A remote maintenance system for an intelligent power distribution cabinet

Abstract

The present invention relates to the technical field of remote maintenance of distribution cabinets, especially a remote maintenance system for intelligent distribution cabinets. By collecting the current, voltage, component temperature, start-up time, and internal and external environment data of the distribution cabinet in real time, a component operation parameter matrix is constructed, and combined with the distribution cabinet environment parameters, a residual neural network algorithm is used for deep learning training. This method can not only detect common fault modes, but also identify abnormal situations caused by component aging, thus solving the problems of false alarms and missed alarms existing in traditional fault detection methods. Through the residual neural network, the system can automatically learn normal and abnormal operation modes from historical data, improving the accuracy of fault detection. Further, the abnormal types are pushed to the operation and maintenance personnel and visually displayed, saving the judgment time and round-trip times of the operation and maintenance personnel for maintaining components and improving the maintenance efficiency.

Description

Technical Field

[0001] The present invention relates to the technical field of remote maintenance of power distribution cabinets, and particularly to a remote maintenance system for intelligent power distribution cabinets. Background Art

[0002] As a crucial device in the power system, the power distribution cabinet undertakes multiple functions such as power distribution, control, and protection. With the continuous increase in power demand, traditional power distribution cabinet equipment is increasingly difficult to meet the requirements of efficient, stable, and safe operation. Traditional power distribution cabinets usually rely on manual inspections, regular maintenance, and manual recording of equipment operation conditions, which not only have low efficiency but also are prone to problems such as missed inspections and misjudgments, thus leading to equipment failures or power interruption accidents and affecting the security and reliability of power supply.

[0003] In the existing fault detection methods in the monitoring system of power distribution cabinets, simple threshold judgments are relied on, and advanced artificial intelligence technologies are not fully utilized for in-depth analysis of equipment status. Especially in a large power distribution cabinet environment, simple rules cannot effectively cope with various potential abnormal patterns, often resulting in a high false alarm rate or missed alarm phenomena. And when the staff of the existing power distribution cabinet monitoring system has not reached the abnormal power distribution cabinet, they cannot intuitively discover the problems existing in the power distribution cabinet and determine the component models that need to be replaced or repaired, thus increasing the number of round trips and reducing the maintenance efficiency. Summary of the Invention

[0004] To solve the above problems, the present invention provides a remote maintenance system for intelligent power distribution cabinets. By performing residual neural network learning and determination based on operation parameters and power distribution cabinet environment parameters, and then re-updating the component management module through maintenance records to adjust the operation parameters, the accuracy of abnormal pattern determination is effectively improved, and the maintenance efficiency is improved by visualizing the abnormal patterns.

[0005] To achieve the above object, the technical solution adopted by the present invention is:

[0006] A remote maintenance system for intelligent power distribution cabinets includes a component management module, an abnormal monitoring module, a model construction module, and a remote maintenance module connected in sequence, and also includes an in-cabinet environment monitoring module. The in-cabinet environment monitoring module is connected to the abnormal monitoring module, and the remote maintenance module is connected to the component management module;

[0007] The component management module is used to update and record the component attribute parameters and component operation parameters in the power distribution cabinet in real time. The component operation parameters include current and voltage, component temperature, and component activation time;

[0008] The in-cabinet environment monitoring module is used to obtain the power distribution cabinet environment parameters in real time, including the in-cabinet environment humidity and the out-of-cabinet environment temperature and humidity;

[0009] The abnormal monitoring module is used to construct an operation parameter matrix on a time series based on the component operation parameters in the power distribution cabinet, train a residual neural network based on the operation parameter matrix and the power distribution cabinet environment parameters, and determine whether there is an abnormality in the real-time component operation parameters through the trained residual neural network to generate a real-time determination result;

[0010] The model construction module is used to construct a three-dimensional framework of the power distribution cabinet, and combine the three-dimensional framework of the power distribution cabinet by matching a preset component library according to the component category, component model and component connection relationship to obtain a real-time three-dimensional model of the power distribution cabinet; the model construction module is also used to perform synchronous mapping on the three-dimensional model of the power distribution cabinet based on the power distribution cabinet environment parameters and the real-time determination result;

[0011] The remote maintenance module is used to receive the three-dimensional model of the power distribution cabinet from the model construction module and visualize it, and is also used to receive maintenance record information to update the component management module.

