Low-voltage internet-of-things switch based on cloud platform and edge computing

Abstract

The application provides a low-voltage internet-of-things switch based on a cloud platform and edge computing, which is connected with a cloud master station and provided with an edge computing module, is connected with a plurality of distributed photovoltaic inverters and / or electric vehicle charging piles to perform centralized management, and identifies different equipment types to determine parameters to be collected and match corresponding control strategies.

Description

Technical Field

[0001] This invention relates to the technical fields of intelligent power management and control, embedded operating system design, Internet of Things (IoT) technology, intelligent power equipment, and system security and stability, and particularly to a low-voltage IoT switch based on cloud platform and edge computing. Background Technology

[0002] With the rapid development of Internet of Things (IoT) technology, the demand in the fields of smart homes and industrial automation is constantly growing. In low-voltage power systems, the need for intelligent, remote control, and data analysis is increasing, and traditional low-voltage switchgear can no longer meet the requirements of modern power systems for efficient, reliable, and intelligent operation. Summary of the Invention

[0003] Considering the rapid development of Internet of Things (IoT) technology, traditional low-voltage switchgear can no longer meet the requirements of modern power systems for efficiency, reliability, and intelligence. Therefore, to address the gaps and shortcomings of existing technologies, this invention proposes a low-voltage IoT switchgear based on a cloud platform and edge computing. It integrates IoT technology, embedded intelligent control modules, wireless communication functions, and cloud master station services, enabling real-time data collection and monitoring of equipment. Through data analysis via edge computing and the cloud master station, it achieves precise optimized management of power equipment.

[0004] Its innovation lies primarily in the integrated application of edge computing and a cloud master station. Through edge computing, the IoT switch can monitor device status in real time, process local data, and make intelligent decisions and optimize control without relying on a cloud master station, improving system response speed and autonomy. Simultaneously, the cloud master station provides remote data management and policy distribution functions, enabling precise control and optimization management of devices, achieving cloud-based data analysis and centralized management of large-scale devices. This combination of edge computing and a cloud master station significantly enhances the system's intelligence, ease of operation, and overall performance, ensuring efficient and stable operation of the equipment in complex power environments.

[0005] The present invention specifically adopts the following technical solution:

[0006] A low-voltage IoT switch based on cloud platform and edge computing: connected to a cloud master station and equipped with an edge computing module, connecting several distributed photovoltaic inverters and / or electric vehicle charging piles for centralized management; by identifying different device types, it determines the parameters that need to be collected and matches the corresponding control strategies.

[0007] Furthermore, the cloud master station sends control commands to the IoT switch according to a preset strategy. After receiving the commands, the IoT switch performs corresponding operations through the edge computing module's adaptive edge computing, including adjusting the power output of the photovoltaic inverter or dynamically adjusting the charging rate of the charging pile. The adaptive edge computing automatically adjusts the control strategy based on real-time data feedback of the device's operating status.

[0008] Furthermore, the edge computing module applies a dual-path recursive prediction and detection algorithm based on the characteristics of the access device to monitor and predict the operating status of the device in real time, so as to determine the changing trends of parameters including voltage and current, and detect whether there are any abnormalities.

[0009] Furthermore, the dual-path recursive prediction and detection algorithm includes:

[0010] A-path utilizes recursive feature matching and time series prediction with neural network models; by comparing historical data with current data, it uses edge computing to predict the future state of the device; in order to monitor whether the current, voltage or power characteristics of the device are at risk of exceeding preset thresholds.

[0011] B-path uses a periodic detection method, performing feature analysis after every 2n rounds of data collection; edge computing analyzes whether the device is operating normally through periodic real-time comparison.

[0012] Furthermore, the recursive feature matching mechanism achieves an iterative process through continuous updating and optimization of historical data; the recursive function `recurrence_match(current_features, historical_features)` is used to compare the currently input feature values ​​with the historical feature set; each time new data is input, this recursive function is called to determine the similarity between the current data and historical data, update the historical records, and return the matching result.

[0013] Maintain a structure to store historical feature data and use indexes to mark important features; when new feature data is input, the recursive function recurrence_match() starts working; the function evaluates the degree of matching between the current data and historical data according to different weights; if the match is successful, proceed to further processing; if the match is not successful, the recursive process ends.

