Intelligent switching control system for weak current equipment based on internet of things

Abstract

This invention relates to the field of low-voltage equipment control technology and discloses an intelligent switching control system for low-voltage equipment based on the Internet of Things (IoT). The system includes a data acquisition module, a collaborative sensing network module, a health prediction module, a parameter co-evolution module, an intelligent switching decision module, a switching execution module, and a pattern memory module. By exchanging state characteristics between devices and predicting health based on time-series analysis, the system triggers the switching process in the early stages of equipment performance degradation. The system extracts degradation pattern characteristics, dynamically calculates the adjustment amount of operating parameters for backup equipment, and selects the optimal backup equipment for smooth switching through multi-objective optimization. Simultaneously, switching cases are recorded to optimize model parameters. This invention realizes a shift from passive switching after a fault to predictive adaptive switching, which helps improve the continuity and reliability of low-voltage system operation.

Description

Technical Field

[0001] This invention relates to the field of low-voltage equipment control technology, and more specifically, to an intelligent switching control system for low-voltage equipment based on the Internet of Things. Background Technology

[0002] In the fields of IoT and low-voltage control technology, the reliability of low-voltage equipment (such as power distribution units and power modules) is crucial to ensuring continuous power supply to critical loads such as security and communications. Traditional and existing low-voltage equipment control systems mostly adopt fault-based switching strategies based on fixed thresholds or simple logic, which generally have the following limitations: First, the system usually only initiates switching after a clear equipment failure, posing a risk of power outage, and the switching process may trigger secondary impacts; second, backup equipment is often put into operation with fixed parameters or simple matching methods, failing to consider the performance degradation history of the failed equipment and changes in load characteristics, which may lead to mismatch or decreased efficiency after switching; finally, each device usually operates independently, lacking collaborative perception and knowledge sharing of the group's state, and the system as a whole lacks the ability to self-learn and continuously optimize.

[0003] For example, Chinese Patent Publication No. CN116054398A discloses a real-time low-voltage power distribution device based on the Internet of Things (IoT). This device monitors the status of each channel through an IoT module and reports it to the cloud, achieving centralized monitoring and data collection. However, this solution focuses on remote monitoring and alarming of the status. Its control logic still follows a "sensing-reporting-response" model, and switching decisions rely on cloud commands. This may result in delays in real-time performance and autonomy, and it does not address predictive switching based on equipment performance degradation trends or dynamic adaptation of backup equipment parameters.

[0004] For example, Chinese Patent Publication No. CN121050288A discloses an IoT-based intelligent building low-voltage control system, which uses multi-sensor fusion to determine the presence of people and thus coordinate with environmental equipment. This solution focuses on environmental perception and energy-saving control, with its core being the identification of personnel activities and equipment linkage. It does not provide solutions for the health status assessment, predictive maintenance, and intelligent switching between low-voltage equipment. The design goals and technical approaches of the two types of systems differ significantly.

[0005] Therefore, the existing technology lacks a weak current equipment control system that can make early health predictions based on the equipment's own and group status, and can adaptively complete intelligent equipment switching and parameter optimization based on the prediction results. Summary of the Invention

[0006] To overcome the aforementioned shortcomings of existing technologies, embodiments of the present invention provide an intelligent switching control system for low-voltage equipment based on the Internet of Things (IoT). This system addresses the problems of delayed fault response, non-intelligent switching processes, and lack of self-evolution capabilities in existing low-voltage equipment control systems due to a lack of predictive capabilities. Specifically, it overcomes limitations such as the inability to predict performance degradation during independent equipment operation, rigid standby equipment activation parameters that fail to adapt to attenuating load characteristics, and the lack of multi-objective trade-offs and historical experience learning in switching decisions.