[0012] Further, the component attribute parameters include component category, component model and component connection relationship.

[0013] Further, the operation parameter matrix includes a current-voltage matrix, a component temperature matrix and a component activation time matrix.

[0014] Further, the construction of the operation parameter matrix on a time series based on the component operation parameters in the power distribution cabinet includes the following steps:

[0015] Obtain component operation parameters from the component management module;

[0016] Based on the component connection relationship, preset value filling areas for the current-voltage matrix, the component temperature matrix and the component activation time matrix;

[0017] Remove outliers and noise from the component operation parameters and perform standardization processing to obtain standardized values;

[0018] Correspondingly fill the standardized values in the value filling areas of the current-voltage matrix, the component temperature matrix and the component activation time matrix.

[0019] Further, the training of the residual neural network based on the operation parameter matrix and the power distribution cabinet environment parameters includes the following steps:

[0020] Obtain the operation parameter matrix and the power distribution cabinet environment parameters to construct a training data set;

[0021] Construct a residual neural network model, use the training data set as input, and extract features from the training data through a convolutional layer to obtain a multi-dimensional feature map;

[0022] Input the multi-dimensional feature map into the residual module and extract high-dimensional features through the residual connection mechanism;

[0023] Input the extracted high-dimensional features into the fully connected layer and output the judgment result through the activation function.

[0024] Further, the judgment results include normal power distribution cabinet, abnormal power distribution cabinet, and abnormal component.

[0025] Further, the calculation formula of the residual neural network model is as follows:

[0026] ;

[0027] where is the predicted judgment result of the power distribution cabinet at time t; is the input feature at time t, including the operation parameter matrix and the power distribution cabinet environment parameter; is the convolution kernel parameter of the th layer; is the bias term; is the convolution kernel weight matrix of the output layer; is the bias term of the output layer; is the activation function; is the output of the previous moment of the th layer in the residual neural network; is the number of network layers.

[0028] Further, the loss function of the residual neural network model is as follows:

[0029] ;

[0030] where is the total loss function; is the true judgment result of the power distribution cabinet at time t; is the predicted judgment result of the power distribution cabinet at time t; is the mean square error between the predicted value and the true value; is the convolution kernel parameter of the th layer; is the regularization coefficient; is the number of samples in the training data set; is the number of network layers.

[0031] Further, the remote maintenance module includes the cloud and the maintenance personnel terminal;

[0032] The remote maintenance module is used to receive a number of three-dimensional models of power distribution cabinets and send the three-dimensional models of power distribution cabinets to the maintenance personnel terminal according to the abnormal judgment results, and is also used to receive the maintenance record information uploaded by the personnel terminal and synchronously update it to the component management module;

[0033] The maintenance staff terminal is used to receive the 3D model of the power distribution cabinet and perform AR visualization, and is also used to receive maintenance record information and upload it to the remote maintenance module in a standardized manner.

[0034] The beneficial effects of the present invention are as follows: By collecting the current and voltage, component temperature, component startup time, and internal and external environment data of the power distribution cabinet in real time, compared with traditional manual inspections and regular maintenance, the system can monitor the working status of the power distribution cabinet at any time, avoiding missed inspections and misjudgments in manual inspections, and ensuring the timeliness and accuracy of equipment operation. By introducing the residual neural network algorithm, through constructing a component operation parameter matrix and combining it with the power distribution cabinet environment parameters for in-depth learning training, not only daily faults are considered, but also the aging problems caused by long-term use of components are considered, effectively solving the problems of low accuracy, serious false alarms and missed alarms in traditional fault detection methods based on threshold judgment. The residual neural network can automatically learn the normal and abnormal operation modes of the equipment from a large amount of historical data. When encountering complex faults or abnormal situations, it can accurately identify and give early warnings in a timely manner, reducing the false alarm rate and missed alarm rate, obtaining the data of the maintenance records and updating them back to the component management module, realizing the recalculation of the component usage time, and further improving the accuracy and synchronization of model judgment. Through the intelligent analysis of the abnormal monitoring module, this solution can not only detect potential faults in the power distribution cabinet, but also push warning information to the operation and maintenance personnel according to the fault type and perform visual display, saving the judgment time and round-trip times of the operation and maintenance personnel for maintenance components, and improving the maintenance efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] Figure 1 is a schematic structural diagram of a remote maintenance system for an intelligent power distribution cabinet in the present invention.

