Outdoor mobile power supply fault remote identification system combined with cloud edge fusion

Abstract

The application provides an outdoor mobile power supply fault remote identification system combined with cloud edge fusion, relates to the technical field of mobile power supplies, and comprises the following steps: a real-time monitoring module acquires power supply monitoring data; a first data transmission module transmits the data to a state inspection edge node; a sensing distortion correction module performs sensing distortion correction according to a sensing distortion correction channel, and acquires power supply state data; a state inspection module inspects the power supply state data by using a power supply state inspection channel, and acquires a state inspection result; when the inspection fails, a second data transmission module transmits the data to a fault detection cloud node; and a fault remote identification module performs fault remote identification according to a fault detection gate channel and a power supply fault detection channel. The application solves the technical problem that the existing technology lacks the collaborative means of cloud computing and edge computing, resulting in insufficient timeliness and accuracy of outdoor mobile power supply fault identification, and improves the fault identification response speed and accuracy.

Description

Technical Field

[0001] This application relates to the field of mobile power technology, specifically to a remote fault identification system for outdoor mobile power banks that combines cloud and edge computing. Background Technology

[0002] Outdoor portable power banks play a vital role in outdoor activities and emergencies due to their portability, high efficiency, environmental friendliness, and long battery life. With the development of intelligent technology, the use and maintenance of outdoor portable power banks have gradually incorporated advanced functions such as intelligent management and remote control, providing users with a more convenient, efficient, and safe user experience.

[0003] Existing remote fault identification methods for outdoor portable power banks primarily rely on edge computing or cloud computing. Edge computing devices are typically small nodes deployed close to the power bank, with limited processor performance and memory capacity, making it difficult to run complex fault identification algorithms and limiting their ability to handle complex fault situations. In cloud computing, data transmission to the cloud is delayed, and even simple faults require deep cloud intervention. This results in significant cloud computing resource consumption when handling simple faults that could be resolved at the edge, leading to wasted computing resources and limiting fault identification efficiency. Furthermore, due to the complexity of the outdoor portable power bank's operating environment, monitoring sensors may be affected by environmental factors (such as temperature, humidity, and electromagnetic interference) and equipment aging, causing deviations in monitoring data and affecting the accuracy of fault identification. Summary of the Invention

[0004] This application provides a cloud-edge integrated remote fault identification system for outdoor mobile power banks, which solves the technical problem that the lack of collaborative means between cloud computing and edge computing in existing technologies leads to insufficient timeliness and accuracy in fault identification of outdoor mobile power banks, and achieves the technical effect of improving fault identification response speed and accuracy.

[0005] In view of the above problems, this application provides a remote fault identification system for outdoor mobile power banks that combines cloud and edge computing. The system includes: a real-time monitoring module for real-time monitoring of the outdoor mobile power bank to obtain power monitoring data; a first data transmission module for transmitting the power monitoring data to a status verification edge node, wherein the status verification edge node includes a sensor distortion correction channel and a power status verification channel deployed on the edge computing node; a sensor distortion correction module for performing sensor distortion correction on the power monitoring data according to the sensor distortion correction channel to obtain power status data; and a status verification module for obtaining the power status of the outdoor mobile power bank. The system comprises: a source control data module, which combines the power status data and the power status verification channel to perform a status verification on the outdoor portable power bank and obtain a power status verification result; a second data transmission module, which transmits the power status data to a fault detection cloud node when the power status verification result is a failure, wherein the fault detection cloud node includes a fault detection gating channel and a power fault detection channel deployed on a cloud computing node; and a remote fault identification module, which performs remote fault identification on the outdoor portable power bank based on the power status data, according to the fault detection gating channel and the power fault detection channel, and obtains a power fault identification report.

[0006] One or more technical solutions provided in this application have at least the following technical effects or advantages:

[0007] The outdoor portable power bank undergoes real-time monitoring via a real-time monitoring module, continuously acquiring real-time power status information to provide power monitoring data for subsequent data processing and fault identification. The first data transmission module transmits the collected power monitoring data to a status verification edge node, where a sensor distortion correction module corrects the power monitoring data to eliminate sensor errors and distortions, ensuring that the edge computing node receives accurate power monitoring data in real time. The status verification module performs a status verification on the outdoor portable power bank using power control data and power status data, deriving a preliminary judgment on whether the power bank is in an abnormal state. Normal power monitoring data is filtered out, reducing the amount of data processed in the cloud and thus improving the overall efficiency of fault identification. When the power status verification result indicates a problem with the power bank, the second data transmission module transmits the power status data to a fault detection cloud node, ensuring that the problematic data can be further analyzed and processed by the cloud node. At the fault detection cloud node, a remote fault identification module performs remote fault identification based on the power status data and generates a power bank fault identification report, achieving remote fault diagnosis.

[0008] In summary, this application, through a sensor distortion correction module and a cloud-edge fusion architecture, can more accurately acquire power status data and perform more comprehensive fault identification, reducing false positives and false negatives. The collaborative work of edge nodes and cloud nodes ensures the rational allocation of computing resources. Edge nodes handle simple status checks, while cloud nodes handle complex fault identification, avoiding waste of computing resources and improving the overall performance of the fault identification system. The layered inspection and identification process, along with the cloud-edge fusion architecture, reduces unnecessary data transmission and computation, enabling rapid fault identification of outdoor portable power banks, improving fault identification efficiency, facilitating timely handling of faulty power banks, and enhancing the safety and reliability of outdoor portable power banks.

