A terminal-independent off-grid internet of things collaborative management method

Abstract

This invention discloses a terminal-independent off-grid IoT collaborative management method, belonging to the field of off-grid IoT technology. This method collects and standardizes terminal energy consumption characteristics, removes parameters related to terminal business functions, constructs a terminal-independent energy consumption node model, and establishes a unified energy consumption management interface. Using this model, energy is allocated through a hierarchical energy storage architecture. Multi-dimensional data is collected to construct a lightweight energy prediction model to match the system's operating mode. Communication parameters and update strategies are adjusted based on the system's operating mode, channel environment, and terminal energy status, and differential updates are performed. The operating level of the near-field interaction module is adjusted to complete offline interaction, caches and transmits behavioral data, forming a collaborative closed loop of energy, communication, and interaction. This method achieves terminal-independent off-grid collaborative management, improves system terminal adaptability, energy utilization rationality, communication stability, and interaction reliability, and provides support for the large-scale deployment of off-grid IoT scenarios.

Description

Technical Field

[0001] This invention relates to the field of off-grid Internet of Things (IoT) technology, and in particular to a terminal-independent off-grid IoT collaborative management method. Background Technology

[0002] Off-grid IoT, as an important branch of the IoT technology system, is the core technological support for realizing digital management and intelligent services in areas without public network coverage and power grid access. Currently, it has been widely deployed in various scenarios such as ecological monitoring in nature reserves, infrastructure management in remote rural areas, smart agricultural production management, outdoor sports event services, and emergency security monitoring. With the continuous increase in the demand for digitalization in outdoor scenarios, various low-power IoT terminals are widely deployed in various remote, network-free areas. Terminal types cover multiple categories such as display terminals, sensing terminals, control terminals, and data acquisition terminals. The continuous development of low-power wide-area communication technology, renewable energy power supply technology, near-field interaction technology, and edge intelligent computing technology provides a technological foundation for the deployment and operation of off-grid IoT systems. The industry as a whole is developing towards large-scale terminal deployment, integrated functionality, and long-term stable operation. Related technical solutions are also continuously adapting to the operational needs of various complex outdoor scenarios, constantly enriching the functional dimensions and application boundaries of off-grid IoT systems.

[0003] Currently, off-grid IoT systems still face multiple technical limitations in actual deployment and operation. Existing systems have deeply integrated energy management logic with terminal business functions, requiring customized development for different types of terminals. This hinders the decoupling of energy management from terminal business, making it difficult to form a universal and standardized energy management framework adaptable to various terminal types. Terminal access and system expansion processes are also complex. Furthermore, existing energy storage architectures and energy scheduling modes struggle to adapt to differentiated energy supply needs, failing to proactively optimize energy scheduling based on terminal energy consumption characteristics and environmental changes, thus hindering effective energy utilization efficiency. Remote update schemes cannot balance communication power consumption and transmission reliability, making it difficult to dynamically adjust transmission strategies based on system operating status and channel environment. In large-scale terminal deployment scenarios, channel congestion and insufficient update success rates are common. Human-machine interaction schemes heavily rely on public networks, making stable human-machine interaction and effective data collection impossible in areas without network coverage. Moreover, the energy supply, remote communication, and human-machine interaction modules within the system are mostly independently designed, lacking deep collaborative mechanisms and failing to form a complete operation and management closed loop. Consequently, the overall environmental adaptability and operational stability of the system cannot be fully guaranteed. Summary of the Invention

[0004] The purpose of this invention is to overcome the shortcomings of the prior art and provide a terminal-independent off-grid Internet of Things collaborative management method.

[0005] The objective of this invention is achieved through the following technical solution: A terminal-independent off-grid IoT collaborative management method is provided, which includes the following steps: S1. Collect energy consumption characteristics of terminals accessing the system, standardize the collected energy consumption data, remove the service function-related parameters of the terminals, construct a terminal-independent energy consumption node model, and build a unified energy consumption management interface. S2. Adopting a terminal-independent energy consumption node model, energy distribution is performed through a three-level energy storage architecture. Multi-dimensional input data is collected to construct a lightweight gradient-enhanced energy prediction model. Energy state data is output through the lightweight gradient-enhanced energy prediction model, and the energy state data is used to match the corresponding system operating mode. S3. Adopt the system working mode, adjust the communication parameters in combination with the collected channel environment data, use the terminal energy status data to adjust and update the strategy, and perform differential update operation; S4. The working level of the near-field interaction module is adjusted by adopting the system working mode. The near-field interaction module completes offline interaction operations, caches and transmits the behavioral data generated by the interaction, and completes the collaborative closed-loop management of energy, communication and interaction.

[0006] Furthermore, step S1 includes the following sub-steps: S1.1. Collect three types of energy consumption characteristic data from the terminals accessing the system. The three types of energy consumption characteristic data include standby power consumption data, transient power consumption data, and daily average energy consumption data. S1.2. Normalize the collected energy consumption data, unify the statistical period, measurement unit and data acquisition node of the energy consumption data, and convert the original energy consumption data of different terminals into standard unit data under the set rated voltage. S1.3. Remove three types of parameters from the terminal: business function type, business execution logic, and business output format, and retain the standardized energy consumption data; S1.4. Construct a terminal-independent energy consumption node model using the retained standardized energy consumption data. The terminal-independent energy consumption node model includes a data input layer, a standardization verification layer, and a model output layer. The data input layer is connected to the standardization verification layer, and the standardization verification layer is connected to the model output layer. Establish a unified energy consumption management interface, which only transmits standardized energy consumption data.

[0007] Furthermore, step S2 includes the following sub-steps: S2.1. Adopting a terminal-independent energy consumption node model, assigning corresponding energy supply scenarios and execution priorities to each energy storage unit of the three-level energy storage architecture; S2.2. Collect historical load data, light prediction data, temperature data, and seasonal factor data as input data for the lightweight gradient boosting energy prediction model; S2.3. Construct a lightweight gradient boosting energy prediction model. The lightweight gradient boosting energy prediction model includes an input feature layer, an iterative training layer composed of multiple decision trees, a result fusion layer, and an output layer. The input feature layer is connected to the iterative training layer, the iterative training layer is connected to the result fusion layer, and the result fusion layer is connected to the output layer. The iterative training layer uses a set number of decision trees for iterative fitting, and sets a set learning rate and tree depth parameters to complete the model training. The lightweight gradient boosting energy prediction model outputs energy state curve data and remaining battery life data within a set future time period. S2.4. Use energy state curve data, remaining range data and real-time status data of energy storage units to match the corresponding system operating mode.