[0007] To achieve the above objectives, the present invention provides the following technical solution: an intelligent switching control system for low-voltage equipment based on the Internet of Things, comprising: The data acquisition module is used to collect the operating status data of multiple low-voltage electrical devices. The operating status data includes electrical parameters and physical status parameters to provide sensing input for the system. The collaborative sensing network module is used to establish communication connections between the multiple weak current devices and exchange status characteristic information of each device, thereby constructing a status sensing network between devices; The health prediction module is used to generate a predicted health status value and corresponding confidence level for each weak current device within a future set time period based on the local operating status data obtained by the data acquisition module and the status feature information of neighboring devices obtained by the collaborative sensing network module, so as to realize the early assessment of the health trend of the devices. The parameter co-evolution module is used to generate an operating parameter adjustment scheme for the backup equipment based on the performance degradation mode characteristics of the weak electrical equipment whose status is marked as about to fail, according to the output of the health prediction module, so that the parameters of the backup equipment are adapted to the degradation characteristics of the main equipment. The intelligent switching decision module is used to generate equipment switching decision instructions through multi-objective optimization calculation based on the health status prediction value, the confidence level and the current status attributes of each backup device, so as to achieve the selection of the global optimal switching target. The switching execution module is used to control the execution of load transfer operations from the primary device to the backup device according to the device switching decision command, so as to complete the seamless switching of the power supply circuit.

[0008] Furthermore, the collaborative sensing network module is specifically configured as follows: Control each low-voltage device to periodically broadcast beacon frames. The beacon frames contain at least the device type identifier and real-time load rate information to realize device presence declaration and basic status notification. Based on the signal strength and load matching degree of the received beacon frames, the neighbor device list is dynamically constructed and updated to filter out neighboring devices with reliable communication and similar operating status. Among the devices included in the neighbor device list, encoded fixed-length state feature vectors are exchanged to achieve efficient and low-overhead sharing of detailed state information.

[0009] Furthermore, the health prediction module includes a local prediction unit and a collaborative prediction unit; The local prediction unit calculates the first health status value of a single weak current device based on its own time-series operating status data, which is used to assess the health trend of its independent operation. The collaborative prediction unit is activated when any of the following conditions are met: the first health status value is lower than a preset first threshold, more than a preset number of abnormal status reports from neighboring devices are received, or a resonant characteristic signal located in a preset frequency band is detected; after activation, by fusing local and neighboring features and performing attention calculation, a second health status value is output to characterize the risk of cascading failures in the device cluster.

[0010] Furthermore, the collaborative prediction unit performs the following operations after being activated: The state feature vectors of this device and its neighboring devices are merged to form a collaborative feature vector, thereby aggregating the state information of the local device group; Attention weights are calculated on the collaborative feature vectors to highlight key anomalous feature components and focus on common fault symptoms; Based on the key abnormal feature components, a second health status value is output to characterize the risk of cascading failures in the equipment cluster, thereby providing a risk warning caused by the environment or related equipment.

[0011] Furthermore, the parameter co-evolution module operates according to the following steps: S1. Extract the performance degradation pattern feature set of the target weak current equipment from historical operating status data to quantify its degradation law; S2. Calculate the equipment similarity between the target low-voltage equipment and the candidate backup equipment to evaluate the adaptation basis for parameter transplantation; S3. Based on the similarity between the performance degradation mode feature set and the device, calculate the adjustment amount Δ of one or more operating parameters of the candidate backup device, where |Δ|≤0.2×P, and P is the baseline value of the corresponding parameter, thereby realizing the directional and quantitative compensation of the parameter; S4. Output the parameter adjustment scheme that includes the adjustment amount Δ.

[0012] Furthermore, the set of performance degradation mode features includes: features characterizing the rate of performance degradation over time, features characterizing specific signs before a failure occurs, and features characterizing the correlation between equipment performance and load size, thus depicting the equipment degradation state from different dimensions.

[0013] Furthermore, the intelligent switching decision module performs the following process: A set of status attributes is defined for each candidate backup device, including health H, capacity margin C, energy efficiency E, and switching cost S, to characterize the status of the backup device in multiple dimensions. A comprehensive evaluation value Z is calculated for each candidate device, wherein the comprehensive evaluation value Z is the sum of the state attribute values ​​after being weighted by the corresponding dynamic weight coefficients, so as to integrate the multi-objective optimization requirements; The candidate backup device that minimizes the comprehensive evaluation value Z is selected as the target switching device, and its switching parameters are determined, thereby achieving a globally optimal decision based on multi-objective trade-offs.

[0014] Furthermore, it also includes a pattern memory module; The pattern memory module is used to record the fault mode identifier, decision parameters and switching results in historical switching events, forming a system experience base. The pattern memory module is used to trigger the update of the internal parameters of the health prediction module and the parameter co-evolution module based on the recorded switching results, so as to achieve incremental optimization of system performance. The pattern memory module is used to initiate the migration of solution parameters corresponding to the fault mode identifier when the same fault mode identifier is detected in different device clusters, thereby promoting cross-regional knowledge sharing.