[0009] The above description is only an overview of the technical solution of this application. In order to better understand the technical means of this application and to implement it in accordance with the contents of the specification, and to make the above and other objects, features and advantages of this application more obvious and understandable, the following are specific embodiments of this application. Attached Figure Description

[0010] Figure 1 This is a schematic diagram of the structure of an outdoor mobile power supply fault remote identification system that combines cloud-edge fusion, as provided in an embodiment of this application.

[0011] Figure 2 This is a schematic diagram illustrating the process of acquiring power status data in an outdoor mobile power supply fault remote identification system that combines cloud-edge fusion, as provided in an embodiment of this application.

[0012] Figure 3 This is a schematic diagram illustrating the process of generating power status prediction results in an outdoor mobile power supply fault remote identification system that combines cloud-edge fusion, as provided in an embodiment of this application.

[0013] Explanation of reference numerals in the attached diagram: Real-time monitoring module 10, first data transmission module 20, sensor distortion correction module 30, status verification module 40, second data transmission module 50, and remote fault identification module 60. Detailed Implementation

[0014] This application provides a cloud-edge fusion-based remote fault identification system for outdoor mobile power banks. It monitors and acquires power monitoring data from outdoor mobile power banks in real time, transmitting this data via a first data transmission module 20 to a status verification edge node. The status verification edge node uses deployed sensor distortion correction channels and power status verification channels to perform sensor distortion correction and power status verification on the monitoring data. When the power status verification result is a failure, the second data transmission module 50 transmits the power status data to a fault detection cloud node. The fault detection cloud node remotely identifies the fault in the outdoor mobile power bank using deployed fault detection gating channels and power fault detection channels, obtaining a power fault identification report. This solves the technical problem of insufficient timeliness and accuracy in outdoor mobile power bank fault identification due to the lack of collaborative means between cloud computing and edge computing in existing technologies, achieving the technical effect of improving fault identification response speed and accuracy.

[0015] like Figure 1 As shown in the figure, this application embodiment provides an outdoor mobile power bank fault remote identification system combining cloud-edge fusion, the system comprising:

[0016] The real-time monitoring module 10 is used to monitor the outdoor portable power bank in real time and obtain power monitoring data.

[0017] Specifically, the real-time monitoring module 10 continuously acquires relevant parameter data of the power bank by connecting to various sensors inside the power bank, such as current sensors, voltage sensors, and temperature sensors. This includes key parameters such as the power bank's capacity, charging status, temperature, and voltage. For example, in an outdoor power bank, a high-precision voltage sensor collects voltage data every 10 seconds, and then transmits the data to the monitoring module through an interface circuit.

[0018] By monitoring the outdoor power bank through the real-time monitoring module 10, the operating data of the outdoor power bank can be obtained in real time and comprehensively, providing a raw data basis for subsequent fault identification.

[0019] The first data transmission module 20 is used to transmit the power monitoring data to the status verification edge node, wherein the status verification edge node includes a sensing distortion correction channel and a power status verification channel deployed on the edge computing node.

[0020] Specifically, a status verification edge node is a system node deployed in an edge computing environment, containing a sensor distortion correction channel and a power status verification channel. Edge computing nodes are computing devices located near outdoor portable power banks, performing preliminary data processing.

[0021] The first data transmission module 20 first determines the format and transmission protocol of the power monitoring data, establishes a communication connection between the real-time monitoring module 10 and the status inspection edge node, and then sends the power monitoring data to the status inspection edge node through wired (such as USB, Ethernet, etc.) or wireless (such as Wi-Fi, Bluetooth, etc.) communication methods to ensure the flow of data in the fault identification system and enable subsequent edge processing.

[0022] The sensing distortion correction module 30 is used to perform sensing distortion correction on the power monitoring data according to the sensing distortion correction channel to obtain power status data.

[0023] Specifically, the sensing distortion correction channel is a logic channel used to correct for potential sensing distortions in power supply monitoring data. For example, if a voltage sensor's readings are too high due to aging, this sensing distortion correction channel can use an algorithm to correct this deviation.

[0024] Upon receiving power monitoring data, the sensor distortion correction module 30 analyzes potential distortions in the data using a pre-set algorithm in the sensor distortion correction channel, and corrects the power monitoring data to obtain corrected power status data. For example, in a case where voltage measurement deviations exist due to ambient temperature, the module finds the corresponding correction value by fitting a temperature-voltage relationship curve, and then corrects the data using simple addition or multiplication operations. This data correction reduces data errors caused by sensor malfunctions, providing more accurate monitoring data for precise status checks and fault identification.

[0025] The status verification module 40 is used to obtain the power control data of the outdoor portable power bank, and to perform status verification on the outdoor portable power bank in combination with the power status data and the power status verification channel to obtain the power status verification result.

[0026] Specifically, power control data refers to the various parameters and commands used to control power output in an outdoor portable power supply management system. This includes setpoints for parameters such as voltage, current, and power, as well as commands for controlling power switches, adjusting output voltage, and protecting circuits. The power status verification channel is a logic channel used for preliminary verification of the outdoor portable power supply's status.