[0008] Furthermore, step S3 includes the following sub-steps: S3.1. Adopt the system working mode and select the corresponding terminal wake-up level. The terminal wake-up levels include periodic polling wake-up level, specific matching wake-up level and broadcast emergency wake-up level. S3.2. Collect channel environment data, including rainfall data, signal-to-noise ratio data, and received signal strength data. Use the channel environment data to predict the channel quality level, and use the channel quality level to adjust communication parameters, including spreading factor, packet size, and retransmission strategy. S3.3. Collect real-time energy status data from the terminal, and use the real-time energy status data from the terminal to adjust and update the strategy. The adjustment logic of the update strategy refers to the standardized energy consumption data of the energy consumption node model that is independent of the terminal. S3.4. Using the adjusted communication parameters and update strategy, the difference between the old and new versions of data is extracted using a differential algorithm to generate a differential update package and execute the differential update operation.

[0009] Furthermore, step S4 includes the following sub-steps: S4.1. The system working mode is adopted to match the corresponding working level to the near-field interaction module. The system working mode is derived from the output data of the lightweight gradient boosting energy prediction model. S4.2. The interaction identification data is stored through the near-field interaction module. The interaction identification data includes scene identification data, encryption token data and language tag data, to complete the offline data interaction operation with the user terminal. S4.3. Locally cache the behavioral data generated by offline data interaction operations. The behavioral data includes interaction timestamp data, terminal identification data, and user behavior feature data. S4.4. According to the set period, the cached behavioral data is batch transmitted back through a low-power self-organizing network, and the transmitted behavioral data is synchronized to the input feature layer of the lightweight gradient boosting energy prediction model.

[0010] Furthermore, in step S1.4, the unified energy management interface shields non-standardized data and business function-related data. Newly connected terminals access the unified energy management interface by reporting standardized energy consumption data. The unified energy management interface synchronizes the terminal-independent energy consumption node model of the connected terminal to the input feature layer of the lightweight gradient boosting energy prediction model. The unified energy management interface collects real-time energy consumption data of the connected terminal according to a set period, updates the terminal-independent energy consumption node model, and synchronizes the updated terminal-independent energy consumption node model to the energy allocation logic of the three-level energy storage architecture.

[0011] Furthermore, in step S2.1, the three-level energy storage architecture includes a primary energy storage unit, a secondary energy storage unit, and a tertiary energy storage unit. The primary energy storage unit corresponds to transient power consumption power supply scenarios, the secondary energy storage unit corresponds to daily load power supply scenarios, and the tertiary energy storage unit corresponds to backup energy reserve scenarios. The execution priority of the primary energy storage unit is higher than that of the secondary energy storage unit, and the execution priority of the secondary energy storage unit is higher than that of the tertiary energy storage unit. The energy allocation logic of the three-level energy storage architecture uses a terminal-independent energy consumption node model for parameter calibration. The calibrated energy allocation logic matches the power consumption level requirements of the system operating mode. The system operating mode is derived from the output data of the lightweight gradient boosting energy prediction model.

[0012] Furthermore, in step S3.3, the real-time energy status data of the terminal is used to divide the energy into multiple energy levels. Different energy levels correspond to different update permissions, update package specifications, and transmission strategies. The corresponding update strategy is matched with the energy level of the terminal. The adjustment logic of the update strategy is bound to the power consumption level of the system working mode. The bound update strategy is synchronized to the differential update package generation logic. The differential update package generation logic uses a differential algorithm to extract the differences between the old and new versions of data and generate a differential update package of the corresponding specification. The system working mode comes from the output data of the lightweight gradient boosting energy prediction model.

[0013] Furthermore, in step S4.2, the user terminal pre-downloads the multilingual content package corresponding to the scenario to the local storage area under a network environment. The user terminal reads the interaction identifier data through the near-field interaction module. The user terminal retrieves the content package corresponding to the interaction identifier data from the local storage area to complete the offline interaction operation. The offline interaction operation does not rely on the public network connection throughout. The behavior data generated by the offline interaction operation is synchronized to the storage area of ​​the near-field interaction module. The data in the storage area of ​​the near-field interaction module is synchronized to the local cache area of ​​the terminal. The cached behavior data is synchronized to the input feature layer of the lightweight gradient boosting energy prediction model.

[0014] Furthermore, in step S4, the collaborative closed-loop management of energy, communication, and interaction includes: adjusting the operating parameters of the communication module and the working level of the near-field interaction module using the terminal's energy status data; updating the input features of the lightweight gradient boosting energy prediction model using behavioral data generated by offline interaction; calibrating the output parameters of the lightweight gradient boosting energy prediction model using actual energy consumption data from differential update operations; updating the system working mode using the output data of the lightweight gradient boosting energy prediction model; adjusting the energy allocation logic of the three-level energy storage architecture, the communication parameters of the communication module, and the working level of the near-field interaction module using the updated system working mode; and updating the terminal-independent energy consumption node model using the updated energy consumption data.

[0015] The beneficial effects of this invention are: (1) By collecting and standardizing the energy consumption characteristics of the terminal, the energy consumption node model independent of the terminal is constructed by stripping the business-related parameters, and a unified energy consumption management interface is built. Combined with three-level energy storage energy allocation, energy prediction, communication adjustment and offline interaction and collaborative closed-loop management, the off-grid Internet of Things system can be adapted to different types of terminals, reducing the terminal customization development process and improving the system terminal adaptation and expansion capabilities. (2) Adopt the terminal-independent energy consumption node model to calibrate the energy distribution logic of the three-level energy storage architecture, combine the energy state data output by the lightweight energy prediction model to match the system working mode, optimize the system energy supply mode, adapt to the differentiated energy consumption needs of different scenarios, improve the rationality of energy utilization, and extend the continuous operation time of the system when there is no external power supply. (3) Combine the channel environment and terminal energy status to adjust communication parameters and update strategies, reduce transmission energy consumption through differential updates, and build a collaborative closed loop with near-field interaction offline management and behavior data backhaul optimization prediction model to improve system communication stability and interaction reliability, providing feasible support for large-scale deployment of off-grid IoT scenarios. Attached Figure Description

[0016] Figure 1 A flowchart illustrating the steps of a terminal-independent off-grid IoT collaborative management method; Figure 2 The flowchart illustrates the specific steps of a terminal-independent off-grid IoT collaborative management method provided in this embodiment. Detailed Implementation

[0017] The technical solution of the present invention will be clearly and completely described below with reference to the embodiments. 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.