[0014] After each recursion, the historical data structure is updated to store the latest feature values. The recursion terminates when the current feature matches a historical feature. If no match is found after all historical features have been traversed, the recursion terminates.

[0015] Furthermore, the calculation process of the recursive feature matching includes:

[0016] Feature distance calculation: During the recursive process, the difference between the current feature value and the historical feature values ​​is calculated; after calculating the similarity between two feature vectors, a match is determined if the distance value is less than the set threshold; otherwise, the recursion continues.

[0017] Weighting: By defining a weight matrix W, the influence of each feature is adjusted so that the matching of high-weight features is given priority during recursion.

[0018] Furthermore, in the dual-path composite judgment, the judgment results of path A and path B are assigned the same or different weights for comprehensive judgment;

[0019] During the composite judgment process, the occurrence time and number of each anomaly are recorded; if the detection results of channel A and / or channel B trigger an anomaly multiple times, the number of anomalies is accumulated to form an anomaly score, and an alarm is triggered when the score exceeds a set threshold.

[0020] Furthermore, when a device malfunction is detected by the dual-path recursive prediction detection algorithm, the IoT switch, based on the adaptive control mechanism, first attempts to automatically adjust the device's operating parameters locally to resolve the issue. The adaptive control dynamically adjusts parameters, including power output or charging rate, according to the actual state and characteristics of the device to ensure optimal operation under different environments. If local adjustment is ineffective, the IoT switch triggers an alarm and uploads the abnormal data to the cloud master station.

[0021] Furthermore, by performing high-frequency sampling on the access devices, feature information including current, voltage, power, zero-crossing length, and higher harmonics is extracted. After preprocessing, a recurrent neural network model is trained on the cloud master station, and the trained model is integrated into the TensorFlow Lite Micro model set in the IoT switch MCU; so as to use the dual-path recursive prediction detection algorithm for real-time data monitoring and anomaly prediction.

[0022] Furthermore, the IoT switch communicates with the cloud master station via an Ethernet communication module through a secure communication protocol, establishes a secure communication connection with the charging pile and photovoltaic inverter through an RS485 communication module or a CAN communication module, and performs authentication through an encryption protocol to ensure the legitimate access of the devices.

[0023] Compared to existing technologies, this invention and its preferred embodiment can connect multiple distributed photovoltaic inverters and electric vehicle charging piles, and realize edge computing and remote control functions through a cloud master station. It connects to distributed power equipment through standardized interfaces (such as RS485, CAN, Ethernet, etc.) to achieve intelligent centralized management.

[0024] Through edge computing, IoT switches can monitor device status in real time, process local data, and make intelligent decisions and optimize control without relying on a cloud master station, improving system response speed and autonomy. Meanwhile, the cloud master station provides remote data management and policy distribution functions, enabling precise control and optimization management of devices, and achieving cloud-based data analysis and centralized management of large-scale devices. This combination of edge computing and the cloud master station significantly improves the system's intelligence, ease of operation, and overall performance, ensuring efficient and stable operation of equipment in complex power environments. Attached Figure Description

[0025] The present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments:

[0026] Figure 1 This is a scenario architecture diagram of Embodiment 1 of the present invention.

[0027] Figure 2 This is a basic module diagram of Embodiment 2 of the present invention.

[0028] Figure 3 This is a schematic diagram illustrating the implementation of edge computing functionality in Embodiment 3 of the present invention. Detailed Implementation

[0029] In the following, specific embodiments of this application will be described in detail with reference to the accompanying drawings. Based on these detailed descriptions, those skilled in the art will be able to clearly understand and implement this application. Without departing from the principles of this application, features from various embodiments can be combined to obtain new implementations, or certain features from some embodiments can be substituted to obtain other preferred implementations.

[0030] To make the features and advantages of this patent more apparent and understandable, specific embodiments are provided below, along with accompanying drawings, for detailed explanation:

[0031] This invention proposes a new generation of low-voltage, high-performance IoT switch operating system and cloud master station strategy to implement edge computing design. It can connect multiple distributed photovoltaic inverters and electric vehicle charging piles, and realize edge computing and remote control functions through the cloud master station. It connects to distributed power equipment through standardized interfaces (such as RS485, CAN, Ethernet, etc.) to achieve intelligent centralized management.