[0015] Furthermore, when performing load transfer operations, the switching execution module controls the power transfer between the primary device and the backup device according to a preset load increment rate to achieve a smooth and seamless switch. After the transfer is completed, it monitors and records the electrical stability parameters of the newly formed power supply circuit to verify the switching effect. When the communication link with the primary device is continuously interrupted for more than a preset number of times or for a preset duration, an emergency switching process is triggered. Based on the real-time capacity margin attributes of each backup device, the backup device with the highest capacity margin is selected to perform a rapid load transfer as the final fault-tolerant means to ensure system continuity.

[0016] Furthermore, the data acquisition module acquires the physical state parameters at a first acquisition frequency, a second acquisition frequency, and a third acquisition frequency. The first acquisition frequency corresponds to the millisecond level and is used to capture electrical transient characteristics. The second acquisition frequency corresponds to the second level and is used to monitor the trend of state changes. The third acquisition frequency corresponds to the minute level and is used to evaluate long-term performance drift. The physical state parameters include the equipment casing temperature, vibration spectrum, and electromagnetic noise intensity in a specific frequency band, so as to comprehensively reflect the physical operating state of the equipment.

[0017] The technical effects and advantages of this invention are as follows: First, addressing the challenge of predicting device health status in advance, this invention utilizes a perception layer comprised of a data acquisition module and a collaborative sensing network module to synchronously collect and exchange multi-dimensional time-series data of the device itself with the status characteristics of neighboring devices. The health prediction module employs a lightweight neural network model to analyze local time-series characteristics and combines this with neighbor features fused through an attention mechanism to calculate a comprehensive health prediction value and confidence level, encompassing both short-term self-risk and cascading cluster risk. This process enables the system to identify degradation trends before substantial device performance failures occur, providing a proactive basis for subsequent operations. This shifts the switching action from post-failure remediation to pre-failure prevention, contributing to improved system continuity.

[0018] Secondly, addressing the issue of fixed parameter configurations for standby equipment upon activation, which cannot match the state of the primary equipment during attenuation, this invention employs a parameter co-evolution module. When the primary equipment is predicted to fail, this module quantitatively extracts features such as attenuation rate and pre-failure precursors from its historical operating data and calculates its similarity to candidate standby equipment in terms of type and capacity. Based on these features and similarities, the adjustment amounts for key parameters such as voltage and current required by the standby equipment are dynamically calculated using a preset evolution formula. This method ensures that the parameters of the standby equipment are adaptively adjusted for the specific attenuation mode of the primary equipment about to be decommissioned before it is put into operation, contributing to improved system stability and operating efficiency after load transfer.

[0019] Third, addressing the issue of simplistic handover decisions in scenarios with multiple backup devices, this invention achieves optimized selection through an intelligent handover decision module. This module constructs a multi-attribute state model for each candidate device, encompassing health, real-time capacity margin, operational energy efficiency, and estimated handover costs. By introducing a multi-objective optimization function with dynamically configurable weights, the comprehensive evaluation value of each candidate device is calculated. By solving this optimization problem, the system automatically selects the globally optimal handover target under the current system strategy, contributing to improved rationality of system resource allocation.

[0020] Fourth, addressing the issue of rigid strategies and inability to adapt to environmental changes during long-term system operation, this invention endows the system with self-learning capabilities through a pattern memory module. This module records the fault characteristics, decision parameters, and final results of each switching event, forming a case library. Clustering algorithms are used to summarize fault patterns, and an incremental learning approach is employed to continuously fine-tune key coefficients in the parameter co-evolution module based on historical success experiences under different fault modes. Simultaneously, the system supports migrating validated optimized parameters between different device clusters. This enables the system to autonomously optimize its decision logic over time, contributing to enhanced long-term adaptability and robustness across various application scenarios. Attached Figure Description

[0021] Figure 1This is a flowchart illustrating the overall system operation and data closed-loop process of the present invention.

[0022] Figure 2 This is a flowchart illustrating the collaborative sensing network construction and information exchange process of the present invention.