[0036] Figure 2 is a flowchart of the steps for training a residual neural network based on the operation parameter matrix and the power distribution cabinet environment parameters in the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0037] Please refer to Figure 1-2 As shown, the present invention relates to a remote maintenance system for an intelligent power distribution cabinet, including a component management module, an abnormal monitoring module, a model construction module, and a remote maintenance module connected in sequence, and further including a cabinet internal environment monitoring module, the cabinet internal environment monitoring module is connected to the abnormal monitoring module, and the remote maintenance module is connected to the component management module;

[0038] The component management module is used to update and record the component attribute parameters and component operation parameters in the power distribution cabinet in real time, and the component operation parameters include current and voltage, component temperature, and component startup time;

[0039] The cabinet internal environment monitoring module is used to obtain the power distribution cabinet environment parameters in real time, including the humidity inside the cabinet and the temperature and humidity outside the cabinet;

[0040] The anomaly monitoring module is used to construct an operation parameter matrix in time series based on the component operation parameters in the power distribution cabinet, train a residual neural network based on the operation parameter matrix and the power distribution cabinet environment parameters, and use the trained residual neural network to determine anomalies in the real-time component operation parameters to generate a real-time determination result;

[0041] The model construction module is used to construct a three-dimensional framework of the power distribution cabinet, and combine the three-dimensional framework of the power distribution cabinet by matching a preset component library according to the component category, component model and component connection relationship to obtain a real-time three-dimensional model of the power distribution cabinet; the model construction module is also used to perform synchronous mapping on the three-dimensional model of the power distribution cabinet based on the power distribution cabinet environment parameters and the real-time determination result;

[0042] The remote maintenance module is used to receive the three-dimensional model of the power distribution cabinet from the model construction module and visualize it, and is also used to receive maintenance record information to update the component management module.

[0043] In some embodiments, the core hardware components of the system include a component management module, an anomaly monitoring module, a model construction module, a remote maintenance module, and a cabinet internal environment monitoring module. First, the component management module is responsible for collecting and recording the operation parameters inside the power distribution cabinet in real time. The components can include circuit breakers, switches, contactors, disconnectors, transformers, relays, and cables. This module is equipped with high-precision current and voltage sensors and infrared temperature sensors, and can monitor the operation status of each component in the power distribution cabinet in real time. The data collected by the sensors include the current, voltage, and temperature parameters of each electrical component. The time when the monitoring component is enabled is inserted through the interface, and these data are transmitted to the central processing unit of the system in real time through a wired or wireless network. The central processing unit updates the attribute parameters and operation status of each component in the power distribution cabinet according to these data in real time.

[0044] Furthermore, the component attribute parameters include component category, component model, and component connection relationship.

[0045] Specifically, the component category refers to the type of component in the distribution panel, such as circuit breakers, switches, contactors, disconnectors, transformers, protective devices, and relays. This parameter helps the system identify the function and operating principle of each component, providing essential information for subsequent fault diagnosis and maintenance. The component model refers to the specific model or specification of each component, such as the rated current and rated voltage of a particular circuit breaker. This parameter is important for identifying the component's technical standards, performance range, and compatibility. By recording this model information, the system can perform targeted analysis based on the model when a fault occurs, determining whether a specific component type needs replacement or repair. Component connection relationships: This parameter describes the physical and logical connections between components in the distribution panel. This parameter describes the electrical connections and dependencies between different components. For example, a switch may be directly connected to a circuit breaker, while a transformer may be interconnected with multiple switches. By recording these connection relationships, the system can better understand the internal circuit structure of the distribution panel, providing a clear electrical diagram for troubleshooting and system optimization. Furthermore, this parameter helps the system quickly locate other potentially affected components when a fault occurs, enabling coordinated detection and remote maintenance.