[0018]

Example

[0019] In some embodiments, step S1 includes the following sub-steps: S1.1. Collect three types of energy consumption characteristic data from the terminals accessing the system. The three types of energy consumption characteristic data include standby power consumption data, transient power consumption data, and daily average energy consumption data. S1.2. Normalize the collected energy consumption data, unify the statistical period, measurement unit and data acquisition node of the energy consumption data, and convert the original energy consumption data of different terminals into standard unit data under the set rated voltage. S1.3. Remove three types of parameters from the terminal: business function type, business execution logic, and business output format, and retain the standardized energy consumption data; S1.4. Construct a terminal-independent energy consumption node model using the retained standardized energy consumption data. The terminal-independent energy consumption node model includes a data input layer, a standardization verification layer, and a model output layer. The data input layer is connected to the standardization verification layer, and the standardization verification layer is connected to the model output layer. Establish a unified energy consumption management interface, which only transmits standardized energy consumption data.

[0020] In some embodiments, step S2 includes the following sub-steps: S2.1. Adopting a terminal-independent energy consumption node model, assigning corresponding energy supply scenarios and execution priorities to each energy storage unit of the three-level energy storage architecture; S2.2. Collect historical load data, light prediction data, temperature data, and seasonal factor data as input data for the lightweight gradient boosting energy prediction model; S2.3. Construct a lightweight gradient boosting energy prediction model. The lightweight gradient boosting energy prediction model includes an input feature layer, an iterative training layer composed of multiple decision trees, a result fusion layer, and an output layer. The input feature layer is connected to the iterative training layer, the iterative training layer is connected to the result fusion layer, and the result fusion layer is connected to the output layer. The iterative training layer uses a set number of decision trees for iterative fitting, and sets a set learning rate and tree depth parameters to complete the model training. The lightweight gradient boosting energy prediction model outputs energy state curve data and remaining battery life data within a set future time period. S2.4. Use energy state curve data, remaining range data and real-time status data of energy storage units to match the corresponding system operating mode.

[0021] In some embodiments, step S3 includes the following sub-steps: S3.1. Adopt the system working mode and select the corresponding terminal wake-up level. The terminal wake-up levels include periodic polling wake-up level, specific matching wake-up level and broadcast emergency wake-up level. S3.2. Collect channel environment data, including rainfall data, signal-to-noise ratio data, and received signal strength data. Use the channel environment data to predict the channel quality level, and use the channel quality level to adjust communication parameters, including spreading factor, packet size, and retransmission strategy. S3.3. Collect real-time energy status data from the terminal, and use the real-time energy status data from the terminal to adjust and update the strategy. The adjustment logic of the update strategy refers to the standardized energy consumption data of the energy consumption node model that is independent of the terminal. S3.4. Using the adjusted communication parameters and update strategy, the difference between the old and new versions of data is extracted using a differential algorithm to generate a differential update package and execute the differential update operation.

[0022] In some embodiments, step S4 includes the following sub-steps: S4.1. The system working mode is adopted to match the corresponding working level to the near-field interaction module. The system working mode is derived from the output data of the lightweight gradient boosting energy prediction model. S4.2. The interaction identification data is stored through the near-field interaction module. The interaction identification data includes scene identification data, encryption token data and language tag data, to complete the offline data interaction operation with the user terminal. S4.3. Locally cache the behavioral data generated by offline data interaction operations. The behavioral data includes interaction timestamp data, terminal identification data, and user behavior feature data. S4.4. According to the set period, the cached behavioral data is batch transmitted back through a low-power self-organizing network, and the transmitted behavioral data is synchronized to the input feature layer of the lightweight gradient boosting energy prediction model.

[0023] In some embodiments, in step S1.4, the unified energy management interface shields non-standardized data and business function-related data. Newly connected terminals access the unified energy management interface by reporting standardized energy consumption data. The unified energy management interface synchronizes the terminal-independent energy consumption node model of the connected terminal to the input feature layer of the lightweight gradient boosting energy prediction model. The unified energy management interface collects real-time energy consumption data of the connected terminal according to a set period, updates the terminal-independent energy consumption node model, and synchronizes the updated terminal-independent energy consumption node model to the energy allocation logic of the three-level energy storage architecture.

[0024] In some embodiments, in step S2.1, the three-level energy storage architecture includes a primary energy storage unit, a secondary energy storage unit, and a tertiary energy storage unit. The primary energy storage unit corresponds to transient power consumption power supply scenarios, the secondary energy storage unit corresponds to daily load power supply scenarios, and the tertiary energy storage unit corresponds to backup energy reserve scenarios. The execution priority of the primary energy storage unit is higher than that of the secondary energy storage unit, and the execution priority of the secondary energy storage unit is higher than that of the tertiary energy storage unit. The energy allocation logic of the three-level energy storage architecture uses a terminal-independent energy consumption node model for parameter calibration. The calibrated energy allocation logic matches the power consumption level requirements of the system operating mode. The system operating mode is derived from the output data of the lightweight gradient boosting energy prediction model.

[0025] In some embodiments, in step S3.3, the real-time energy status data of the terminal is used to divide the energy into multiple energy levels. Different energy levels correspond to different update permissions, update package specifications and transmission strategies. The corresponding energy level of the terminal is matched with the corresponding update strategy. The adjustment logic of the update strategy is bound to the power consumption level of the system working mode. The bound update strategy is synchronized to the differential update package generation logic. The differential update package generation logic uses a differential algorithm to extract the differences between the old and new versions of data and generate a differential update package of the corresponding specification. The system working mode comes from the output data of the lightweight gradient boosting energy prediction model.

[0026] In some embodiments, in step S4.2, the user terminal pre-downloads the multilingual content package corresponding to the scenario to the local storage area in a network environment. The user terminal reads the interaction identifier data through the near-field interaction module. The user terminal retrieves the content package corresponding to the interaction identifier data from the local storage area to complete the offline interaction operation. The offline interaction operation does not rely on the public network connection throughout. The behavior data generated by the offline interaction operation is synchronized to the storage area of ​​the near-field interaction module. The data in the storage area of ​​the near-field interaction module is synchronized to the local cache area of ​​the terminal. The cached behavior data is synchronized to the input feature layer of the lightweight gradient boosting energy prediction model.

[0027] In some embodiments, in step S4, completing the coordinated closed-loop management of energy, communication, and interaction includes: adjusting the operating parameters of the communication module and the working level of the near-field interaction module using the terminal's energy status data; updating the input features of the lightweight gradient boosting energy prediction model using behavioral data generated by offline interaction; calibrating the output parameters of the lightweight gradient boosting energy prediction model using actual energy consumption data from differential update operations; updating the system working mode using the output data of the lightweight gradient boosting energy prediction model; adjusting the energy allocation logic of the three-level energy storage architecture, the communication parameters of the communication module, and the working level of the near-field interaction module using the updated system working mode; and updating the terminal-independent energy consumption node model using the updated energy consumption data.