[0032] Based on the tiered operating system, it has been deeply optimized to meet the low power consumption requirements of IoT switches, including regional memory allocation, multi-task scheduling and device driver model, to ensure that the system runs efficiently in a multi-task environment.

[0033] In terms of data communication, the IoT switch communicates with the cloud master station via secure communication protocols (such as TLS-encrypted MQTT, CoAP, etc.), supporting the issuance of policy commands from the cloud. Edge computing processes and analyzes data in real time on the device, enabling dynamic adjustments to the device's operating status based on locally processed data, such as adjusting the power output of photovoltaic inverters or controlling the charging rate of charging piles, thereby optimizing the utilization of power resources. Simultaneously, when the cloud master station is unavailable, the embedded module can continue to execute basic control functions according to local rules, ensuring stable system operation.

[0034] 1. System Architecture:

[0035] Hardware layer: Employs an embedded high-performance microcontroller (RISC-V series) that supports multi-tasking, features a multi-core architecture, and integrates an NPU (Neural Processing Unit). This allows for simultaneous control of multiple photovoltaic inverters and charging piles, ensuring stable system operation under high loads. The NPU provides dedicated hardware acceleration for executing machine learning tasks, enhancing system performance when processing complex data and performing intelligent analysis.

[0036] Operating System Layer: The operating system layer is a new generation of low-voltage, high-performance operating system, ensuring efficient real-time response and multi-task scheduling, and supporting edge computing and remote control functions in complex scenarios. The operating system integrates the TensorFlow Lite Micro package to support machine learning task processing and inference, optimizing edge computing capabilities.

[0037] Middleware layer: mainly realizes the integration of communication protocol unification, data processing and device management services, and ensures efficient interaction between IoT switches and multiple distributed photovoltaic inverters, charging piles and cloud master stations.

[0038] Application layer: Provides user interaction and intelligent control, with main functions including remote control, real-time monitoring, data acquisition and analysis, and equipment management and identification.

[0039] It enables centralized management of multiple distributed photovoltaic inverters and electric vehicle charging piles through standardized interfaces (such as RS485, CAN, Ethernet), supports plug-and-play functionality, and ensures system scalability and compatibility by automatically identifying and configuring newly connected devices.

[0040] The IoT switch automatically obtains the device ID and status information by scanning and identifying different photovoltaic inverters and charging piles. Next, it categorizes the devices according to their type (e.g., photovoltaic inverter or charging pile) to determine the key parameters that need to be collected. Subsequently, by identifying the device brand, the IoT switch matches the corresponding control algorithm.

[0041] 2. IoT switch operation procedure:

[0042] Device Startup and Initialization: Upon startup, the IoT switch performs self-tests on the communication module, power measurement module, and storage device, completing hardware initialization. The embedded operating system loads task scheduling, device drivers, and the communication protocol stack based on the kernel. The IoT switch establishes a secure communication connection with the charging pile and photovoltaic inverter via Bluetooth or RS485, and performs authentication through encryption protocols to ensure legitimate device access.

[0043] Device identification and data acquisition: Devices are identified and data is synchronized via the RS485 protocol. Parameters such as the power generation and output voltage of the photovoltaic inverter, as well as information such as the charging status and charging current of the charging pile, are collected. This data is stored and cached locally for real-time processing by edge computing and transmission to the cloud master station.

[0044] Edge Computing and Status Monitoring: The IoT switch first identifies the type and brand of the connected power equipment and applies a dual-path recursive predictive detection algorithm based on equipment characteristics to monitor and predict the equipment's operating status in real time. This algorithm allows the system to accurately determine the changing trends of key parameters such as voltage and current, and detect any anomalies. The IoT switch then executes corresponding optimization algorithms based on the equipment characteristics. For example, the system applies customized algorithms to adjust the power output or charging rate of photovoltaic inverters or charging piles of different brands to ensure optimal operating efficiency. Simultaneously, the system introduces a dual-path recursive predictive detection algorithm for precise monitoring and predictive analysis of the equipment's operating status. This algorithm can detect changing trends of parameters such as voltage and current during equipment operation and predict potential anomalies in advance. When anomalies such as excessively high voltage or current overload are detected, edge computing, combined with the prediction results, automatically selects appropriate protection measures based on the characteristics of different equipment brands. This algorithm not only improves the accuracy of status monitoring but also reduces reliance on the cloud master station, ensuring the system can quickly respond to anomalies locally, achieving real-time optimization and protection.