[0046] Furthermore, the operating parameter matrix includes a current and voltage matrix, a component temperature matrix, and a component activation time matrix.

[0047] Specifically, the current and voltage matrix records the current and voltage data for each component in the power distribution cabinet. Each matrix element represents the current and voltage value of a component at a specific point in time (or time period). Using this matrix, the system can monitor the operating status of electrical components in real time and determine whether any abnormal conditions, such as overload or short circuit, exist. The matrix not only helps the system identify potential faults but also analyzes historical data to predict future equipment performance trends. The component temperature matrix records the temperature data of each component in the power distribution cabinet. Component temperature is a key factor affecting its proper operation; excessively high temperatures can cause equipment damage or failure. Each element in the temperature matrix corresponds to the temperature value of a component at a specific moment. This data allows the system to analyze whether the equipment is overheating and provides a basis for subsequent fault warnings and maintenance. The accumulation of long-term temperature data also helps the system identify signs of equipment aging or potential safety hazards. The component operating time matrix records the operating time or operating hours of each component in the power distribution cabinet. The operating time of each component is often used to measure its service life. By monitoring the operating time matrix, the system can assess the degree of component aging or when it is nearing end of life. This matrix provides an important time reference for fault prediction and equipment maintenance. Especially when performing periodic maintenance, enabling the time matrix can help the system decide which components need to be inspected or replaced.

[0048] Further, the construction of the operation parameter matrix on the time series based on the component operation parameters in the power distribution cabinet includes the following steps:

[0049] Obtain the component operation parameters from the component management module;

[0050] Preset the value filling areas for the current-voltage matrix, the component temperature matrix, and the component enabled time matrix based on the component connection relationship;

[0051] Remove outliers and noise from the component operation parameters and perform standardization processing to obtain standardized values;

[0052] Correspondingly fill the standardized values in the value filling areas of the current-voltage matrix, the component temperature matrix, and the component enabled time matrix respectively.

[0053] In some embodiments, first, the system obtains the operating parameters of each component in the power distribution cabinet from the component management module. These parameters include current, voltage, component temperature, and component activation time, etc. Data acquisition is carried out through various sensors and acquisition modules, ensuring the real-time and accuracy of the data. The acquired raw data usually presents in the form of a time series and may contain some outliers and noise, which requires subsequent processing steps to clean and standardize the data. Next, the system presets the filling areas for different operating parameter matrices according to the component connection relationship. Specifically, at the beginning of constructing the current-voltage matrix, component temperature matrix, and component activation time matrix, corresponding filling areas need to be predefined so that the parameters of each component can be accurately mapped to the corresponding matrix positions. In this process, the component connection relationship provides a reference for setting the structure of each parameter matrix. For example, through the electrical wiring diagram, it can be determined which current-voltage data between components should be mapped to the same matrix area. The purpose of this is to ensure the logical consistency of the data and facilitate subsequent matrix processing and analysis. Before data filling, the system removes outliers and noise from all the obtained component operating parameters. To this end, the system adopts outlier detection algorithms such as the Z-Score method or the interquartile range method to detect and eliminate abnormal data beyond the normal range. In addition, in order to eliminate the differences between data with different dimensions and improve the accuracy of data analysis, all component operating parameters will be standardized after removing outliers. The purpose of standardization is to convert data with different dimensions into a unified standard format, so that each parameter has equal weight in subsequent analysis, avoiding disproportionate influence on the analysis results due to some parameters being larger or smaller in value. After standardization processing, the system fills the obtained standardized values into the corresponding filling areas of the current-voltage matrix, component temperature matrix, and component activation time matrix. This process is achieved through efficient matrix operations, ensuring that the real-time data of each component is accurately inserted into the corresponding matrix according to the time series and forming a complete time series dataset with the operating data of other components. The construction of these matrices provides data support for subsequent anomaly detection and fault diagnosis based on deep learning models.