[0023] Figure 3 This is a branch structure diagram of the dual judgment and early warning mechanism of the health prediction module of the present invention. Detailed Implementation

[0024] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0025] Example 1 As attached Figures 1 to 3 The following is a description of an intelligent switching control system for low-voltage equipment based on the Internet of Things (IoT). The specific implementation details are as follows: I. System Overall Architecture and Operation Flow This system is applied to an IoT environment containing multiple low-voltage devices. The system operation flow is as follows: 1. Data Acquisition: Each device acquires its own electrical and physical status data at different time frequencies.

[0026] 2. Collaborative sensing: Devices periodically exchange compressed state feature summaries via a low-power wireless network.

[0027] 3. Health Prediction: Each device combines local data with neighbor features and uses a locally deployed lightweight neural network model to calculate its own health score for a future period of time. and prediction confidence .

[0028] 4. Parameter Evolution: When a device satisfies... and At the same time, the system analyzes its historical performance degradation patterns, extracts features, and calculates customized operating parameter adjustments for candidate backup equipment. .

[0029] 5. Intelligent Decision-Making: Based on the real-time status of all backup equipment, the system calculates a comprehensive evaluation value through multi-objective optimization. , choose to The device with the lowest value is selected as the switching target.

[0030] 6. Switching Execution: Based on the decision results, the control circuit uses PWM modulation to control the power switch, smoothly transferring the load from the original equipment to the standby equipment whose parameters have been pre-adjusted.

[0031] 7. Learning and Optimization: The system records the decision parameters and results of each switch and continuously optimizes the prediction and parameter evolution model through incremental learning.

[0032] Upon initial power-on or reset, each device performs a self-test. After passing the self-test, the device enters a listening state based on preset network channel parameters. If no neighbor beacon is received within a preset timeout period (e.g., 30 seconds), the device considers itself the initial root node and begins independently broadcasting its own beacon frames to gradually guide the formation of a neighbor network. At startup, all devices' gain coefficients in their mode memory modules... All initialized to default values .

[0033] II. Specific Implementation Methods of Each Module 1. Implementation of the data acquisition module This module is used to acquire raw data about the device's operation. Electrical parameters include output voltage. (Unit: V) and output current (Unit: A). Physical state parameters include: device casing temperature obtained through a temperature sensor. (unit: ) and internal hot spot temperature The principal frequency was obtained by acquiring it through a vibration sensor and analyzing it using Fast Fourier Transform (FFT). (Unit: Hz); Noise amplitude obtained by an electromagnetic probe within a center frequency of 1MHz and a bandwidth of 100kHz. (Unit: dB) V).

[0034] Data collection is performed at three fixed frequencies: (Millisecond level): Used for sampling and The instantaneous waveform.

[0035] (Second-level): Used for sampling , , , It monitors changes in trends down to the second level.

[0036] (Minute-level): Used for recording , Average over one minute , And calculate the average efficiency for that minute. ,in This refers to the input power.

[0037] 2. Implementation of the Collaborative Sensing Network Module This module enables the sharing of status information between devices. Each device is assigned a type code. (For example: (Represents 12V / 2A DC devices). Each device is cycled. A beacon frame is broadcast every second, with the frame structure: [C, r, A]. The real-time load rate... , This is the current output current. Rated current; It is a 2-byte short network address.

[0038] The neighbor discovery logic is as follows: Device Received equipment After receiving the beacon frame, read its signal strength. .like And the load rate difference between the two devices Then join in Neighbor list middle.

[0039] State feature vector Generation and exchange: Each device will collect data within the last 60 seconds. The mean and variance of the four parameter sequences are calculated to obtain eight eigenvalues. These eight eigenvalues ​​are normalized to the interval [0, 255] and arranged sequentially to form a fixed-length state feature vector of 8 bytes (64 bits). Each piece of equipment The second will be its own vector Unicast to list All neighboring devices.

[0040] Device unicast state feature vector At this time, a reliable transmission mode with Automatic Repeat Request (ARQ) is used. After successfully receiving the vector, the receiver returns a short acknowledgment frame. If the sender does not receive an acknowledgment within a specified time, a limited number of retransmissions are performed. Meanwhile, the vector... A Cyclic Redundancy Check (CRC) code is embedded in the data, which the receiver uses to verify data integrity and discard erroneous data.