[0045] Anomaly Handling and Alarm Mechanism: When an anomaly is detected by the dual-path recursive predictive detection algorithm, the IoT switch will, based on its adaptive control mechanism, first attempt to automatically adjust the device's operating parameters locally to resolve the issue. Adaptive control dynamically adjusts key parameters such as power output or charging rate according to the device's actual state and characteristics, ensuring optimal operation under different environments. If local adjustments are ineffective, the IoT switch will trigger an alarm and upload detailed anomaly data to the cloud master station. Users will receive remote notifications via mobile apps and other means, enabling remote intervention. The IoT switch also features automatic recovery and fault handling functions. Through adaptive adjustment and support from the cloud master station or edge computing, it attempts to restore the device to normal operation, records the cause of the fault and the handling process, ensuring system stability and reliability.

[0046] Data storage and historical records: Critical equipment data, such as power parameters and alarm information, is stored locally for easy maintenance and fault analysis. The cloud main station stores uploaded data long-term, allowing users to view historical records and perform energy efficiency assessments and optimizations via the cloud.

[0047] Equipment maintenance and upgrades: Remote upgrades are supported through the cloud main station, and technicians can also perform local debugging and parameter settings via Bluetooth to ensure continuous and efficient operation of the equipment. Configuration updates and optimizations are also performed in conjunction with the analysis results of the cloud main station and edge computing.

[0048] 3. Communication modules supported by IoT switches:

[0049] Ethernet communication module: Used to connect the IoT switch to the cloud master station to achieve real-time data transmission and remote control. It updates the device status and power data in real time on the cloud master station and receives remote control commands, supporting remote monitoring and management of the device, such as adjusting the inverter's operating mode or modifying the charging pile's charging strategy.

[0050] RS485 communication module: Used for multi-point serial communication in industrial environments to exchange data with photovoltaic inverters and charging pile equipment. It supports long-distance data transmission and multiple device connections, and is suitable for communication scenarios requiring high stability and reliability.

[0051] CAN communication module: Used to connect and communicate multiple devices, especially in automotive or complex industrial control system environments. It ensures high-speed data transmission between IoT switches and devices such as photovoltaic inverters and charging piles, making it suitable for control systems requiring high real-time performance and stability, and guaranteeing reliable data transmission.

[0052] 4. Implementation of edge computing:

[0053] The cloud master station sends control commands to the IoT switch according to a preset strategy. Upon receiving the commands, the IoT switch executes corresponding operations through adaptive edge computing, such as adjusting the power output of the photovoltaic inverter or dynamically adjusting the charging rate of the charging pile. The IoT switch employs an adaptive control algorithm, automatically adjusting the control strategy based on real-time data feedback of the device's operating status. Through the edge computing platform, the system can monitor key parameters in real time, such as voltage, current, and power, and dynamically adjust the device's operating mode according to actual needs. For example, the photovoltaic inverter automatically optimizes its power output under different weather conditions, while the charging pile adaptively adjusts the charging rate based on the vehicle's battery status, ensuring efficient and safe energy management. Simultaneously, the IoT switch uploads the collected operating data to the cloud master station via a communication module for remote monitoring and data analysis, enabling intelligent control and adaptive management of the equipment.

[0054] Preliminary preparations include high-frequency sampling during device charging to extract features such as current, voltage, power, zero-crossing length, and higher harmonics. These features are then processed and a recurrent neural network (RNN) model is trained on the cloud master station. The trained model is then integrated into the TensorFlow Lite Micro model on the MCU. The Ladder OS supports TensorFlow Lite Micro, enabling the model to run computations on edge devices. Finally, a dual-path recursive prediction and detection algorithm is used for real-time data monitoring and anomaly prediction.

[0055] The system periodically scans connected devices through a standardized communication interface. Once a sub-device is discovered, the system establishes a connection with it using the device's communication protocol (Modbus, DL / T 645), reads the device's unique ID (MAC address), and identifies the device type and brand.