[0054] Further, the training of the residual neural network based on the operating parameter matrix and the power distribution cabinet environmental parameters includes the following steps:

[0055] Obtain the operating parameter matrix and the power distribution cabinet environmental parameters to construct a training dataset;

[0056] Construct a residual neural network model, use the training dataset as input, and extract features from the training data through the convolutional layer to obtain a multi-dimensional feature map;

[0057] Input the multi-dimensional feature map into the residual module and extract high-dimensional features through the residual connection mechanism;

[0058] Input the extracted high-dimensional features into the fully connected layer and output the determination result through the activation function.

[0059] In some embodiments, first, the system obtains the real-time operation data of each component in the power distribution cabinet from the component management module, including operation parameters such as current, voltage, component temperature, and activation time, and forms an operation parameter matrix with these data. At the same time, environmental data inside and outside the power distribution cabinet, such as temperature and humidity, external environmental temperature and humidity, etc., are collected as environmental parameters. After preliminary cleaning of these data, an original data set for neural network training is formed. During the data set construction process, the system will uniformly format and preprocess the data to ensure the dimensional consistency of each parameter and the accuracy of the data. Then, based on the above training data set, the system constructs a residual neural network model. In the neural network model, first, the input data is subjected to feature extraction through convolutional layers. The convolutional layers use different convolutional kernels to perform multi-dimensional sliding window processing on the data, and can extract local features in the data, such as current and voltage fluctuation features and the change trend of environmental temperature, etc. Through multiple convolutional operations, the network can gradually learn high-level abstract features, thereby obtaining the dynamic feature maps of the power distribution cabinet components. Then, these multi-dimensional feature maps are input into the residual module. The core idea of the residual module is to establish a direct path between different layers of the network through the residual connection mechanism to avoid the problem of gradient disappearance during the training of deep networks. The residual connection enables information to be passed layer by layer in the network, ensuring the retention of key features and the effective extraction of deep features. The residual module can effectively capture complex time series features, such as the gradual changes caused by component aging or the instantaneous fluctuations brought about by sudden failures. Next, the extracted high-dimensional features will be passed to the fully connected layers. In the fully connected layers, the neural network compresses and integrates the multi-dimensional feature vectors, and further abstracts the global features of the operation state of the power distribution cabinet. At this time, the features are non-linearly mapped through activation functions (such as ReLU or Sigmoid) to enhance the expression ability and prediction ability of the network. The activation function helps to introduce non-linear features, ensuring that the model can learn the complex rules in the data. Especially in complex power distribution cabinet fault modes, it can improve the accuracy and generalization ability of the system. Finally, the activated output is optimized through a loss function (such as cross entropy or mean square error) to adjust the weights of the neural network. Through the backpropagation algorithm, the network continuously updates the weights and biases until the output of the model can accurately reflect the normal operation state and potential faults of the power distribution cabinet. After this process is completed, the residual neural network model can efficiently detect various abnormal situations in the power distribution cabinet, thereby realizing precise fault warning and remote maintenance of the power distribution cabinet.

[0060] Further, the determination results include that the power distribution cabinet is normal, the power distribution cabinet is abnormal, and the component is abnormal.