[0028] Example 2 This embodiment provides a specific implementation process for a terminal-independent off-grid IoT collaborative management method. Addressing the terminal adaptation, energy supply, communication updates, and interactive control needs of off-grid IoT systems, this embodiment constructs a terminal-independent system operation framework through operations such as standardized energy consumption data processing, hierarchical energy storage scheduling, environmental adaptive communication adjustment, and offline interactive data management. This enables the collaborative operation of energy, communication, and interactive modules, improving the adaptability and operational stability of the off-grid IoT system. Figure 2 As shown, the specific implementation process is as follows: Step 1. Standardized modeling and unified interface construction for terminal energy consumption: Step 1.1. Collect three types of energy consumption characteristic data from the terminals connected to the system. These three types of energy consumption characteristic data include standby power consumption data, transient power consumption data, and daily average energy consumption data. Energy consumption characteristic data are the basic data characterizing the energy consumption status of the terminal. Standby power consumption data is the energy consumption data when the terminal is not performing any services. Transient power consumption data is the short-term energy consumption data when the terminal performs sudden operations. Daily average energy consumption data is the cumulative energy consumption data of the terminal within a set period. The collection operation is completed through preset data collection nodes. The collection nodes are the system's preset energy consumption data acquisition ports. The collection nodes collect the current and voltage parameters in the terminal's power supply circuit in real time, and generate the corresponding energy consumption data through integral calculation. This ensures that the format and collection dimensions of the collected data are consistent, providing a unified raw data foundation for subsequent data processing. Step 1.2. Normalize the collected energy consumption data. Normalization involves converting energy consumption data of different specifications and acquisition methods into a unified standard format, standardizing the statistical period, unit of measurement, and data acquisition nodes of the energy consumption data, and converting the raw energy consumption data of different terminals into standard unit data under a set rated voltage. Abnormal fluctuations in the collected data are first removed, and then energy consumption data at different voltage levels are converted into unified standard values ​​using a linear conversion formula. This eliminates data format deviations caused by differences in terminal hardware types and business models, allowing energy consumption data from different terminals to be compared and processed under the same standard, providing standardized data support for subsequent modeling operations. Step 1.3. Remove three types of parameters from the terminal: business function type, business execution logic, and business output format. Business function type refers to the type of business the terminal undertakes; business execution logic refers to the operational process by which the terminal completes the business; and business output format refers to the way the terminal presents the business results. These three types of parameters are not directly related to energy consumption management. Using data filtering rules, remove the fields corresponding to these parameters, retaining only the standardized energy consumption data. Separate the association between energy consumption data and terminal business attributes, retaining only the data content related to energy consumption management, so that the energy consumption data only serves the system's energy scheduling logic. Step 1.4. Construct a terminal-independent energy consumption node model using the retained standardized energy consumption data. This model stores only standardized energy consumption data and does not contain any terminal service-related information. The model includes a data input layer, a standardization verification layer, and a model output layer. The data input layer connects to the standardization verification layer, which in turn connects to the model output layer. The data input layer receives the standardized energy consumption data, sorts it according to fixed fields, and transmits it to the standardization verification layer. The standardization verification layer checks the standard format of the energy consumption data, verifying whether the data conforms to the preset numerical range and field length, and removes non-standard data items. The model output layer outputs fixed-length hexadecimal format model data, establishing a unified energy consumption management interface. The energy management interface is a transmission interface that only transmits standardized energy consumption data. The unified energy management interface uses a custom serial transmission protocol to transmit only standardized energy consumption data. It masks non-standardized data and data related to business functions. Newly connected terminals access the unified energy management interface by reporting standardized energy consumption data. The unified energy management interface synchronizes the terminal-independent energy consumption node model of the connected terminal to the input feature layer of the lightweight gradient-enhanced energy prediction model. The unified energy management interface collects real-time energy consumption data from the connected terminals according to a set period, updates the terminal-independent energy consumption node model, and synchronizes the updated model to the energy allocation logic of the three-tier energy storage architecture, providing real-time updated energy consumption data for energy allocation. In some specific implementations, the standardized verification layer of the terminal-independent energy consumption node model sets fixed verification rules. The verification range for standby power consumption data is set to microampere to milliampere level, the verification range for transient power consumption data is set to milliampere to hundreds of milliampere level, and the verification range for daily average energy consumption data is set to milliampere-hour to ampere-hour level. Data exceeding the range is directly discarded. The model data length output by the model output layer is fixed at sixteen bytes. The transmission rate of the unified energy consumption management interface is set to kilobits per second, and the transmission cycle is set to hours. When a new terminal accesses the system, it only needs to upload sixteen bytes of standardized energy consumption data to complete the adaptation. This implementation method solves the problems of inconsistent energy consumption data formats and complex access processes for different terminals. By fixing the verification rules and transmission format, it simplifies the terminal access operation and ensures the validity and transmission efficiency of energy consumption data.

[0029] In some embodiments, the terminal-independent energy consumption node model can use a fixed-format text structure to store data. The text structure only contains numerical items of standby power consumption data, transient power consumption data, and daily average energy consumption data, without adding any business-related comments or additional fields, which further simplifies the model structure, reduces the system data processing load, and improves the efficiency of model calling and transmission.