[0056] The system will adapt the corresponding management strategies and control algorithms based on the type, brand and protocol of the sub-devices and add the devices to the device management list.

[0057] The overall detection process employs a multi-task balancing mechanism that combines monitoring with adaptive control using a dual-path recursive prediction detection algorithm. Each path is responsible for a specific detection task, and the feature data from both paths are combined for a final judgment to determine whether to issue a warning or alarm.

[0058] Recursive Feature Matching Mechanism: The core of the recursive feature matching mechanism lies in the iterative process of feature matching through continuous updating and optimization of historical data. The recursive function `recurrence_match(current_features, historical_features)` compares the currently input feature values ​​with the historical feature set. Each time new data is input, the system calls this recursive function to determine the similarity between the current data and historical data, updates the historical records, and returns the matching result.

[0059] First, the system maintains a structure to store historical feature data and uses indexes to mark important features. When new feature data such as current, voltage, and power are input, the recursive function `recurrence_match()` begins its work. The function evaluates the degree of match between the current data and historical data based on different weights (such as the importance of historical features). If a match is successful, further processing begins; otherwise, the recursive process ends.

[0060] After each recursion, the system updates the historical data structure and stores the latest feature values ​​in the appropriate location. The recursion terminates when the current feature is sufficiently close to a historical feature, indicating a successful match; otherwise, the recursion terminates if no match is found after traversing all historical features.

[0061] The calculation process of recursive feature matching:

[0062] Feature distance calculation: During the recursive process, the system calculates the difference between the current feature value and historical feature values. The similarity between two feature vectors can be calculated using methods such as Euclidean distance and cosine similarity. The calculation formula is as follows: If the distance value is less than the set threshold, it is considered a match; otherwise, the recursion continues.

[0063] Weighting: Different features may have different levels of importance. Current and voltage have a greater weight than power and harmonics. A weight matrix W can be defined to adjust the influence of each feature. In this way, the matching of high-weight features will be given priority during recursion.

[0064] Dual-path recursive prediction and detection process:

[0065] A-Route Process

[0066] Recursive / Memory Cursor: Define a data structure to store historical data (current, voltage, power, etc.), use timestamps or indices to mark important feature locations, and point the cursor to these feature values. When new data is input, the cursor is used to determine its matching degree with historical features.

[0067] Current feature analysis: If the currently monitored feature value meets the value condition (>n / m), then further processing is performed; otherwise, the path is exited.

[0068] RNN (LSTM) models predict potential abnormal electricity usage behavior by collecting historical current, voltage, and power data, combined with device-specific electricity usage patterns, as a training set. The model is trained on time-series data using either a Recurrent Neural Network (RNN) or a Long Short-Term Memory (LSTM) network, enabling it to learn and capture patterns in device usage behavior. When new data is input into the model, it can predict and identify abnormal signals such as illegal electricity use in real time.

[0069] System Feature Analysis 1: Continue to monitor feature data, compare historical records, and determine whether feature values ​​have experienced abnormal fluctuations.

[0070] Anomaly detection: If an anomaly is detected by comparing historical records, the situation is recorded and the device status is updated.

[0071] Alarm conditions: If an anomaly is detected multiple times, the system will accumulate the number of anomalies and ultimately decide whether to trigger an alarm based on the alarm conditions.

[0072] Route B Process

[0073] Anomalies are analyzed every 2n rounds: Path B uses different strategies and defines a round counter. Every 2n data collections, a large-scale feature analysis is triggered.

[0074] Actual detection value analysis: The system determines whether the current detection value exceeds the safety threshold. If no anomalies are found, the path ends; if the detection value exceeds the threshold, the system continues processing.

[0075] System Feature Analysis 2: Similar to path A, the system enters the feature analysis process, using the ISTM model to perform complex time series prediction.

[0076] Data cleaning and supplementary analysis: The system will clean up noise and anomalies in the data and perform further analysis and verification.

[0077] Exception recording and review: Similar to route A, exceptions on route B will be recorded and reviewed to ensure accuracy.