[0027] First, the power control data of the portable power bank is acquired, such as charging and discharging status information from the power management system. This data, along with calibrated power status data, is then input into the power status verification channel. Within this channel, the power status is assessed based on preset status judgment logic, such as setting normal voltage and current ranges, generating a power status verification result indicating whether the power bank is functioning normally. For example, if the power status data falls within the normal operating range, the verification result is "verified," indicating the portable power bank is in a normal state; if the data does not fall within the normal operating range, the verification result is "verified failed," indicating the portable power bank is in an abnormal state, requiring fault identification and diagnosis.

[0028] By using the status inspection module 40 to perform preliminary status inspections on outdoor power banks at edge nodes, potential faulty power banks can be quickly identified, thus determining the targets for subsequent cloud-based fault identification.

[0029] The second data transmission module 50 is used to transmit the power status data to the fault detection cloud node when the power status test result is unsuccessful. The fault detection cloud node includes a fault detection gating channel and a power fault detection channel deployed on the cloud computing node.

[0030] Specifically, a fault detection cloud node is a system node deployed in a cloud computing environment, comprising a fault detection gating channel and a power fault detection channel. A cloud computing node is a hardware device for remote fault identification, possessing powerful computing resources capable of performing complex calculations.

[0031] When the power status check fails, the second data transmission module 50 establishes a communication connection between the edge computing node and the fault detection cloud node. Following a predetermined transmission protocol and method, such as HTTP or TCP / IP, the power status data is transmitted to the fault detection cloud node. For example, if HTTP is used, the data is packaged into a format conforming to HTTP requirements and then sent to the cloud over the network. By selectively transmitting potentially faulty data to the cloud, unnecessary data transmission is avoided, saving network resources and reducing the amount of data the cloud needs to process, thereby improving the efficiency of fault identification.

[0032] The remote fault identification module 60 is used to remotely identify faults in the outdoor portable power bank based on the power status data, according to the fault detection gating channel and the power fault detection channel, and obtain a power fault identification report.

[0033] Specifically, a fault detection gating channel is a logical channel that controls and filters fault detection, such as determining which data needs further in-depth testing. A power supply fault detection channel is a logical channel used for in-depth fault detection of power supplies.

[0034] After receiving power status data from the cloud, the remote fault identification module 60 judges and filters the power status data according to the control logic of the fault detection gating channel before inputting it into the power fault detection channel. For example, it first judges the completeness and validity of the data, and then inputs valid data into the power fault detection channel. Within the power fault detection channel, it utilizes the powerful computing resources of the cloud to run fault identification algorithms, such as machine learning algorithms, to perform deep fault identification on the power supply and generate a power fault identification report. For example, it uses TensorFlow to build a neural network model to analyze data such as voltage, current, and temperature of the power supply and identify faults such as internal short circuits in the battery and damage to circuit components. The power fault identification report contains specific information about the outdoor portable power bank fault, such as the time of the fault, the location of the fault, and the cause of the fault.

[0035] The remote fault identification module 60 enables rapid fault identification and response, reducing the need for on-site maintenance and improving maintenance efficiency.

[0036] Furthermore, the sensing distortion correction channel includes a sensing anomaly identification model and a sensing distortion correction model, such as... Figure 2 As shown, the sensing distortion correction module 30 is also used to perform the following steps:

[0037] Step P31: Traverse the power monitoring data and extract the first power monitoring parameter.

[0038] Step P32: Collect the sensor status information corresponding to the first parameter of the power supply monitoring to obtain the first sensor status data.

[0039] Step P33: Input the first data of the sensing state into the sensing anomaly identification model to obtain the first sensing anomaly identification result.

[0040] Step P34: Based on the first sensor anomaly identification result and the first power monitoring parameter, obtain the first power monitoring correction parameter according to the sensor distortion correction model, and add the first power monitoring correction parameter to the power status data.

[0041] Specifically, the sensor distortion correction channel deployed on edge computing nodes consists of a sensor anomaly identification model and a sensor distortion correction model. The sensor anomaly identification model identifies whether there are any anomalies in the power monitoring data, while the sensor distortion correction model corrects the distorted portions of the power monitoring data after confirming the presence of anomalies. Both the sensor anomaly identification model and the sensor distortion correction model are built based on machine learning algorithms, such as support vector machines and linear regression algorithms.

[0042] Taking a sensor anomaly recognition model as an example, firstly, a large amount of sensor data under normal conditions and data under abnormal sensor conditions is collected to construct a historical dataset for sensor anomaly recognition. Then, this historical dataset is divided into a training set and a test set, for example, with 70% of the data used as the training set and 30% as the test set. The Support Vector Machine (SVM) algorithm is selected to train the sensor anomaly recognition model. The training set is used to train the SVM model, determining the model's parameters, such as the kernel function (linear kernel, Gaussian kernel, etc.) and penalty parameters. During training, the model gradually learns how to distinguish the feature differences between normal and abnormal states of various types of sensors. Finally, the trained model is evaluated using the test set, such as by calculating metrics like accuracy, recall, and F1 score to measure its performance. If the model performance does not meet the requirements, the optimizer is used to adjust the model parameters or a new algorithm is selected, and training and evaluation are repeated until the model converges.