[0030] Step 2. Tiered energy storage energy allocation and energy prediction model scheduling: Step 2.1. Employing a terminal-independent energy consumption node model, assign corresponding energy supply scenarios and execution priorities to each energy storage unit of the three-tier energy storage architecture. The three-tier energy storage architecture comprises three levels of energy storage units to adapt to different types of energy supply demands. The three-tier architecture includes primary, secondary, and tertiary energy storage units. Primary energy storage units correspond to transient power consumption scenarios, meeting the short-term sudden energy consumption demands of terminals. Secondary energy storage units correspond to daily load scenarios, meeting the energy consumption demands of terminals during normal operation. Tertiary energy storage units correspond to backup energy reserve scenarios, used for other... When the energy storage unit is insufficient, it provides supplementary energy. The execution priority of the first-level energy storage unit is higher than that of the second-level energy storage unit, and the execution priority of the second-level energy storage unit is higher than that of the third-level energy storage unit. The system controls the output path of each level of energy storage unit through electronic switches, prioritizing the first-level energy storage unit path, then the second-level energy storage unit path, and finally the third-level energy storage unit path. The energy distribution logic of the three-level energy storage architecture uses a terminal-independent energy consumption node model for parameter calibration. The calibrated energy distribution logic matches the power consumption level requirements of the system's operating mode, ensuring that the energy supply matches the actual energy consumption demand of the terminal. Step 2.2. Collect historical load data, illumination prediction data, temperature data, and seasonal factor data as input data for the lightweight gradient boosting energy prediction model. Historical load data refers to the actual operating energy consumption data of the terminal in previous continuous cycles. Illumination prediction data refers to the predicted data of ambient illumination conditions. Temperature data refers to the real-time temperature data of the terminal's operating environment. Seasonal factor data refers to the parameter data characterizing the impact of seasonal changes on energy supply. The acquisition module obtains temperature data and illumination prediction data through environmental sensors, reads historical load data through the storage unit, and generates seasonal factor data through the system clock. Multiple types of data together provide a comprehensive basis for energy prediction, ensuring the reference value of the prediction results. Step 2.3. Construct a lightweight gradient boosting energy prediction model. This model is a lightweight prediction model based on the gradient boosting algorithm, adapted to the low-computing-power hardware requirements of off-grid scenarios. The lightweight gradient boosting energy prediction model includes an input feature layer, an iterative training layer composed of multiple decision trees, a result fusion layer, and an output layer. The input feature layer connects to the iterative training layer, the iterative training layer connects to the result fusion layer, and the result fusion layer connects to the output layer. The input feature layer receives multi-dimensional input data, normalizes the data to the zero-to-one interval, and then transmits it to the iterative training layer. The iterative training layer completes data fitting and iterative calculation through multiple decision trees. The system employs a set number of decision trees for iterative fitting, with predefined learning rates and tree depth parameters for model training. The learning rate controls the iteration step size, and the tree depth controls the computational level of a single decision tree. The decision tree weights are updated after each iteration. The result fusion layer weighted and aggregates the outputs from multiple decision trees. The output layer outputs continuous prediction data, including energy state curves and remaining range data for a predetermined future duration, obtained through a lightweight gradient boosting energy prediction model. The system's operating mode is derived from the output data of this model, providing data support for matching the system's operating mode. Step 2.4. Using energy state curve data, remaining range data, and real-time status data of energy storage units, the corresponding system operating mode is matched. Energy state curve data is continuous data characterizing the future energy change trend of the system; remaining range data is predicted data characterizing the sustainable operating time of the system under the current energy state; and real-time status data of energy storage units is the current energy storage capacity and charge / discharge status data of each level of energy storage unit. The system compares these three types of data with preset operating mode thresholds to determine the currently suitable operating mode. The system operating mode will directly serve as the basis for subsequent communication parameter adjustments and interaction module control, realizing the linkage between energy state and system operation. In some specific implementations, the iterative training layer of the lightweight gradient boosting energy prediction model is set with fifty decision trees, a learning rate of 0.01, and a tree depth of six layers. The input feature layer converts the four types of input data into four-dimensional feature vectors. The decision tree weights are updated once per iteration of the iterative training layer. After training, the model accuracy is verified using validation set data. The output duration of the energy state curve data is set to 72 hours, and the output accuracy of the remaining battery life data is set to the hour level. This implementation solves the problems of insufficient computing power and low prediction accuracy in off-grid scenarios. By configuring lightweight model parameters, it adapts to low-computing-power hardware while ensuring the duration and accuracy requirements of energy prediction.

[0031] In some embodiments, the lightweight gradient boosting energy prediction model can be replaced by a lightweight neural network prediction model. The lightweight neural network prediction model is a neural network model adapted to low computing power devices. The lightweight neural network prediction model includes an input layer, a hidden layer, and an output layer. The input layer receives multi-dimensional input data, the hidden layer completes data calculation and feature extraction through a preset number of neuron nodes, and the output layer outputs energy state curve data and remaining battery life data. The prediction model type can be flexibly selected according to the computing power conditions of the system hardware.

[0032] Step 3. Adaptive communication parameter adjustment and differential update execution: Step 3.1. Adopt the system working mode and select the corresponding terminal wake-up level. The terminal wake-up level is the level setting that triggers the terminal to start communication operations. It is used to adapt to update tasks with different urgency levels. The terminal wake-up levels include periodic polling wake-up level, specific matching wake-up level, and broadcast emergency wake-up level. The periodic polling wake-up level is a wake-up method in which the terminal actively starts according to a preset period, which is suitable for non-urgent content update tasks. The specific matching wake-up level is a wake-up method in which the terminal starts after receiving a dedicated signaling, which is suitable for regular information update tasks. The broadcast emergency wake-up level is a wake-up method in which the terminal quickly starts after receiving a broadcast command, which is suitable for emergency information distribution tasks. The system switches the wake-up level according to the urgency of the update task. Different wake-up levels can balance the relationship between the real-time performance of the update task and the standby power consumption of the terminal. Step 3.2. Collect channel environment data. Channel environment data characterizes the operating status and transmission quality of the communication channel. This data includes rainfall data, signal-to-noise ratio (SNR) data, and received signal strength data. Rainfall data reflects the rainfall conditions in the terminal's operating environment and its impact on communication. SNR data is the ratio of communication signal to noise. Received signal strength data is the strength of the communication signal received by the terminal. The communication module collects rainfall data through sensors and obtains SNR and received signal strength data through the communication chip. Channel environment data is used to predict the channel quality level, and the channel quality level is used to adjust communication parameters. Communication parameters are the core parameters controlling the communication transmission effect. These parameters include spreading factor, packet size, and retransmission strategy. The spreading factor controls the extent of communication signal spread. Packet size is the segmentation and transmission specification of communication data. The retransmission strategy is the retransmission rule after communication data transmission failure. Parameter adjustments are made to adapt to different channel transmission conditions. Step 3.3. Collect real-time energy status data from the terminal. Divide the terminal's real-time energy status data into multiple energy levels. Different energy levels correspond to different update permissions, update package specifications, and transmission strategies. Update permissions define the range of update operations the terminal can perform. Update package specifications define the size of the updated data. Transmission strategies define the data transmission rate and retransmission rules. Match the corresponding update strategy to the terminal's energy level. The adjustment logic of the update strategy references standardized energy consumption data from a terminal-independent energy consumption node model. The adjustment logic of the update strategy is bound to the power consumption level of the system's operating mode. The bound update strategy is synchronized to the differential update package generation logic to ensure that the update operation is compatible with the terminal's energy status. Step 3.4. Using the adjusted communication parameters and update strategy, the difference between the old and new versions of data is extracted using a differential algorithm. The differential algorithm is a general algorithm for extracting data version differences. First, the binary content of the old and new versions of data is read, and the data differences are compared byte by byte. The difference positions and contents are marked, and only the data with differences is extracted to generate a differential update packet. The differential update packet is an update data packet containing only the difference data, which can significantly reduce the amount of data transmitted. The differential update operation is performed, and the differential update packet is transmitted to the terminal. The terminal uses a differential restoration algorithm to merge the differential data with the old version data, completing the data restoration and realizing remote data update of the terminal. In some specific implementations, the communication parameters... Based on the channel quality level, there are four configuration levels. When the channel quality is excellent, the spreading factor is set to seven, the packet size is set to 256 bytes, and the retransmission count is set to one. When the channel quality is good, the spreading factor is set to nine, the packet size is set to 128 bytes, and the retransmission count is set to three. When the channel quality is medium, the spreading factor is set to eleven, the packet size is set to sixty-four bytes, and forward error correction is enabled. When the channel quality is poor, the spreading factor is set to twelve, the packet size is set to thirty-two bytes, and a multi-redundancy transmission method is adopted. This implementation method solves the problems of low communication transmission success rate and channel congestion in harsh environments. By configuring the parameters in a hierarchical manner, it adapts to different channel conditions and improves transmission stability.