[0078] Composite Feature Screening

[0079] Dual-path analysis: Path A is high-frequency feature detection, focusing on real-time data monitoring and anomaly prediction; Path B is based on periodic analysis, performing large-scale feature analysis after every 2n rounds of sampling. The combination of the two, through data analysis methods of different frequencies complementing each other, ensures accuracy.

[0080] Weighting Mechanism: In dual-path composite judgment, the judgment results of path A and path B can be assigned different weights. At the final decision level of the system, the high-frequency data detected in real time by path A and the results of periodic detection by path B are combined and analyzed using the following weighted formula: Comprehensive Judgment = WA⋅Path A Result + WB⋅Path B Result.

[0081] Anomaly accumulation mechanism: During the composite judgment process, the system records the occurrence time and number of each anomaly. If the detection results of channel A or channel B trigger an anomaly multiple times, the system will accumulate these anomaly counts to form an anomaly score. When the score exceeds a set threshold, an alarm will be triggered.

[0082] Time series analysis and multi-layer detection: Through RNN or LSTM models, the system can perform predictive analysis on time series data to capture changes in device usage behavior. The multi-layer screening mechanism in both paths ensures the accuracy of the detection results, making it particularly suitable for appliance identification and anomaly detection in complex environments.

[0083] If an anomaly is detected and alarm conditions are met, or if no anomaly is found after several rounds of analysis, the detection path is exited. In the dual-path analysis process, path A mainly provides rapid feedback on real-time anomalies, while path B combines in-depth analysis with historical data. The predictive capability of the LSTM model is integrated into both paths, and the model continuously optimizes the accuracy of anomaly prediction by recursively updating historical data.

[0084] Adaptive control applications:

[0085] Application of A-channel results: When an anomaly detected by A-channel requires adjustment, the system automatically optimizes equipment parameters based on A-channel's adaptive control mechanism to ensure optimal equipment operation under dynamic loads and environmental changes. Adaptive control adjusts equipment parameters based on A-channel's real-time detection results to address specific anomalies.

[0086] Application of B-channel results: Anomaly detection results from the B-channel also trigger adaptive control, optimizing the equipment through periodic, large-scale analysis of the results. The system adjusts the equipment's operating state based on the B-channel analysis results to avoid faults or overloads, thereby improving the overall stability and performance of the equipment.

[0087] System Optimization: Adaptive control dynamically adjusts equipment parameters to ensure optimal system performance under varying loads and environmental conditions. It responds in real-time to changes in equipment operating status, reducing the probability of failures and optimizing long-term stability and efficiency. Under a multi-task framework, adaptive control enables self-balancing of optimization schemes and control commands for multiple electrical devices, ensuring optimal allocation and use of computing power and enhancing the system's intelligent control capabilities.

[0088] Data Upload and Monitoring: The adjusted equipment status and optimization results will be uploaded to the cloud main station via the communication module for remote monitoring and further data analysis. The system will record the adjustment and optimization process to ensure system stability and reliability, achieving intelligent equipment management and efficient operation and maintenance.

[0089] The following three specific application examples will provide a more detailed introduction and demonstration of the above-mentioned solution of the present invention:

[0090] Example 1: Figure 1 The diagram shows the system architecture, which includes a hardware layer, an operating system layer, a middleware layer, and an application layer. The hardware layer includes ARM Cortex-M architecture chips, and the operating system layer is a tiered operating system.

[0091] The IoT switch establishes a connection with the electrical equipment via the CAN bus, enabling one-to-many device scanning. The format is JSON, and the type of the item is specified by the type field.

[0092] The IoT switch sends a request to each sub-device to read device information via a standard protocol, obtaining the device's unique ID, model, brand, and other identifiers. Different brands of photovoltaic inverters may use different data frame structures, and the IoT switch will automatically switch to the corresponding protocol parsing module based on the device type.

[0093] The identified brands and models are matched with corresponding control algorithms and management strategies.

[0094] Real-time data from each device is uploaded to the cloud main site in JSON format to enable real-time monitoring.

[0095] Example 2: The strategy of connecting IoT switches to the cloud master station of multiple distributed photovoltaic inverter devices enables edge computing functionality, such as... Figure 2 As shown.

[0096] The equipment connects to the photovoltaic inverter via RS485 or CAN bus using the Modbus protocol to establish a communication connection.