[0043] The training process for the sensor distortion correction model is similar to that for the sensor anomaly identification model, involving the collection of a large amount of power supply monitoring data containing sensor distortion conditions. This data should cover different degrees and types of sensor distortion, such as voltage measurement deviations and current measurement deviations of varying magnitudes. A regression algorithm can be used to construct the sensor distortion correction model. The model is trained using a training set to determine the coefficients in the regression equation. The trained model is then validated using a test set, and metrics such as mean squared error (MSE) are calculated to evaluate the model's accuracy. If the model's accuracy does not meet expectations, feature selection can be adjusted or the amount of data increased, and the model can be retrained until a model that can accurately correct sensor distortion is obtained.

[0044] A monitoring parameter is randomly selected from the acquired power monitoring data and designated as the first power monitoring parameter. After determining the first power monitoring parameter, the corresponding sensor status information is acquired, including the sensor's operating temperature and operating time. This sensor status information is typically stored in specific registers or configuration files associated with the sensor. By querying the relevant sensor interfaces or reading the configuration file, the sensor status information corresponding to the first power monitoring parameter is collected, then organized and combined to obtain the first sensor status data. The obtained first sensor status data provides more background information, offering a more comprehensive basis for sensor anomaly identification and improving the accuracy of anomaly identification.

[0045] The initial sensor status data is input into a pre-built sensor anomaly identification model. This model determines whether the sensor is malfunctioning based on the characteristics of the input sensor status data. For example, the initial sensor status data might be a vector containing information such as sensor temperature and whether it's in a calibration cycle. This vector is input into a sensor anomaly identification model built on a Support Vector Machine (SVM). The model analyzes the input vector based on previously learned feature patterns from normal and abnormal data to arrive at the first sensor anomaly identification result. This first anomaly identification provides specific information about the sensor status corresponding to the first power monitoring parameter, including whether the sensor is malfunctioning, the type of malfunction, and the degree of malfunction.

[0046] After obtaining the first sensor anomaly identification result, if the result indicates the presence of a sensor anomaly, the first power monitoring parameter and the first sensor anomaly identification result are used as input data and fed into the sensor distortion correction model. The sensor distortion correction model calculates the first power monitoring correction parameter based on its internal algorithm and the input data. This correction parameter is then added to the power status data.

[0047] Traverse the power monitoring data, extract parameters one by one, and repeat the above operations (steps P31 to P34) to determine the correction parameter corresponding to each monitoring parameter. Integrate these correction parameters according to the data structure of the power monitoring data to obtain the final power status data.

[0048] By correcting data with sensor anomalies, the accuracy of power status data is improved, enabling subsequent status checks and fault identification operations to be based on more accurate data, thereby improving the accuracy of outdoor portable power bank status assessment.

[0049] Furthermore, step P34 includes:

[0050] Step P341: Evaluate the sensing distortion based on the first sensing anomaly identification result to obtain the first sensing distortion.

[0051] Step P342: Determine whether the first sensing distortion is greater than or equal to the predetermined sensing distortion.

[0052] Step P343: If the first sensing distortion is greater than or equal to the predetermined sensing distortion, activate the sensing distortion correction model.

[0053] Step P344: Input the first sensor anomaly identification result and the first power monitoring parameter into the sensor distortion correction model to obtain the first power monitoring correction parameter.

[0054] Specifically, sensing distortion is a quantitative indicator used to measure the degree of sensor anomaly. The predetermined sensing distortion is a predefined threshold based on the accuracy requirements of the sensing data. Data analysis software evaluates the first sensing anomaly identification result, determines the degree of sensor anomaly, and generates a specific sensing distortion, i.e., the first sensing distortion. For example, the first sensing anomaly identification result can be compared with the sensor state data under ideal conditions, and the relative deviation between the two can be calculated as the first sensing distortion.

[0055] The first sensing distortion is compared with a predetermined sensing distortion. If the first sensing distortion is greater than or equal to the predetermined distortion, it indicates an anomaly in the sensor. The sensing distortion correction model is then activated, and the first sensing anomaly identification result and the first power monitoring parameter are input into the model for correction, resulting in the first power monitoring correction parameter. If the first sensing distortion is less than the predetermined distortion, the first power monitoring parameter is output as the first power monitoring correction parameter. By comparing with a predetermined threshold, it is possible to quickly determine whether the distortion level of the current sensing data meets the standard requiring correction, thus deciding whether to activate the correction model and avoiding unnecessary correction operations or missed corrections. Obtaining accurate correction parameters through the sensing distortion correction model enables effective correction of the power monitoring data, improving data accuracy and providing a more reliable data foundation for subsequent power status judgment, fault diagnosis, and other operations.

[0056] Furthermore, the power state verification channel includes a power state deviation risk assessment model and a state deviation risk verification model. The state verification module 40 is also used to perform the following steps:

[0057] Step P41: Based on the power control data, predict the health status of the outdoor portable power bank to obtain the power status prediction result.

[0058] Step P42: Input the power state data and the power state prediction result into the power state deviation risk assessment model to obtain the power state deviation risk coefficient.

[0059] Step P43: Input the power state deviation risk coefficient into the state deviation risk test model, wherein the state deviation risk test model includes a power state deviation risk test operator, and the power state deviation risk test operator includes a power state test result of failing if the power state deviation risk coefficient is greater than or equal to the power state deviation risk threshold, and a power state test result of passing if the power state deviation risk coefficient is less than the power state deviation risk threshold.

[0060] Step P44: Output the power state check result according to the power state deviation risk check operator.