[0061] Specifically, when the power distribution cabinet is normal: when the operating parameters of all components are within the predetermined safety threshold range and the environmental parameters of the power distribution cabinet are also in a normal state, the residual neural network will output a judgment result of "the power distribution cabinet is normal". This result indicates that the overall operating state of the power distribution cabinet is good and no further intervention measures are required. At this time, the remote maintenance module only needs to upload and store the daily data for subsequent status monitoring and trend analysis. When the power distribution cabinet is abnormal: when the system detects that the overall parameters of the power distribution cabinet (such as too high temperature, current overload, etc.) exceed the safety threshold and the abnormal trend involves multiple key components, the residual neural network will output a judgment result of "the power distribution cabinet is abnormal". This situation usually reflects that there may be potential risks in the overall power distribution cabinet, such as overheating, overload or other systematic failures. This judgment result will be pushed to the operation and maintenance personnel terminal in real time through the remote maintenance module, accompanied by the AR visualization display of the 3D model, enabling the operation and maintenance personnel to view the abnormal situation of the power distribution cabinet from a global perspective, quickly locate the potential problem source, and arrange maintenance or repair operations in a timely manner. When a component is abnormal: when the operating parameters of one or a few specific components deviate from the normal range while other components remain normal, the residual neural network will separately identify this component as "component abnormal" and generate the corresponding judgment result. For example, the system may detect that the temperature of a certain circuit breaker rises abnormally or the voltage is unstable, while the operating states of other components remain within the normal range. In this case, the remote maintenance module will push the abnormal information and related operating parameters of this specific component to the maintenance personnel terminal in an AR visualization manner, combined with highlighting the location and numerical information of the faulty component, enabling the operation and maintenance personnel to accurately identify the problem component and conduct targeted inspections and maintenance, avoiding affecting the overall function of the power distribution cabinet.

[0062] Furthermore, the calculation formula of the residual neural network model is as follows:

[0063] ;

[0064] where, is the predicted judgment result of the power distribution cabinet at time t; is the input feature at time t, including the operating parameter matrix and the environmental parameters of the power distribution cabinet; is the convolution kernel parameter of the th layer; is the bias term; is the convolution kernel weight matrix of the output layer; is the bias term of the output layer; is the activation function; is the output of the previous moment of the th layer in the residual neural network; is the number of network layers.

[0065] Specifically, the network performs layer-by-layer feature extraction on the input features through convolutional layers, forming a residual path with cumulative addition between layers. The output of each layer is not only affected by the current input features but also superimposed with the output features of the previous layer to ensure the effective transmission of deep information. The weight matrix of the final output layer performs weighted processing on the accumulated features to generate the final determination result. This residual connection mechanism avoids the problem of gradient disappearance by accumulating features across layers and enhances the stability and recognition ability of the model when dealing with high-dimensional complex fault patterns. Finally, after the residual features accumulated through multiple layers are input into the L-th layer, they are processed by the activation function to generate the final predicted determination result of the power distribution cabinet at time t .

[0066] Furthermore, the loss function of the residual neural network model is as follows:

[0067] ;

[0068] where is the total loss function; is the true determination result of the power distribution cabinet at time t; is the predicted determination result of the power distribution cabinet at time t; is the mean square error between the predicted value and the true value; is the convolution kernel parameter of the -th layer; is the regularization coefficient; is the number of samples in the training dataset; is the number of network layers.

[0069] Specifically, this loss function combines the mean square error term and the regularization term, optimizes the prediction accuracy of the model by minimizing the mean square error, and at the same time constrains the convolution kernel parameters through the regularization term to prevent the model from overfitting. During the training process of the residual neural network, the weights and bias terms are continuously adjusted to make the loss function reach the minimum value, thereby achieving high-precision prediction and anomaly detection of the power distribution cabinet status.

[0070] Furthermore, the remote maintenance module includes the cloud and the maintenance personnel terminal;

[0071] The remote maintenance module is used to receive several 3D models of power distribution cabinets and send the 3D models of power distribution cabinets to the maintenance personnel terminal according to the anomaly determination results, and is also used to receive the maintenance record information uploaded by the personnel terminal and synchronously update it to the component management module;

[0072] The maintenance personnel terminal is used to receive the 3D models of power distribution cabinets and perform AR visualization, and is also used to receive the maintenance record information and perform standardized upload to the remote maintenance module.

