutional neural network structure and combines time series to extract dynamic features. The first feature judgment network constraint conditions include an input constraint based on the first feature judgment network, an output constraint based on the first feature judgment network, an activation function constraint based on the first feature judgment network, and a gradient constraint based on the first feature judgment network; The second feature judgment network is used to process static parameters. The goal of the second feature judgment network objective function is to extract the static characteristics of the device through the static parameters and output the second state analysis result. A fully connected neural network is used, and the neural network does not contain a convolutional layer. The second feature judgment network constraint conditions include an input constraint based on the second feature judgment network, an output constraint based on the second feature judgment network, an activation function constraint based on the second feature judgment network, and a data normalization constraint based on the second feature judgment network; The goal of the network objective function of the third state analysis network is to fuse the analysis results of dynamic and static features and output the final device state; The constraints of the third state analysis network include input constraints based on the outputs of the first feature judgment network and the second feature judgment network, and fusion constraints refer to output constraints for the third state analysis network. The third state analysis network must include a fully connected layer and an attention mechanism. The network objective function in the fourth decision network structure is to determine which strategy to implement based on the final operating state type; The fourth decision network structure includes a system in which the power equipment resides as an environment, an output of the third state analysis network as a state, and an intelligent agent configured to receive the output of the third state analysis network and execute actions, wherein the executed actions include several maintenance strategies; the environment generates corresponding rewards based on the executed actions and feeds the rewards back to the intelligent agent, wherein the rewards include a task completion reward, a penalty cost, an exploration reward, and a constraint penalty; Training the first analysis model according to the second parameter to obtain a trained first analysis model, which is recorded as a second analysis model; In response to the real-time data of the target power equipment being input into the second analysis model, an analysis result of the power equipment operating status is obtained.

2. A method for analyzing the operating status of electric power equipment according to claim 1, characterized in that: The first analysis model further includes: The two input ends are input ends of the first feature judgment network architecture and the second feature judgment network architecture respectively; One of the first feature judgment network architecture and the second feature judgment network architecture is used to process the second dynamic data in the second parameter, and the other is used to process the second static data in the second parameter; The output of the first feature judgment network architecture is a first state analysis result; The output of the second feature judgment network architecture is a second state analysis result; The first state analysis result and the second state analysis result are inputs of the third state analysis network structure.

3. A method for analyzing the operating status of electric power equipment according to claim 2, characterized in that: The first analysis model further includes: The third state analysis network structure is used to perform a third state analysis on the first state analysis result and the second state analysis result; Presetting a third state analysis logic, wherein the third state analysis network structure is established according to the third state analysis logic; The result of the third state analysis is the final analysis state of the target power equipment; The final analysis state is input as input to the fourth decision network structure.

4. A method for analyzing the operating status of electric power equipment according to claim 3, characterized in that: The first analysis model further includes: The fourth decision network structure is used to decide a state processing strategy according to the final analysis state; The state processing strategy includes several processing strategies for different target power equipment operating states.