[0061] Specifically, the power state verification channel deployed on edge computing nodes includes a power state deviation risk assessment model and a state deviation risk verification model. The power state deviation risk assessment model is used to evaluate the risk arising from the degree of deviation of the power state from the normal state. It calculates a coefficient representing the degree of deviation risk based on the input power state data and power state prediction results. The state deviation risk verification model is used to verify the power state deviation risk coefficient to determine whether the power state is acceptable. It includes a power state deviation risk verification operator, which defines the rules for judging whether the power state passes the verification. The power state deviation risk coefficient is a quantified value used to represent the degree of risk brought about by the power state deviating from the normal state. The larger the value, the higher the risk; the smaller the value, the lower the risk.

[0062] Based on power control data, the health status of outdoor portable power banks is predicted, yielding a power status prediction result. This prediction result is a quantitative value representing the future health status of the portable power bank, ranging from 0 to 1, where 0 indicates complete failure and 1 indicates complete health. The health status prediction can be compared with historical data to foreshadow future power bank status. By analyzing and predicting power control data, we can understand the health status trend of outdoor portable power banks in advance, providing a basis for subsequent risk assessment and status inspection, and helping to identify potential problems in a timely manner and take appropriate measures.

[0063] Power state data and power state prediction results are input into the power state deviation risk assessment model. This model is pre-trained using machine learning algorithms and processes the input data according to predefined algorithms and parameters, ultimately outputting a power state deviation risk coefficient. Deep learning frameworks such as TensorFlow or PyTorch can be used to build the power state deviation risk assessment model, utilizing the framework's model loading, data preprocessing, and prediction functions to obtain the power state deviation risk coefficient.

[0064] The power supply status deviation risk coefficient is input into the status deviation risk inspection model. The power supply status deviation risk inspection operator in the model compares the input risk coefficient with the power supply status deviation risk threshold. This power supply status deviation risk threshold is a pre-set critical value used to classify whether the power supply status is acceptable. If the power supply status deviation risk coefficient is greater than or equal to the power supply status deviation risk threshold, the power supply status inspection result is failed, indicating that there may be a problem with the power supply; if it is less than the power supply status deviation risk threshold, the power supply status inspection result is passed, indicating that the power supply status is normal. This power supply status inspection result can also include a more detailed report, such as indicating which factors caused the failure (e.g., voltage deviating too much from the normal range) in the case of failure.

[0065] Through the above steps, the status inspection module 40 can accurately predict and assess the health status of the outdoor power bank, promptly identify potential problems, and thus improve the safety and reliability of the outdoor power bank.

[0066] Furthermore, such as Figure 3 As shown, step P41 includes:

[0067] Step P411: Retrieve the power health status record database of the outdoor portable power bank.

[0068] Step P412: Based on the power control data, optimize the power health status record library according to the power control registration threshold to establish a power health status prediction space.

[0069] Step P413: Perform a central tendency evaluation based on the power health status prediction space to generate the power status prediction result.

[0070] Specifically, the data storage device of the interactive outdoor portable power bank retrieves the power health status record library of the outdoor portable power bank. This power health status record library refers to a database that stores historical health status data of the outdoor portable power bank, including multiple power health status record groups, such as records of past parameters such as voltage, current, temperature, and battery capacity, as well as the corresponding power health status.

[0071] A power control registration threshold is a pre-defined standard or range used to filter and evaluate data in a power supply health status record library. Based on the power control registration threshold, power status records in the health status record library that match the power control data within the threshold are searched. These selected power status records are then combined to form a power supply health status prediction space. This prediction space encompasses the power supply's health status under different conditions and can be used for subsequent predictive analysis. By establishing a power supply health status prediction space, the dimensionality and complexity of the data can be reduced, while simultaneously improving the accuracy of power supply status prediction results.

[0072] Central tendency analysis is performed on the data in the power supply health status prediction space. Statistical analysis software is used to find the central location or typical values ​​of the data in the power supply health status prediction space, such as central tendency indicators like the mean, median, and mode. The determined central tendency indicator values ​​are used as the power supply status prediction results.

[0073] By performing central trend analysis on historical power supply health records, we can make full use of historical data to determine the current health status of outdoor portable power banks, thereby obtaining accurate power supply status inspection results.

[0074] Furthermore, step P412 includes:

[0075] Step P412-1: Extract the first power health status record group according to the power health status record library, wherein the first power health status record group includes the first power control record and the first power health status record.

[0076] Step P412-2: Perform a similarity evaluation on the power control data and the first power control record to obtain the first power control registration coefficient.

[0077] Step P412-3: Determine whether the first power control registration coefficient is greater than or equal to the power control registration threshold.

[0078] Step P412-4: If the first power control registration coefficient is greater than or equal to the power control registration threshold, set the first power health status record as the first power health status prediction result, and add the first power health status prediction result to the power health status prediction space.

[0079] Step P412-5: Based on the power control data and the power control registration threshold, continue to optimize and select the power health status record library to generate the power health status prediction space.

[0080] Specifically, when establishing the power supply health status prediction space, a power supply health status record is first randomly selected from multiple power supply health status record groups in the power supply health status record library, and denoted as the first power supply health status record group. This first power supply health status record group contains the control parameter record at a certain historical moment (i.e., the first power supply control record) and the corresponding power supply health status information at that moment (i.e., the first power supply health status record).