[0041] 3. Implementation of the health prediction module This module contains two prediction units that run on the device's microcontroller. The neural network model... and Supervised training was completed using historical operational data before system deployment. The training dataset consists of historical normal operation data, known fault precursor data, and synthetic data generated through data augmentation (such as adding noise), and is divided into training, validation, and test sets. The training objective is to minimize the loss between the predicted results and the actual health status.

[0042] Model and The specific structure is fixed after training and stored in the device memory in a lightweight format.

[0043] 3.1 Local Prediction Unit: The input to this unit is the device's own data sequence over the past 10 minutes, including... , , , First, the mean of each parameter over 10 consecutive 1-minute sub-windows is calculated, forming a feature set containing 40 features. .

[0044] Will Input a lightweight fully connected neural network model . It has a three-layer structure: an input layer with 40 nodes, a hidden layer with 16 nodes (using the ReLU activation function), and an output layer with 1 node (using the Sigmoid activation function). Output a scalar That is, the first health status value. .

[0045] 3.2 Collaborative Prediction Unit: This unit is activated when any of the following conditions are met: Condition one: ,in .

[0046] Condition 2: In two consecutive During the period, received more than Neighboring devices report via beacon frames The alarm.

[0047] Condition 3: Local analysis revealed The value remained within the preset resonant frequency band for 10 seconds. Inside.

[0048] After activation, the unit performs the following steps: Fusion: Combines the current feature vectors of this device. With all vectors received from neighbors By concatenating the features, we obtain the extended feature vector. .

[0049] Attention calculation: A dot product attention mechanism is used. First, ... Generate query vector through linear transformation Key vector Sum value vector .

[0050] Calculate the attention score matrix ,in Let be the dimension of the key vector. Finally, calculate the weighted key anomaly feature components. .

[0051] Prediction: Will Input another neural network model with the same structure Output scalar That is, the second health status value. .

[0052] Finally, the module outputs a predicted health status value. and confidence level : . The calculation method is as follows ,in and The models are respectively and Output and The corresponding internal confidence probability. Health status fault threshold (0.25), confidence threshold (0.8), first threshold. Key parameters such as (0.3) are determined during the model training phase by adjusting them on the validation set to maximize the fault detection rate while controlling the false alarm rate to an acceptable level (e.g., below 5%). These thresholds can be fine-tuned in the management interface after system deployment based on actual operational statistics.

[0053] 4. Implementation of the parameter co-evolution module When a certain device satisfy and This module starts when [the specified time is specified].

[0054] S1. Feature Extraction: From Three features are extracted from historical data from the past 24 hours. First, the time series data is standardized.

[0055] (Decay rate characteristics): Calculation In the last 4 hours (total) Average efficiency (minutes) Regarding time ( The linear regression slope of ( ). The least squares method is used: .

[0056] (Characteristics of early signs of failure): Statistics Output voltage in the last 30 minutes The instantaneous sampled value is lower than the rated value. of The number of times.

[0057] (Load correlation characteristics): Calculate the current load rate Below, equipment Internal temperature rise The same as in the past week Historical average temperature rise in the region Relative deviation: .

[0058] S2. Calculate similarity: for candidate backup devices Calculate device similarity . ,in . For type similarity, if ,but Otherwise . For power capacity similarity, , This is the rated power.

[0059] S3. Calculate the adjustment amount: to adjust Output voltage setting value For example, its benchmark value for The rated output voltage. Adjustment amount. The calculation formula is: .in, This is a configurable gain coefficient with an initial default value. This calculation must satisfy the constraints. .

[0060] S4. Output Scheme: Module output structured parameter adjustment scheme: {Target: Backup: ;Adjustment }; 5. Implementation of the intelligent switching decision module When it is necessary to repair faulty equipment When selecting a backup device, this module considers all candidate devices. An assessment will be conducted.

[0061] First, define each The four state attributes: Health : Taken from its health prediction value .

[0062] capacity margin : ,in This represents the current.

[0063] Energy efficiency : ,in This is the rated peak efficiency of the equipment.

[0064] Switching Cost : .in, To estimate the switchover time, To estimate the energy consumption during the switching process. and As a normalized baseline value, and As weight.