[0097] Read device information: The IoT switch sends an identification request to each connected photovoltaic inverter to obtain the device's unique ID, brand, and model number, ensuring successful device identification.

[0098] Device Identification: The IoT switch sends a device identification frame to the connected photovoltaic (PV) inverter via RS485 or CAN bus. This frame contains a device query command, requesting information such as the device ID, brand, and model. Upon receiving the request, the PV inverter returns a response frame containing its unique ID, brand identifier, and device model. The response frame also includes the device's initial status data. The IoT switch parses the returned response frame and, based on the device ID and brand identifier, determines the specific model and manufacturer of the PV inverter.

[0099] Algorithm selection: The IoT switch identifies the brand and invokes the corresponding scheduling strategy. For example, for an inverter from brand A, a Dynamic Maximum Power Point Tracking (MPPT) algorithm might be used to adjust the inverter's output power in real time. For an inverter from brand B, a voltage limiting algorithm might be needed to prevent overload operation.

[0100] Protocol conversion: The core of protocol conversion is to unify the parsing of the raw data from inverters of different brands into a standard format and to convert the control commands of edge computing back to their respective protocols.

[0101] The IoT switch collects real-time data from connected devices (such as voltage, current, and power parameters of photovoltaic inverters). This data is first processed and stored locally via edge computing. Then, the encrypted data is sent to the cloud master station via common IoT communication protocols (such as MQTT, CoAP, and HTTP). The cloud master station sends encrypted control commands to the IoT switch via MQTT, CoAP, or HTTP protocols. The IoT switch then uses edge computing to convert the commands to the corresponding brand's protocol based on the brand identified by the device before issuing them to the inverter device.

[0102] Example 3: The strategy of connecting IoT switches to the cloud master station of multiple charging pile devices enables edge computing functionality, such as... Figure 3 As shown:

[0103] Device Access and Identification: The IoT switch communicates with multiple charging pile devices through a standardized communication interface. The system periodically scans connected devices, identifies their type and brand by reading their device IDs, and adapts to the corresponding communication protocols. The IoT switch assigns specific control policies based on the identified device type, adds the devices to the management list, and manages them uniformly.

[0104] Cloud Master Station Policy Distribution: The IoT switch establishes a connection with the cloud master station via Ethernet. The cloud master station distributes control policies, such as power allocation, priority settings, or charging strategies. After parsing the cloud policies, the IoT switch executes local control, including adjusting the charging pile's power and allocating loads.

[0105] Edge computing: When dealing with multiple charging pile devices, the IoT switch combines a dual-path (A-path and B-path) data analysis framework and uses local edge computing technology to achieve real-time monitoring and anomaly handling of device operation status, reduce dependence on the cloud, and ensure that the devices operate efficiently within a safe range.

[0106] A-path process: The A-path focuses on a time series prediction system using recursive feature matching and RNN (LSTM) models. By comparing historical data with current data, edge computing is used to predict the future state of devices. Once the current, voltage, or power characteristics of a device are detected to exceed a preset threshold, the IoT switch will immediately start an analysis model at the edge layer to assess whether there is abnormal power consumption behavior and respond quickly by triggering a local alarm mechanism.

[0107] B-path Process: The B-path employs a periodic detection method, performing in-depth feature analysis after every 2n rounds of data collection. Edge computing analyzes the normal operating status of the equipment through periodic real-time comparisons. Simultaneously, the B-path possesses powerful data cleaning and verification capabilities, ensuring more accurate data processing after each detection. If an anomaly is detected, the system records it locally and continuously tracks it during subsequent verification processes to ensure stable equipment operation.

[0108] The above description is a preferred embodiment of the present invention. It should be noted that, for those skilled in the art, several improvements and modifications can be made without departing from the principle of the present invention, and these improvements and modifications are also considered to be within the scope of protection of the present invention.

[0109] This patent is not limited to the above-described preferred embodiment. Anyone can derive other forms of low-voltage IoT switches based on cloud platforms and edge computing under the guidance of this patent. All equivalent changes and modifications made within the scope of this patent application shall fall within the scope of this patent.