[0081] A similarity evaluation is performed on the power control data and the first power control record, and the similarity evaluation result is used as the registration coefficient for the first power control. Similarity evaluation can use similarity calculation methods such as cosine similarity and Pearson correlation distance. Taking the cosine similarity method as an example, the power control data and the first power control record are converted into vector representations, and then the similarity value between the two vectors is calculated according to the cosine similarity formula. The result obtained is the registration coefficient for the first power control.

[0082] Compare the first power control registration coefficient with the power control registration threshold, and determine whether the first power control registration coefficient is greater than or equal to the power control registration threshold. If the first power control registration coefficient is greater than or equal to the power control registration threshold, it indicates that the first power control record is quite similar to the power control data, and the health status of the current outdoor portable power bank can be predicted using the first power health status record corresponding to the first power control record. In this case, the first power health status record is set as the first result of the power health status prediction, and the first result of the power health status prediction is added to the power health status prediction space.

[0083] Then, continue to select new power health status record groups from the power health status record library, repeat the above similarity evaluation and threshold judgment operations, and filter new power health status records from the power health status record library to add to the power health status prediction space.

[0084] By following the steps above, records similar to the current power control data can be selected from a large number of historical power health status records, thus constructing a representative power health status prediction space and providing a data foundation for accurate power status prediction.

[0085] Furthermore, the remote fault identification module 60 is also used to perform the following steps:

[0086] Step P61: Input the power state deviation risk coefficient into the fault detection gating channel to obtain the fault detection gating coefficient.

[0087] Step P62: Based on the fault detection gating coefficient, perform feature activation on the power supply fault detection channel to obtain M fault detection activation models, wherein the power supply fault detection channel includes P power supply fault detection models corresponding to the outdoor mobile power supply, and M and P are both positive integers, 1≤M≤P.

[0088] Step P63: Input the power status data into the M fault detection activation models to obtain M fault detection results.

[0089] Step P64: Generate the power supply fault identification report based on the M fault detection results.

[0090] Specifically, the fault detection gating channel is a module or mechanism dedicated to adjusting fault detection-related operations based on the power state deviation risk coefficient. It converts the input power state deviation risk coefficient into a fault detection gating coefficient, thereby regulating the subsequent fault detection process. The fault detection gating coefficient is a coefficient obtained after processing the power state deviation risk coefficient through the fault detection gating channel. This coefficient affects subsequent operations on the power fault detection channel, such as determining which fault detection models need to be activated.

[0091] The fault detection gating channel receives the power state deviation risk coefficient as input and processes it using an internal machine learning algorithm to generate a corresponding fault detection gating coefficient. The power state deviation risk coefficient and the fault gating coefficient are directly proportional; the higher the power state deviation risk coefficient, the larger the fault detection gating coefficient. For example, the fault detection gating channel can be built based on a neural network. First, power state deviation risk coefficients under different power fault states are collected, and corresponding fault detection gating coefficients are assigned to these coefficients. This fault detection gating coefficient can be a value between 0 and 1. Then, the weights and biases of the neural network are initialized, and a loss function, such as mean squared error, is determined to measure the difference between the model's prediction and the actual result. An optimizer, such as Adam, is then selected to adjust the network parameters. Afterward, the preprocessed data is input into the neural network for training. During training, the data propagates forward, and the output is obtained through calculations at each layer. The error is then calculated based on the loss function, and the weights and biases are adjusted using a backpropagation algorithm. This process is iterated until the network converges and achieves good performance, ultimately resulting in the constructed neural network for the fault detection gating channel.

[0092] The power failure detection channel is a collection of models for detecting faults in outdoor portable power banks. These models are based on different machine learning algorithm architectures, such as neural networks, support vector machines, and decision trees. For ease of explanation, P is used to refer to the total number of power failure detection models in the channel. To ensure the accuracy of the fault detection results, P is usually set to a positive integer greater than or equal to 3. The construction process of these detection models is as follows: First, a large amount of power failure sample data is collected, including power state data under different fault types and fault degrees. This power failure sample data is divided into training and validation sets (e.g., randomly divided with 80% of the data as the training set and 20% as the validation set). Multiple different machine learning models are built using deep learning frameworks, including but not limited to recurrent neural networks, support vector machines, and decision trees. These models are trained using the training set to learn how to distinguish different types of power failures. Then, a validation machine is used to verify the model performance. Based on the validation results, an optimizer is used to continuously adjust the internal parameters of the model until the model converges, resulting in multiple power failure detection models.

[0093] Based on the fault detection gating coefficient, feature activation is performed on the power supply fault detection channel to obtain M activated fault detection models. The larger the fault detection gating coefficient, the more power supply fault detection models are activated within the power supply fault detection channel, and the larger the value of M. Here, M is a positive integer representing the number of activated power supply fault detection models, and 1 ≤ M ≤ P. For example, the fault detection gating coefficient can be multiplied by the total number of power supply fault detection models P, and the calculation result can be verified as the number of activated power supply fault detection models M.

[0094] Power status data is input into M fault detection activation models. Each model analyzes and processes the input data according to its own algorithm and parameters, ultimately outputting M corresponding fault detection results. The M fault detection results are summarized and analyzed to generate a power fault identification report. This report provides a comprehensive description of the outdoor portable power bank's fault status, including whether a fault exists, possible fault types, and fault severity.

[0095] The above steps improve the flexibility of computing resource allocation by matching the number of corresponding fault detection models according to the degree of power supply anomaly, and allocating different computing resources to different degrees of anomalies, thereby ensuring the accuracy of fault detection results while saving resources.