[0065] Next, calculate each Comprehensive evaluation value : ; in, , , , The dynamic weighting coefficients satisfy the following conditions: The system can preset multiple strategy modes, corresponding to different combinations of coefficients.

[0066] The final decision is: choose to make The device with the smallest value As the target switching device. Switching parameter set. Include The identity identifier and all adjustment schemes calculated by the parameter co-evolution module.

[0067] When the main controller fails to communicate with a power supply control unit that is predicted to be healthy multiple times (e.g., 3 times), regardless of its predicted health value... In any case, the system will mark the device as 'communication unavailable' and trigger an emergency handover process for the load it supplies. The emergency handover will directly select the current capacity margin. The highest standby equipment performs a rapid switchover (e.g., using a higher load increment rate). Once the faulty device is isolated, the system will generate an alarm in the maintenance log.

[0068] 6. Implementation of switching execution modules This module is used to perform physical handover operations.

[0069] Received contains After receiving the instruction, the module first sends a message to the backup device. Send parameter configuration instructions. The output voltage is set via its digital-to-analog converter (DAC). .

[0070] The load transfer process is as follows: Master equipment and backup equipment The output terminal is connected to a MOSFET switch. and Connect the load after parallel connection.

[0071] The controller generates two complementary PWM signals with linearly adjustable duty cycles, which drive the controllers respectively. and Set the load increment rate. .exist Within seconds, The PWM duty cycle linearly decreases from 100% to 0%, while The duty cycle increases linearly from 0% to 100%.

[0072] After the transfer is completed, the module monitors the new circuit in the following... Peak voltage fluctuation within seconds .like If the switch is successful, the electrical stability parameters are recorded. and actual switching time .

[0073] 7. Implementation of the Pattern Memory Module This module runs on the regional gateway and is used for continuous system optimization.

[0074] Record: Generate a record after each event switch. .

[0075] (Fault Mode Identifier): The standardized feature vector extracted from this event. Perform K-means clustering analysis. Before performing clustering analysis, analyze the feature vectors. Standardization is performed. The number of clusters is preset. This value is an empirical value derived from historical fault data through analysis combining the profile coefficient and the elbow rule. When system records are insufficient in the initial stages of operation, all events are temporarily categorized into a default category; cluster analysis is then performed once sufficient records are available. After clustering, the feature vectors are assigned to the nearest cluster, and the cluster number is [the value of the cluster]. .

[0076] (Decision parameter vector): .

[0077] (Result Vector): .

[0078] Update: For each failure mode When the accumulated number of records reaches When a condition is met, a model update is triggered. Stochastic gradient descent is used, with a learning rate of [missing information]. Update the gain coefficient specific to this fault mode. .

[0079] Migration: The gateway periodically compares the failure mode characteristics of different physical regions (e.g., region A, region B). For each cluster center in region A... Calculate its characteristics compared to all new events in region B. Euclidean distance If it exists If so, then region B is determined to have the same pattern as region A. Similar failures. The system automatically adjusted the coefficients of region A. Synchronize to the configuration library in region B.

[0080] The above description is merely a specific embodiment of the present invention. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this invention should be included within the protection scope of this invention.

Claims

1. A smart switching control system for low-voltage equipment based on the Internet of Things, characterized in that, include: The data acquisition module is used to collect operating status data of multiple low-voltage electrical devices, including electrical parameters and physical status parameters. The collaborative sensing network module is used to establish communication connections between the multiple low-voltage devices and exchange status characteristic information of each device; The health prediction module is used to generate a predicted health status value and corresponding confidence level for each weak current device within a future set time period based on the local operating status data obtained by the data acquisition module and the neighbor device status feature information obtained by the collaborative sensing network module. The parameter co-evolution module is used to generate an operating parameter adjustment scheme for backup equipment based on the performance degradation mode characteristics of the weak electrical equipment whose status is marked as about to fail, as output by the health prediction module. The intelligent switching decision module is used to generate a device switching decision instruction through multi-objective optimization calculation based on the health status prediction value, the confidence level, and the current status attributes of each backup device. The switching execution module is used to control the execution of load transfer operations from the primary device to the backup device according to the device switching decision command.