[0073] ​It should be noted that the remote maintenance module is designed to consist of a cloud end and a maintenance personnel end to achieve efficient remote monitoring and maintenance support for the power distribution cabinet. The remote maintenance module combines cloud computing and augmented reality (AR) technologies to ensure that the real-time status of the power distribution cabinet can be promptly communicated and visually displayed, and supports the real-time synchronization and update of maintenance records. During daily operation, the cloud end of the remote maintenance module is responsible for receiving and storing the 3D models of multiple power distribution cabinets. These models are generated by the model construction module and cover the real-time operation status and spatial distribution information of all components within the power distribution cabinet. Whenever the abnormal monitoring module detects an abnormality in the operating parameters or environmental parameters of the power distribution cabinet and generates a corresponding abnormal determination result, the cloud end will immediately push the 3D model of the corresponding power distribution cabinet and the determination result to the maintenance personnel end. This real-time synchronization mechanism enables the operation and maintenance personnel to quickly obtain the abnormal information of the power distribution cabinet, and they can intuitively understand the specific location and fault details of the components within the power distribution cabinet without having to arrive at the scene, thus improving the maintenance response speed. At the maintenance personnel end, the received 3D model of the power distribution cabinet will be presented in an AR visualization manner, enabling the maintenance personnel to see the component distribution and fault location inside the power distribution cabinet on augmented reality devices (such as smart glasses, tablets or mobile phones). Through AR technology, the operation and maintenance personnel can accurately locate the faulty components in the model and observe the relevant parameters of the components, such as current, voltage, temperature, etc., so as to provide effective decision-making support. For example, when the temperature of a certain circuit breaker in the power distribution cabinet rises abnormally, the AR visualization at the maintenance personnel end will highlight the circuit breaker and display the temperature value and other associated parameters in real time, facilitating the operation and maintenance personnel to evaluate the specific situation. In addition, after the maintenance personnel handle the fault or complete the maintenance operation, they can enter detailed maintenance record information through the maintenance personnel end. These record information include maintenance content, operation steps, component replacement situation and test results, etc. After being entered, the system will standardize the maintenance record to ensure the unity and standardization of the data. The processed maintenance record information will be uploaded to the cloud end and synchronized and updated to the component management module through the cloud end. After receiving the updated maintenance record, the component management module will combine and archive this information with the historical record of the component to provide reference for subsequent equipment monitoring and maintenance. This design of the remote maintenance module ensures the complete traceability and recording of maintenance information, and can be synchronized and updated in real time after the component fails or is replaced, enabling the system to effectively accumulate the maintenance historical data of the power distribution cabinet during long-term operation, so as to provide more reliable support in future fault prediction and operation and maintenance decision-making. This module realizes the remote monitoring, fault diagnosis, AR visualization and automated management of maintenance records of the power distribution cabinet through efficient data transmission and information synchronization, greatly improving the intelligent level and maintenance efficiency of the system.

[0074] The above embodiments are only descriptions of the preferred embodiments of the present invention, and do not limit the scope of the present invention. Without departing from the design spirit of the present invention, various deformations and improvements made by those of ordinary skill in the art to the technical solutions of the present invention shall fall within the protection scope determined by the claims of the present invention.

Claims

1. A remote maintenance system for an intelligent power distribution cabinet, characterized in that, It includes a component management module, an anomaly monitoring module, a model construction module, and a remote maintenance module connected in sequence, and also includes a cabinet internal environment monitoring module. The cabinet internal environment monitoring module is connected to the anomaly monitoring module, and the remote maintenance module is connected to the component management module; The component management module is used to update the component attribute parameters and component operation parameters in the power distribution cabinet in real time. The component operation parameters include current and voltage, component temperature, and component activation time; The cabinet internal environment monitoring module is used to obtain the power distribution cabinet environment parameters in real time, including the humidity inside the cabinet and the temperature and humidity outside the cabinet; The anomaly monitoring module is used to construct an operation parameter matrix in the time series based on the component operation parameters in the power distribution cabinet, perform residual neural network training based on the operation parameter matrix and the power distribution cabinet environment parameters, and perform anomaly determination on the real-time component operation parameters through the trained residual neural network to generate a real-time determination result; The model construction module is used to construct a three-dimensional framework of the power distribution cabinet, and combine the three-dimensional framework of the power distribution cabinet according to the component category, component model, and component connection relationship to match a preset component library to obtain a real-time three-dimensional model of the power distribution cabinet. The model construction module is also used to perform synchronous mapping on the three-dimensional model of the power distribution cabinet based on the power distribution cabinet environment parameters and the real-time determination result; The remote maintenance module is used to receive the three-dimensional model of the power distribution cabinet from the model construction module and visualize it, and is also used to receive maintenance record information to update the component management module.