[0096] Furthermore, step P64 includes:

[0097] Step P641: Merge the M fault detection results to generate a power supply fault identification result.

[0098] Step P642: Evaluate the impact of the power supply fault based on the power supply fault identification results to obtain the power supply fault impact coefficient.

[0099] Step P643: If the power failure impact coefficient is greater than or equal to the power failure impact threshold, activate the fault propagation prediction model within the fault detection cloud node.

[0100] Step P644: Input the power fault identification result into the fault propagation prediction model to obtain the power fault propagation prediction result.

[0101] Step P645: Organize the power fault identification results and the power fault propagation prediction results to obtain the power fault identification report.

[0102] Specifically, data fusion techniques, such as voting mechanisms and weighted averaging, are used to fuse M fault detection results to generate a power supply fault identification result. This result contains detailed information on all possible faults of the outdoor portable power bank, including the location and type of the fault.

[0103] The power supply fault identification results are evaluated for fault impact, quantifying the potential impact of the fault and obtaining a power supply fault impact coefficient. A higher impact coefficient indicates a greater impact on the power system and related equipment, and vice versa. The fault impact evaluation is based on a pre-defined fault type-impact coefficient mapping table, considering the fault type and its potential severity. For example, a severe fault like a battery short circuit may be assigned a higher impact coefficient, while minor voltage fluctuation faults may be assigned a lower one. The power supply fault impact coefficient provides a quantitative basis for subsequent decisions on whether further fault propagation prediction is needed and what countermeasures to take, helping to rationally allocate resources for fault handling.

[0104] The power failure impact threshold is a pre-set value used to determine whether the fault propagation prediction model needs to be activated. When the power failure impact coefficient reaches or exceeds this threshold, it indicates a significant impact from the fault, requiring further prediction of its propagation. The power failure impact coefficient is compared with the power failure impact threshold. If the power failure impact coefficient is greater than or equal to the power failure impact threshold, the fault propagation prediction model within the fault detection cloud node is activated via network communication or a related invocation mechanism. This fault propagation prediction model is a machine learning model pre-trained using historical fault propagation data to predict the propagation path and potential impact of faults in outdoor portable power supply systems.

[0105] The power supply fault identification results are formatted according to the input requirements of the fault propagation prediction model. For example, if the model requires fault type and severity data in a specific format, the power supply fault identification results need to be parsed and transformed. Then, the processed results are input into the fault propagation prediction model. Internally, the model analyzes and calculates the input based on its own algorithms and parameters, ultimately outputting the power supply fault propagation prediction result. For example, if the power supply fault identification result is a battery short circuit fault, inputting it into the fault propagation prediction model predicts that the fault will propagate to adjacent battery modules within 10 minutes, causing a sharp drop in the voltage of these battery modules. The power supply fault propagation prediction result provides a basis for taking proactive measures to prevent further fault deterioration and helps to develop more effective fault handling strategies.

[0106] The power fault identification results and power fault propagation prediction results are organized according to a certain structure to generate a power fault identification report. For example, a dictionary structure can be created, using information such as the presence and type of fault from the power fault identification results as key-value pairs, and information such as the direction and speed of fault propagation from the power fault propagation prediction results as additional key-value pairs. For instance, if the power fault identification result indicates a battery short circuit fault, and the power fault propagation prediction result indicates that the fault will propagate to adjacent battery modules within 10 minutes, then the power fault identification report would be "A battery short circuit fault exists, and the fault will propagate to adjacent battery modules within 10 minutes."

[0107] Power failure identification reports provide maintenance personnel with comprehensive fault information, helping them to quickly and accurately formulate response strategies and improve the response speed and processing efficiency for outdoor portable power bank failures.

[0108] In summary, the remote fault identification system for outdoor mobile power supplies combining cloud and edge computing provided in this application has the following technical effects:

[0109] In this embodiment, the real-time monitoring module 10 monitors the power supply in real time and collects power supply monitoring data. Then, the first data transmission module 20 transmits the data to a status verification edge node. At the status verification edge node, the sensor distortion correction module 30 uses a sensor anomaly identification model and a sensor distortion correction model to correct the power supply monitoring data to ensure data accuracy. The status verification module 40 combines the power supply control data and the corrected power supply status data, and performs a status verification using a power supply status deviation risk assessment model and a status deviation risk verification model to assess the health status of the power supply. If the verification result indicates a problem with the power supply, the second data transmission module 50 transmits the power supply status data to a fault detection cloud node. At the fault detection cloud node, the remote fault identification module 60 remotely identifies faults in the power supply through a fault detection gating channel and a power supply fault detection channel, and generates a fault identification report, thus achieving remote fault diagnosis.

[0110] Overall, this application embodiment, through the sensing distortion correction module 30 and the cloud-edge fusion architecture, can more accurately acquire power status data and perform more comprehensive fault identification, reducing false positives and false negatives. The collaborative work of edge nodes and cloud nodes ensures the rational allocation of computing resources. Edge nodes handle simple status checks, while cloud nodes handle complex fault identification, avoiding waste of computing resources and improving the overall performance of the fault identification system. The layered inspection and identification process, along with the cloud-edge fusion architecture, reduces unnecessary data transmission and computation, enabling rapid fault identification of outdoor portable power banks, improving fault identification efficiency, facilitating timely handling of faulty power banks, and enhancing the safety and reliability of outdoor portable power banks.