[0033] In some specific implementations, the differential algorithm adopts the bsdiff differential algorithm to perform binary comparison of the new and old firmware data. The difference extraction precision is set to the single-byte level, and the size of the differential update package can be reduced to less than one-tenth of the original complete update package. The terminal restoration operation is completed through the bspatch restoration algorithm. The restoration process only calls the terminal's basic storage unit and does not occupy additional running memory. This implementation method solves the problems of large data volume, long transmission time, and high terminal power consumption in remote updates. Through fine differential extraction and low-power restoration, the system load of the update process is reduced.

[0034] In some embodiments, the differential algorithm can be replaced by the bsdiff differential algorithm, which is a commonly used binary data differential algorithm. It can achieve more refined difference extraction for binary format firmware and data content, further reduce the size of the differential update package, reduce the amount of data transmitted in communication, and reduce the power consumption of the terminal during the update process.

[0035] Step 4. Offline interaction control and collaborative closed-loop formation: Step 4.1. Adopt the system operating mode and match the corresponding operating level to the near-field interaction module. The near-field interaction module is a functional module that realizes near-range wireless interaction. It realizes data interaction based on near-field communication technology. The system operating mode comes from the output data of the lightweight gradient boosting energy prediction model. The operating level is the running permission setting of the near-field interaction module. Different operating levels correspond to different interactive function activation states. The operating level is directly related to the terminal energy status. The system closes unnecessary interactive functions according to the energy status to ensure the adaptability of interactive functions and energy supply. Step 4.2. The near-field interaction module stores interaction identifier data. This data is the core identifier for offline interaction and includes scene identifier data, encryption token data, and language tag data. Scene identifier data is a unique identifier representing the interaction scene; encryption token data is encrypted verification data to ensure the security of the interaction data; and language tag data is identifier data representing the interaction language. The near-field interaction module writes these three types of data into the storage area in a fixed order, completing the offline data interaction operation with the user terminal. In a network environment, the user terminal pre-downloads the corresponding multilingual content package to its local storage area. The user terminal reads the interaction identifier data through the near-field interaction module and retrieves the content package corresponding to the interaction identifier data from its local storage area to complete the offline interaction operation. The offline interaction operation does not rely on a public network connection. The behavioral data generated by the offline interaction operation is synchronized to the storage area of ​​the near-field interaction module, and the data in the storage area of ​​the near-field interaction module is synchronized to the terminal's local cache area, completing the initial storage of the interaction data. Step 4.3. Perform local caching on the behavioral data generated by offline data interaction operations. Behavioral data is the foundational data representing user interaction behavior. It includes interaction timestamp data, terminal identification data, and user behavior characteristic data. Interaction timestamp data records the time of the interaction, terminal identification data is the unique identifier of the user's terminal, and user behavior characteristic data represents the user's interaction methods and frequency. The caching operation writes the behavioral data to the terminal's non-volatile storage unit in chronological order, overwriting the earliest historical data to prevent data loss during interaction and provide a complete data source for subsequent data backhaul. Step 4.4. According to the set period, the cached behavioral data is batch transmitted back through a low-power self-organizing network. The low-power self-organizing network is a wireless communication network composed of low-power terminals, which has the characteristics of low power consumption and self-organizing network operation. The terminal wakes up the communication module at the set time period, packages the cached behavioral data and transmits it to the aggregation node. The transmitted behavioral data is synchronized to the input feature layer of the lightweight gradient boosting energy prediction model, providing supplementary input data for the lightweight gradient boosting energy prediction model, enriching the input dimension of the model, and optimizing the prediction accuracy of the lightweight gradient boosting energy prediction model. In some specific implementations, the total length of the interaction identifier data of the near-field interaction module is set to sixteen bytes, of which the scene identifier data occupies four bytes, the encryption token data occupies eight bytes, the language tag data occupies one byte, and one byte is reserved as an extension bit. The interaction response time is controlled within 0.3 seconds. The content package pre-downloaded by the user terminal is stored according to language. After reading the interaction identifier data, the corresponding language content is directly matched. This implementation method solves the problems of slow interaction response and large data transmission volume in the absence of network. Through the extremely simple data format design, the interaction efficiency is improved and the terminal interaction power consumption is reduced.

[0036] In some specific implementations, the cache capacity of behavioral data is set to 1,000 records, the return cycle is set to 24 hours, and the return period is set to a time when the ambient light is sufficient and the system energy consumption is low. The returned data is transmitted in an encrypted format, and the aggregation node forwards it directly to the cloud platform after receiving it. This implementation method solves the problems of insufficient offline data storage capacity and excessive return energy consumption. By fixing the cache capacity and timed return, the relationship between data storage and energy consumption is balanced.

[0037] In some embodiments, the backhaul operation of behavioral data can be performed during periods of low system energy consumption, avoiding the runtime of high-energy-consuming services performed by the terminal, reducing the impact of the backhaul operation on the normal operation of the terminal, reducing the additional energy consumption caused by data backhaul, and improving the overall stability of the system operation.

[0038] The collaborative closed-loop management of energy, communication, and interaction includes: adjusting the operating parameters of the communication module and the working level of the near-field interaction module using the energy status data of the terminal; the communication module is a functional module that enables remote communication with the terminal, and data transmission is achieved based on low-power wireless communication technology; updating the input characteristics of the lightweight gradient boosting energy prediction model using behavioral data generated by offline interaction; calibrating the output parameters of the lightweight gradient boosting energy prediction model using actual energy consumption data from differential update operations; updating the system operating mode using the output data of the lightweight gradient boosting energy prediction model; adjusting the energy allocation logic of the three-level energy storage architecture, the communication parameters of the communication module, and the working level of the near-field interaction module using the updated system operating mode; updating the terminal-independent energy consumption node model using the updated energy consumption data; and resynchronizing the updated data of the terminal-independent energy consumption node model to the energy allocation logic of the three-level energy storage architecture. This establishes a bidirectional data transmission and logically mutually constraining operational relationship between the modules, ensuring that energy supply, communication transmission, and interactive operations remain in a state of coordinated adaptation.