2. The remote maintenance system for an intelligent power distribution cabinet according to claim 1, characterized in that, The component attribute parameters include component category, component model, and component connection relationship.

3. The remote maintenance system of an intelligent power distribution cabinet according to claim 2, characterized in that, The operation parameter matrix includes a current and voltage matrix, a component temperature matrix, and a component activation time matrix.

4. The remote maintenance system of an intelligent power distribution cabinet according to claim 3, characterized in that, The constructing the operation parameter matrix in the time series based on the component operation parameters in the power distribution cabinet includes the following steps: Obtain the component operation parameters from the component management module; Preset the value filling areas for the current and voltage matrix, the component temperature matrix, and the component activation time matrix based on the component connection relationship; Remove outliers and noise from the component operation parameters and perform standardization processing to obtain standardized values; Correspondingly fill the standardized values in the value filling areas of the current and voltage matrix, the component temperature matrix, and the component activation time matrix respectively.

5. The remote maintenance system of an intelligent power distribution cabinet according to claim 1, characterized in that, The performing residual neural network training based on the operation parameter matrix and the power distribution cabinet environment parameters includes the following steps: Obtain the operation parameter matrix and the power distribution cabinet environment parameters to construct a training data set; Construct a residual neural network model, use the training data set as input, and perform feature extraction on the training data through the convolutional layer to obtain a multi-dimensional feature map; Input the multi-dimensional feature map into the residual module and extract high-dimensional features through the residual connection mechanism; Input the extracted high-dimensional features into the fully connected layer and output the determination result through the activation function.

6. The remote maintenance system of an intelligent power distribution cabinet according to claim 5, characterized in that, The determination result includes that the power distribution cabinet is normal, the power distribution cabinet is abnormal, and the component is abnormal.

7. The remote maintenance system of an intelligent power distribution cabinet according to claim 6, characterized in that, The calculation formula of the residual neural network model is as follows: ； Among them, is the predicted determination result of the power distribution cabinet at time t; is the input feature at time t, including the operation parameter matrix and the power distribution cabinet environment parameters; is the convolution kernel parameter of the layer; is the bias term; is the convolution kernel weight matrix of the output layer; is the bias term of the output layer; is the activation function; is the output of the previous moment of the layer in the residual neural network; is the number of network layers.

8. The remote maintenance system of an intelligent power distribution cabinet according to claim 7, characterized in that, The loss function of the residual neural network model is as follows: ; in, is the total loss function; is the actual judgment result of the distribution cabinet at time t; is the prediction result of the distribution cabinet at time t; is the mean square error between the predicted value and the true value; For the The convolution kernel parameters of the layer; is the regularization coefficient; is the number of samples in the training data set; is the number of network layers.

9. The remote maintenance system of an intelligent power distribution cabinet according to claim 1, characterized in that, The remote maintenance module includes a cloud end and a maintenance personnel end; The remote maintenance module is used to receive a number of three-dimensional models of distribution cabinets and send the three-dimensional models of distribution cabinets to the maintenance personnel terminal according to the abnormal determination result, and is also used to receive the maintenance record information uploaded by the personnel terminal and synchronously update it to the component management module; The maintenance personnel terminal is used to receive the three-dimensional model of the distribution cabinet and perform AR visualization, and is also used to receive the maintenance record information and perform standardized upload to the remote maintenance module.

Images