5. A power equipment operating status analysis system using the method according to claim 1, characterized in that: include: a data acquisition and classification module, configured to acquire a first parameter of a target power device and perform a first classification on the first parameter; The first classification of the first parameter includes: The first parameter includes a first internal parameter and a second external parameter; The first category includes: Preset the first parameter classification library and the second parameter classification judgment logic; Acquire classified parameters and unclassified parameters in the first internal parameter and the second external parameter according to the first parameter classification library; Performing classification judgment on the unclassified parameters according to the second parameter classification judgment logic, and obtaining a classification judgment result if judgment logic exists for the unclassified parameters; If it does not exist, make a materiality judgment; Preset importance judgment criteria, if the importance judgment criteria are met, then regenerate the second parameter classification judgment logic; If the importance judgment criteria are not met, the corresponding unclassified parameters are not considered; The second parameter classification judgment logic includes: Presetting a first period threshold and a second fluctuation threshold; If there is a parameter among the unclassified parameters whose data change within the first period threshold is greater than the second fluctuation threshold, then the parameter is a dynamic parameter, otherwise it is a static parameter; The first cycle threshold is a value adjusted according to the needs of relevant technical personnel. The second fluctuation threshold is initially an empirical value. After the initial judgment, the second fluctuation threshold is updated according to the period set in the first cycle threshold; The second fluctuation threshold includes several different branch thresholds, and the number of branch thresholds is determined according to the number of dynamic parameters or static parameters obtained from historical data; Obtaining a first dynamic parameter and a first static parameter according to the classified parameters and the classification judgment result; The first category also includes: The initial first parameter classification library is a blank library; The initial first classification is performed only through the second parameter classification judgment logic, and the initial first classification is performed based on the second parameter classification judgment logic on the initial first parameter; Establishing a non-blank first parameter classification library based on the parameters in the initial classification judgment result; The first classification is used to divide the first parameter into a first dynamic parameter and a first static parameter; a preprocessing module, configured to perform a first preprocessing on the first parameter after the first classification to obtain a second parameter, where the second parameter includes a second dynamic parameter and a second static parameter; A model building module, configured to build a first analysis model, wherein the first analysis model includes two input terminals, a four-layer network architecture, and an output terminal; The four-layer network architecture includes a first feature judgment network architecture and a second feature judgment network architecture connected in parallel, a third state analysis network structure connected in series with the first feature judgment network architecture and the second feature judgment network architecture, and a fourth decision network structure connected in series with the third state analysis network structure; The four-layer network architecture includes different network objective functions and network constraints. The different network objective functions and network constraints of the four-layer network architecture are as follows: The first feature judgment network is used to process dynamic parameters. The objective function of the first feature judgment network is to extract the dynamic characteristics of the equipment operation status through the second dynamic parameters and output the first state analysis results. It uses a 1D convolutional neural network structure and combines time series to extract dynamic features. The first feature judgment network constraint conditions include an input constraint based on the first feature judgment network, an output constraint based on the first feature judgment network, an activation function constraint based on the first feature judgment network, and a gradient constraint based on the first feature judgment network; The second feature judgment network is used to process static parameters. The goal of the second feature judgment network objective function is to extract the static characteristics of the device through the static parameters and output the second state analysis result. A fully connected neural network is used, and the neural network does not contain a convolutional layer. The second feature judgment network constraint conditions include an input constraint based on the second feature judgment network, an output constraint based on the second feature judgment network, an activation function constraint based on the second feature judgment network, and a data normalization constraint based on the second feature judgment network; The goal of the network objective function of the third state analysis network is to fuse the analysis results of dynamic and static features and output the final device state; The constraints of the third state analysis network include input constraints based on the outputs of the first feature judgment network and the second feature judgment network, and fusion constraints refer to output constraints for the third state analysis network. The third state analysis network must include a fully connected layer and an attention mechanism. The network objective function in the fourth decision network structure is to determine which strategy to implement based on the final operating state type; The fourth decision network structure includes a system in which the power equipment resides as an environment, an output of the third state analysis network as a state, and an intelligent agent configured to receive the output of the third state analysis network and execute actions, wherein the executed actions include several maintenance strategies; the environment generates corresponding rewards based on the executed actions and feeds the rewards back to the intelligent agent, wherein the rewards include a task completion reward, a penalty cost, an exploration reward, and a constraint penalty; A model training module, configured to train the first analysis model according to the second parameter to obtain a trained first analysis model, which is recorded as a second analysis model; The analysis module is used to obtain an analysis result of the operating status of the power equipment in response to real-time data of the target power equipment input into the second analysis model.

6. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method for analyzing the operating status of electric power equipment according to any one of claims 1 to 4 are implemented.

7. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of a method for analyzing the operating status of electric power equipment according to any one of claims 1 to 4 are implemented.

Images