[0039] In some embodiments, the collaborative closed-loop management of energy, communication, and interaction can add a real-time status feedback link. The terminal collects its own energy consumption data, communication transmission data, and interaction operation data in real time, and synchronizes the real-time collected data to the system control node. The control node dynamically adjusts the operating parameters of the lightweight gradient improvement energy prediction model based on the real-time data, forming a closed-loop control process of real-time iterative optimization, which further improves the adaptability of system operation.

[0040] In some embodiments, a two-tier energy storage architecture can be used instead of a three-tier energy storage architecture to achieve energy supply. The two-tier architecture includes transient energy storage units and long-term energy storage units. The transient energy storage units correspond to transient power consumption data in the terminal-independent energy consumption node model, while the long-term energy storage units correspond to standby power consumption data and daily average energy consumption data. The energy scheduling logic still uses the output data of the lightweight gradient boosting energy prediction model. This scheme simplifies the energy storage structure, reduces hardware deployment costs, and retains the core steps of terminal-independent energy consumption adaptation and energy prediction scheduling, solving the technical problems of complex hardware and high deployment costs in off-grid scenarios. In some embodiments, a centralized update protocol can be used instead of environment-adaptive differential updates to achieve remote updates. The specific sub-steps are as follows: Step 3.1. Use a terminal-independent energy consumption node model to divide terminal groups, and group terminals with similar energy consumption characteristics into the same update group; Step 3.2. Use a lightweight gradient boosting energy prediction model to determine the update period, and select a period with sufficient energy to start the update operation; Step 3.3. Use a unified update package to transmit data in batches to terminals in the same group, and complete the data distribution through broadcast. Step 3.4. After receiving the update package, the terminal performs a local update operation, and the update result is synchronized to the energy consumption node model that is independent of the terminal. By simplifying the communication process through grouped batch updates, the core applications of the terminal-independent energy consumption node model and the lightweight gradient boosting energy prediction model are retained, thus solving the complex technical problems of large-scale terminal update scheduling.

[0041] In some embodiments, local storage interaction can be used to replace the near-field interaction module to achieve offline interaction. The user terminal reads preset scene data through the local storage unit. The scene data is bound to the scene identification data in the terminal-independent energy consumption node model. The behavioral data generated by the interaction is directly stored in the local storage of the user terminal and then synchronized to the system platform after accessing the public network. This solution eliminates the near-field interaction hardware module, retains the core functions of offline interaction and data acquisition, and realizes the linkage between interaction and energy by combining the terminal-independent energy consumption node model, thus solving the technical problem of high hardware cost for interaction in network-free environments.

[0042] In some embodiments, a rule-based prediction algorithm can be used to replace the lightweight gradient boosting energy prediction model to achieve energy scheduling. The specific sub-steps are as follows: Step 2.1. Collect standardized energy consumption data and environmental data of the terminal-independent energy consumption node model, and establish a rule matching library; Step 2.2. Compare the real-time collected data with the preset rules in the rule matching library to determine the energy scheduling scheme; Step 2.3. Adjust the energy allocation logic and system operating mode of the three-level energy storage architecture based on the rule matching results; Step 2.4. Update the actual running data to the rule matching library and continuously optimize the rule matching accuracy; This solution reduces the model's computing power requirements through rule matching, retains the core structure of the terminal-independent energy consumption node model and the three-level energy storage architecture, and solves the technical problem of insufficient hardware computing power in off-grid scenarios.

[0043] This embodiment achieves terminal-independent off-grid IoT collaborative management through standardized terminal energy consumption modeling, hierarchical energy storage allocation, adaptive communication updates, offline interactive control, and collaborative closed-loop construction, improving the overall operational performance of the off-grid IoT system from multiple dimensions. By standardizing terminal energy consumption data and constructing a terminal-independent energy consumption node model, the association between terminal business attributes and energy consumption control is separated, enabling the system to adapt to different types of terminal access, reducing the complexity of terminal customization development, and improving the system's terminal adaptability and scalability. Through hierarchical energy allocation in a three-tiered energy storage architecture and intelligent scheduling of a lightweight gradient-enhanced energy prediction model, the system's energy supply method is optimized, adapting to differentiated energy consumption needs in different operating scenarios, improving the rationality of energy utilization, and extending the system's continuous operating time in scenarios without external power supply. Through environment-adaptive communication parameter adjustment and differential update operations, the communication energy consumption for remote terminal updates is reduced, improving the stability of communication transmission in complex environments and reducing data transmission losses and failure probabilities. By implementing offline interaction control and behavioral data caching and feedback through the near-field interaction module, stable interactive operations were achieved in environments without a public network. Simultaneously, the collection of interactive data and model application were completed, providing real data support for the optimization of the system's predictive model. The energy, communication, and interaction collaborative closed loop constructed in this embodiment enables stable data linkage and logical connections between various functional modules. The processing results of preceding steps continuously provide execution basis for subsequent steps, comprehensively improving the operational stability and environmental adaptability of the off-grid IoT system. This provides a feasible implementation method for the large-scale deployment and long-term stable operation of off-grid IoT scenarios and also provides a reference path for the standardized construction of off-grid IoT systems.

[0044] The above description is merely a preferred embodiment of the present invention. It should be understood that the present invention is not limited to the forms disclosed herein and should not be construed as excluding other embodiments. It can be used in various other combinations, modifications, and environments, and can be altered within the scope of the concept described herein through the above teachings or related technologies or knowledge. Modifications and variations made by those skilled in the art that do not depart from the spirit and scope of the present invention should be within the protection scope of the appended claims.

Claims

1. A terminal-independent off-grid IoT collaborative management method, characterized in that, Includes the following steps: S1. Collect energy consumption characteristics of terminals accessing the system, standardize the collected energy consumption data, remove the service function-related parameters of the terminals, construct a terminal-independent energy consumption node model, and build a unified energy consumption management interface. S2. Adopting a terminal-independent energy consumption node model, energy distribution is performed through a three-level energy storage architecture. Multi-dimensional input data is collected to construct a lightweight gradient-enhanced energy prediction model. Energy state data is output through the lightweight gradient-enhanced energy prediction model, and the energy state data is used to match the corresponding system operating mode. S3. Adopt the system working mode, adjust the communication parameters in combination with the collected channel environment data, use the terminal energy status data to adjust and update the strategy, and perform differential update operation; S4. The working level of the near-field interaction module is adjusted by adopting the system working mode. The near-field interaction module completes offline interaction operations, caches and transmits the behavioral data generated by the interaction, and completes the collaborative closed-loop management of energy, communication and interaction.

2. The method according to claim 1, characterized in that, Step S1 includes the following sub-steps: S1.

1. Collect three types of energy consumption characteristic data from the terminals accessing the system. The three types of energy consumption characteristic data include standby power consumption data, transient power consumption data, and daily average energy consumption data. S1.