Claims

1. A low-voltage IoT switch based on cloud platform and edge computing, characterized in that: It connects to the cloud master station and is equipped with an edge computing module, connecting several distributed photovoltaic inverters and / or electric vehicle charging piles for centralized management; by identifying different device types, it determines the parameters that need to be collected and matches the corresponding control strategies. The edge computing module applies a dual-path recursive prediction and detection algorithm based on the characteristics of the access device to monitor and predict the operating status of the device in real time, so as to determine the changing trends of parameters including voltage and current, and detect whether there are any abnormalities. The dual-path recursive prediction and detection algorithm includes: A-path utilizes recursive feature matching and time series prediction with neural network models; by comparing historical data with current data, it uses edge computing to predict the future state of the device; in order to monitor whether the current, voltage or power characteristics of the device are at risk of exceeding preset thresholds. B-path uses a periodic detection method, performing feature analysis after every 2n rounds of data collection; edge computing analyzes whether the equipment is operating normally through periodic real-time comparison; The recursive feature matching mechanism achieves an iterative process through continuous updating and optimization of historical data. The recursive function `recurrence_match(current_features, historical_features)` compares the currently input feature values ​​with the historical feature set. Each time new data is input, this recursive function is called to determine the similarity between the current data and historical data, update the historical records, and return the matching result. Maintain a structure to store historical feature data and use indexes to mark important features; when new feature data is input, the recursive function recurrence_match() starts working; the function evaluates the degree of matching between the current data and historical data according to different weights; if the match is successful, proceed to further processing; if the match is not successful, the recursive process ends. After each recursion, the historical data structure is updated to store the latest feature values. The recursion terminates when the current feature matches a historical feature. If no match is found after all historical features have been traversed, the recursion terminates. The calculation process for the recursive feature matching includes: Feature distance calculation: During the recursive process, the difference between the current feature value and the historical feature values ​​is calculated; after calculating the similarity between two feature vectors, a match is determined if the distance value is less than the set threshold; otherwise, the recursion continues. Weighting: By defining a weight matrix W, the influence of each feature is adjusted so that the matching of high-weight features is given priority during recursion.

2. The low-voltage IoT switch based on cloud platform and edge computing according to claim 1, characterized in that: The cloud master station sends control commands to the IoT switch according to a preset strategy. After receiving the commands, the IoT switch performs corresponding operations through the edge computing module, including adjusting the power output of the photovoltaic inverter or dynamically adjusting the charging rate of the charging pile. The adaptive edge computing automatically adjusts the control strategy based on real-time data feedback of the device's operating status.

3. A low-voltage IoT switch based on cloud platform and edge computing according to claim 1, characterized in that: In the dual-path composite judgment, the judgment results of path A and path B are assigned the same or different weights for comprehensive judgment. During the composite judgment process, the occurrence time and number of each anomaly are recorded; if the detection results of channel A and / or channel B trigger an anomaly multiple times, the number of anomalies is accumulated to form an anomaly score, and an alarm is triggered when the score exceeds a set threshold.

4. A low-voltage IoT switch based on cloud platform and edge computing according to claim 1, characterized in that: When a device malfunction is detected by the dual-path recursive prediction detection algorithm, the IoT switch, based on an adaptive control mechanism, first attempts to automatically adjust the device's operating parameters locally to resolve the issue. The adaptive control dynamically adjusts parameters, including power output or charging rate, according to the actual state and characteristics of the device to ensure optimal operation under different environments. If local adjustment is ineffective, the IoT switch triggers an alarm and uploads the abnormal data to the cloud master station.

5. A low-voltage IoT switch based on cloud platform and edge computing according to claim 1, characterized in that: By sampling the access devices at high frequency, feature information including current, voltage, power, zero-crossing length and higher harmonics is extracted. After preprocessing, a recurrent neural network model is trained on the cloud master station, and the trained model is integrated into the TensorFlow Lite Micro model set in the IoT switch MCU. A dual-path recursive prediction detection algorithm is used for real-time data monitoring and anomaly prediction.

6. A low-voltage IoT switch based on cloud platform and edge computing according to claim 1, characterized in that: The IoT switch communicates with the cloud master station via an Ethernet communication module through a secure communication protocol. It establishes a secure communication connection with the charging pile and photovoltaic inverter through an RS485 communication module or a CAN communication module, and performs authentication through an encryption protocol to ensure the legitimate access of the devices.

Images