[0111] The above description of the disclosed embodiments enables those skilled in the art to make or use this application. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of this application. Therefore, this application is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A remote fault identification system for outdoor mobile power banks that integrates cloud and edge computing, characterized in that: The system includes: The real-time monitoring module is used to monitor the outdoor portable power bank in real time and obtain power monitoring data. The first data transmission module is used to transmit the power monitoring data to the status verification edge node, wherein the status verification edge node includes a sensor distortion correction channel and a power status verification channel deployed on the edge computing node. The sensor distortion correction module is used to perform sensor distortion correction on the power monitoring data according to the sensor distortion correction channel to obtain power status data. The status verification module is used to obtain the power control data of the outdoor portable power bank, and to perform status verification on the outdoor portable power bank by combining the power status data and the power status verification channel to obtain the power status verification result. The second data transmission module is used to transmit the power status data to the fault detection cloud node when the power status test result is unsuccessful. The fault detection cloud node includes a fault detection gating channel and a power fault detection channel deployed on the cloud computing node. The remote fault identification module is used to remotely identify faults in the outdoor portable power bank based on the power status data, according to the fault detection gating channel and the power fault detection channel, and obtain a power fault identification report. The sensing distortion correction channel includes a sensing anomaly identification model and a sensing distortion correction model. The execution steps of the sensing distortion correction module include: Traverse the power monitoring data and extract the first power monitoring parameter; Collect sensor status information corresponding to the first power monitoring parameter to obtain first sensor status data; The first data of the sensing state is input into the sensing anomaly identification model to obtain the first sensing anomaly identification result; Based on the first sensor anomaly identification result and the first power monitoring parameter, the first power monitoring correction parameter is obtained according to the sensor distortion correction model, and the first power monitoring correction parameter is added to the power status data. The power status verification channel includes a power status deviation risk assessment model and a status deviation risk verification model. The execution steps of the status verification module include: Based on the power control data, the health status of the outdoor portable power supply is predicted to obtain the power status prediction result. The power state data and the power state prediction results are input into the power state deviation risk assessment model to obtain the power state deviation risk coefficient, wherein the power state deviation risk coefficient is a quantitative value used to represent the degree of risk brought about by the power state deviating from the normal state. The execution steps of the remote fault identification module include: Input the power state deviation risk coefficient into the fault detection gating channel to obtain the fault detection gating coefficient; Based on the fault detection gating coefficient, feature activation is performed on the power fault detection channel to obtain M fault detection activation models. The power fault detection channel includes P power fault detection models corresponding to the outdoor mobile power supply. M and P are both positive integers, and 1≤M≤P. The power status data is input into the M fault detection activation models to obtain M fault detection results; Based on the M fault detection results, a power supply fault identification report is generated.

2. The system as described in claim 1, characterized in that, The execution steps of the sensing distortion correction module include: Based on the first sensing anomaly identification result, the sensing distortion is evaluated to obtain the first sensing distortion. Determine whether the first sensing distortion is greater than or equal to a predetermined sensing distortion. If the first sensing distortion is greater than or equal to the predetermined sensing distortion, the sensing distortion correction model is activated. The first sensor anomaly identification result and the first power monitoring parameter are input into the sensor distortion correction model to obtain the first power monitoring correction parameter.

3. The system as described in claim 1, characterized in that, The power state verification channel includes a power state deviation risk assessment model and a state deviation risk verification model. The execution steps of the state verification module further include: The power state deviation risk coefficient is input into the state deviation risk test model, wherein the state deviation risk test model includes a power state deviation risk test operator, and the power state deviation risk test operator includes a power state test result of failing if the power state deviation risk coefficient is greater than or equal to the power state deviation risk threshold, and a power state test result of passing if the power state deviation risk coefficient is less than the power state deviation risk threshold. The power state deviation risk detection operator is used to output the power state detection result.

4. The system as described in claim 3, characterized in that, The execution steps of the status verification module include: Retrieve the power health status record database of the outdoor portable power bank; Based on the power control data, the power health status record library is optimized and selected according to the power control registration threshold to establish a power health status prediction space. The power state prediction result is generated by performing a central tendency evaluation based on the power health state prediction space.

5. The system as described in claim 4, characterized in that, The execution steps of the status verification module include: According to the power health status record library, extract the first power health status record group, wherein the first power health status record group includes the first power control record and the first power health status record. A similarity evaluation is performed on the power control data and the first power control record to obtain the first power control registration coefficient; Determine whether the first power control registration coefficient is greater than or equal to the power control registration threshold; If the first power control registration coefficient is greater than or equal to the power control registration threshold, the first power health status record is set as the first power health status prediction result, and the first power health status prediction result is added to the power health status prediction space. Based on the power control data and the power control registration threshold, the power health status record library is further optimized and selected to generate the power health status prediction space.

6. The system as described in claim 1, characterized in that, The execution steps of the remote fault identification module include: By integrating the M fault detection results, a power supply fault identification result is generated; Based on the power fault identification results, the fault impact is evaluated to obtain the power fault impact coefficient. If the power failure impact coefficient is greater than or equal to the power failure impact threshold, the fault propagation prediction model in the fault detection cloud node is activated. The power fault identification result is input into the fault propagation prediction model to obtain the power fault propagation prediction result; The power fault identification results and the power fault propagation prediction results are combined to obtain the power fault identification report.

Images