2. Normalize the collected energy consumption data, unify the statistical period, measurement unit and data acquisition node of the energy consumption data, and convert the original energy consumption data of different terminals into standard unit data under the set rated voltage. S1.

3. Remove three types of parameters from the terminal: business function type, business execution logic, and business output format, and retain the standardized energy consumption data; S1.

4. Construct a terminal-independent energy consumption node model using the retained standardized energy consumption data. The terminal-independent energy consumption node model includes a data input layer, a standardization verification layer, and a model output layer. The data input layer is connected to the standardization verification layer, and the standardization verification layer is connected to the model output layer. Establish a unified energy consumption management interface, which only transmits standardized energy consumption data.

3. The method according to claim 1, characterized in that, Step S2 includes the following sub-steps: S2.

1. Adopting a terminal-independent energy consumption node model, assigning corresponding energy supply scenarios and execution priorities to each energy storage unit of the three-level energy storage architecture; S2.

2. Collect historical load data, light prediction data, temperature data, and seasonal factor data as input data for the lightweight gradient boosting energy prediction model; S2.

3. Construct a lightweight gradient boosting energy prediction model. The lightweight gradient boosting energy prediction model includes an input feature layer, an iterative training layer composed of multiple decision trees, a result fusion layer, and an output layer. The input feature layer is connected to the iterative training layer, the iterative training layer is connected to the result fusion layer, and the result fusion layer is connected to the output layer. The iterative training layer uses a set number of decision trees for iterative fitting, and sets a set learning rate and tree depth parameters to complete the model training. The lightweight gradient boosting energy prediction model outputs energy state curve data and remaining battery life data within a set future time period. S2.

4. Use energy state curve data, remaining range data and real-time status data of energy storage units to match the corresponding system operating mode.

4. The method according to claim 1, characterized in that, Step S3 includes the following sub-steps: S3.

1. Adopt the system working mode and select the corresponding terminal wake-up level. The terminal wake-up levels include periodic polling wake-up level, specific matching wake-up level and broadcast emergency wake-up level. S3.

2. Collect channel environment data, including rainfall data, signal-to-noise ratio data, and received signal strength data. Use the channel environment data to predict the channel quality level, and use the channel quality level to adjust communication parameters, including spreading factor, packet size, and retransmission strategy. S3.

3. Collect real-time energy status data from the terminal, and use the real-time energy status data from the terminal to adjust and update the strategy. The adjustment logic of the update strategy refers to the standardized energy consumption data of the energy consumption node model that is independent of the terminal. S3.

4. Using the adjusted communication parameters and update strategy, the difference between the old and new versions of data is extracted using a differential algorithm to generate a differential update package and execute the differential update operation.

5. The method according to claim 1, characterized in that, Step S4 includes the following sub-steps: S4.

1. The system working mode is adopted to match the corresponding working level to the near-field interaction module. The system working mode is derived from the output data of the lightweight gradient boosting energy prediction model. S4.

2. The interaction identification data is stored through the near-field interaction module. The interaction identification data includes scene identification data, encryption token data and language tag data, to complete the offline data interaction operation with the user terminal. S4.

3. Locally cache the behavioral data generated by offline data interaction operations. The behavioral data includes interaction timestamp data, terminal identification data, and user behavior feature data. S4.

4. According to the set period, the cached behavioral data is batch transmitted back through a low-power self-organizing network, and the transmitted behavioral data is synchronized to the input feature layer of the lightweight gradient boosting energy prediction model.

6. The method according to claim 2, characterized in that, In step S1.4, the unified energy management interface shields non-standardized data and business function-related data. Newly connected terminals access the system by reporting standardized energy consumption data. The unified energy management interface synchronizes the terminal-independent energy consumption node model of the connected terminal to the input feature layer of the lightweight gradient boosting energy prediction model. The unified energy management interface collects real-time energy consumption data of the connected terminal according to a set period, updates the terminal-independent energy consumption node model, and synchronizes the updated terminal-independent energy consumption node model to the energy allocation logic of the three-level energy storage architecture.

7. The method according to claim 3, characterized in that, In step S2.1, the three-level energy storage architecture includes a primary energy storage unit, a secondary energy storage unit, and a tertiary energy storage unit. The primary energy storage unit corresponds to transient power consumption power supply scenarios, the secondary energy storage unit corresponds to daily load power supply scenarios, and the tertiary energy storage unit corresponds to backup energy reserve scenarios. The execution priority of the primary energy storage unit is higher than that of the secondary energy storage unit, and the execution priority of the secondary energy storage unit is higher than that of the tertiary energy storage unit. The energy allocation logic of the three-level energy storage architecture uses a terminal-independent energy consumption node model for parameter calibration. The calibrated energy allocation logic matches the power consumption level requirements of the system operating mode. The system operating mode is derived from the output data of the lightweight gradient boosting energy prediction model.

8. The method according to claim 4, characterized in that, In step S3.3, the real-time energy status data of the terminal is used to divide the energy into multiple energy levels. Different energy levels correspond to different update permissions, update package specifications and transmission strategies. The corresponding energy level of the terminal is matched with the corresponding update strategy. The adjustment logic of the update strategy is bound to the power consumption level of the system working mode. The bound update strategy is synchronized to the differential update package generation logic. The differential update package generation logic uses a differential algorithm to extract the differences between the old and new versions of data and generate a differential update package of the corresponding specification. The system working mode comes from the output data of the lightweight gradient boosting energy prediction model.

9. The method according to claim 5, characterized in that, In step S4.2, the user terminal pre-downloads the multilingual content package corresponding to the scenario to the local storage area under a network environment. The user terminal reads the interaction identifier data through the near-field interaction module. The user terminal retrieves the content package corresponding to the interaction identifier data from the local storage area to complete the offline interaction operation. The offline interaction operation does not rely on the public network connection throughout. The behavior data generated by the offline interaction operation is synchronized to the storage area of ​​the near-field interaction module. The data in the storage area of ​​the near-field interaction module is synchronized to the local cache area of ​​the terminal. The cached behavior data is synchronized to the input feature layer of the lightweight gradient boosting energy prediction model.

10. The method according to claim 1, characterized in that, In step S4, the coordinated closed-loop management of energy, communication, and interaction includes: adjusting the operating parameters of the communication module and the working level of the near-field interaction module using the terminal's energy status data; updating the input features of the lightweight gradient boosting energy prediction model using behavioral data generated by offline interaction; calibrating the output parameters of the lightweight gradient boosting energy prediction model using actual energy consumption data from differential update operations; updating the system working mode using the output data of the lightweight gradient boosting energy prediction model; adjusting the energy allocation logic of the three-level energy storage architecture, the communication parameters of the communication module, and the working level of the near-field interaction module using the updated system working mode; and updating the terminal-independent energy consumption node model using the updated energy consumption data.

Images