2. The intelligent switching control system for low-voltage equipment based on the Internet of Things according to claim 1, characterized in that, The collaborative sensing network module is specifically configured as follows: Control each low-voltage electrical device to periodically broadcast beacon frames, wherein the beacon frames contain at least the device type identifier and real-time load rate information; The neighbor device list is dynamically constructed and updated based on the signal strength and load matching degree of the received beacon frames; Among the devices included in the neighbor device list, encoded fixed-length state feature vectors are exchanged.

3. The intelligent switching control system for low-voltage equipment based on the Internet of Things according to claim 1 or 2, characterized in that, The health prediction module includes a local prediction unit and a collaborative prediction unit; The local prediction unit calculates the first health status value of a single weak current device based on its own time-series operating status data. The collaborative prediction unit is activated when any of the following conditions are met: the first health status value is lower than a preset first threshold, more than a preset number of abnormal status reports from neighboring devices are received, or a resonant characteristic signal located in a preset frequency band is detected.

4. The intelligent switching control system for low-voltage equipment based on the Internet of Things according to claim 3, characterized in that, Once activated, the collaborative prediction unit performs the following operations: The state feature vectors of this device and neighboring devices are merged to form a collaborative feature vector; Attention weights are calculated on the collaborative feature vector to highlight key anomalous feature components; Based on the key abnormal feature components, a second health status value is output to characterize the risk of cascading failures in the equipment cluster.

5. The intelligent switching control system for low-voltage equipment based on the Internet of Things according to claim 1, characterized in that, The parameter co-evolution module operates according to the following steps: S1. Extract the performance degradation mode feature set of the target low-voltage equipment from historical operating status data; S2. Calculate the equipment similarity between the target low-voltage equipment and the candidate backup equipment; S3. Based on the similarity between the performance degradation mode feature set and the device, calculate the adjustment amount Δ of one or more operating parameters of the candidate backup device, where |Δ|≤0.2×P, and P is the baseline value of the corresponding parameter; S4. Output the parameter adjustment scheme that includes the adjustment amount Δ.

6. The intelligent switching control system for low-voltage equipment based on the Internet of Things according to claim 5, characterized in that, The set of performance degradation mode features includes: features characterizing the rate of performance decline over time, features characterizing specific signs before a failure occurs, and features characterizing the correlation between equipment performance and load size.

7. The intelligent switching control system for low-voltage equipment based on the Internet of Things according to claim 1, characterized in that, The intelligent switching decision module performs the following process: Define a set of status attributes for each candidate backup device, including health H, capacity margin C, energy efficiency E, and switching cost S; A comprehensive evaluation value Z is calculated for each candidate device, wherein the comprehensive evaluation value Z is the sum of the state attribute values ​​after being weighted by the corresponding dynamic weight coefficients; Select the candidate backup device that minimizes the comprehensive evaluation value Z as the target switching device, and determine its switching parameters.

8. The intelligent switching control system for low-voltage equipment based on the Internet of Things according to claim 1, characterized in that, It also includes a pattern memory module; The pattern memory module is used to record the fault mode identifier, decision parameters, and switching results in historical switching events; The pattern memory module is used to trigger the updating of the internal parameters of the health prediction module and the parameter co-evolution module based on the recorded switching results; The pattern memory module is used to initiate the migration of solution parameters corresponding to the fault mode identifier when the same fault mode identifier is detected in different device clusters.

9. The intelligent switching control system for low-voltage equipment based on the Internet of Things according to claim 1 or 7, characterized in that, When performing load transfer operations, the switching execution module controls the power transfer between the master device and the backup device according to a preset load increment rate. After the transfer is completed, it monitors and records the electrical stability parameters of the newly formed power supply circuit. When the communication link with the master device is continuously interrupted for more than a preset number of times or for a preset duration, an emergency switching process is triggered. Based on the real-time capacity margin attributes of each backup device, the backup device with the highest capacity margin is selected to perform a fast load transfer.

10. The intelligent switching control system for low-voltage equipment based on the Internet of Things according to claim 1, characterized in that, The data acquisition module acquires the physical state parameters at a first acquisition frequency, a second acquisition frequency, and a third acquisition frequency. The first acquisition frequency corresponds to the millisecond level, the second acquisition frequency corresponds to the second level, and the third acquisition frequency corresponds to the minute level. The physical state parameters include the device casing temperature, vibration spectrum, and electromagnetic noise intensity in a specific frequency band.

Images