Mobile power supply multi-node electric quantity scheduling and equalization control method based on LoRa technology

Abstract

The invention belongs to the technical field of internet of things power supply management, and particularly relates to a mobile power supply multi-node electric quantity scheduling and balance control method based on a LoRa technology, which comprises the steps of constructing mobile power supply nodes and collecting data, constructing a LoRa communication network and transmitting data, and performing data processing and state modeling by an application server. And carrying out electric quantity scheduling and balance control based on a hierarchical multi-objective optimization strategy. The electric quantity dispatching and balancing control algorithm aims at minimizing the standard deviation of the state of charge, maximizing the available capacity of the system, optimizing the charging and discharging efficiency, balancing the cycle life and generating a dispatching instruction to be issued and executed. Through adoption of the technical scheme, low-power-consumption wide-area communication and global optimal configuration of distributed electric quantity resources can be realized, the overall available capacity and average life of the system are remarkably improved, and the operation and maintenance cost is reduced.

Description

Technical Field

[0001] This invention belongs to the field of Internet of Things power management technology, specifically a method for multi-node power scheduling and balancing control of mobile power supply based on LoRa technology. Background Technology

[0002] With the widespread adoption of mobile electronic devices, portable power supplies have expanded from providing power to single devices to multi-node distributed scenarios such as public rentals, the sharing economy, and emergency power supply. In these scenarios, achieving efficient scheduling and balanced control of power across multiple nodes is a core requirement for ensuring coordinated system operation, extending the overall lifespan of devices, improving user experience, and optimizing energy utilization.

[0003] In existing technologies, the battery management system for single-cell mobile power banks is relatively mature. It can achieve charge and discharge protection and cell balancing by monitoring parameters such as voltage, current, and temperature, effectively solving the safety and lifespan issues of individual battery packs. However, this type of technology only manages the core state of a single cell, and information interaction is limited to internal components of the device, lacking efficient external communication capabilities to support information sharing and collaborative decision-making among multiple nodes.

[0004] Existing multi-node communication solutions have inherent bottlenecks: Wi-Fi offers high transmission speeds but consumes a lot of power and has limited coverage, resulting in high costs for large-scale deployment; Bluetooth has low power consumption but short communication distances, making it unable to support cross-regional multi-node networking; while cellular networks offer wide coverage, their module costs and energy consumption are high, failing to meet the cost and battery life requirements of power banks. The lack of a communication layer that combines low power consumption, wide coverage, and low cost makes real-time communication between multiple nodes difficult, hindering the effective implementation of global power optimization scheduling and restricting the performance improvement of distributed systems.

[0005] Therefore, this invention provides a method for multi-node power scheduling and equalization control of mobile power supplies based on LoRa technology. Summary of the Invention

[0006] In order to overcome the shortcomings of the prior art, at least one technical problem raised in the background art is solved.

[0007] The technical solution adopted by this invention to solve its technical problem is: a multi-node power scheduling and equalization control method for mobile power supplies based on LoRa technology, comprising: Construction and Data Acquisition of Mobile Power Nodes: Each mobile power node includes a battery pack, a battery management system (BMS), a microcontroller unit (MCU), and a LoRa communication module. The battery pack consists of multiple lithium-ion cells connected in series and parallel, used for storing electrical energy and supplying power. The BMS is used for overcharge protection, over-discharge protection, overcurrent protection, short circuit protection, over-temperature protection, and cell voltage equalization management of the battery pack. The MCU communicates with the BMS via a high-speed serial interface to obtain real-time operating parameters of the battery pack, including but not limited to: total battery pack voltage, individual cell voltage, charging and discharging current, battery pack temperature, cycle count, remaining capacity, and state of charge estimated by the BMS. The MCU is also responsible for processing local logic, controlling charging and discharging paths, and managing the LoRa communication module. The LoRa communication module is connected to the MCU via an antenna, used to encapsulate the operating parameter data collected and processed by the MCU into LoRa data packets, and periodically transmit them to the LoRa gateway in a low-power, long-range wireless manner. The LoRa communication module employs spread spectrum modulation technology, operates in the unlicensed Industrial, Scientific, and Medical (ISM) band, and supports an Adaptive Data Rate (ADR) mechanism to dynamically adjust transmission power, spreading factor, and bandwidth based on link quality, thereby optimizing energy consumption and communication reliability. The data packet structure includes a unique node identifier, data type identifier, timestamp, encryption checksum, and specific operational parameter data.

[0008] LoRa Communication Network Setup and Data Transmission: The LoRa communication network includes at least one LoRa gateway, one LoRa network server, and one application server. The LoRa gateway receives LoRa data packets from the mobile power supply nodes and transmits these packets to the LoRa network server via Ethernet, cellular networks (e.g., 4G or 5G), or satellite links. The LoRa gateway is configured as a multi-channel receiver, capable of simultaneously receiving data transmissions from multiple mobile power supply nodes at different frequencies and spreading factors. The LoRa network server is responsible for deduplicating, decrypting, verifying integrity, calibrating timestamps, and managing routing of the received data packets, and forwarding verified data packets to the application server. The LoRa network server is also responsible for managing the parameters of the LoRa communication network, including but not limited to: dynamically adjusting the ADR parameters of the mobile power supply nodes, allocating transmission time slots, and handling network access authentication.

[0009] Application Server Data Processing and State Modeling: The application server receives and stores the mobile power node operating parameter data from the LoRa network server. The application server has a built-in data parsing module for parsing raw data packets and extracting various operating parameters of the mobile power node. The application server also includes a historical database and a real-time database, used to store long-term historical data and current real-time data, respectively. The application server is configured with a state of charge (SOC) and state of health (SOH) estimation module, which, based on the received operating parameter data, combines battery models (e.g., equivalent circuit models, electrochemical models, or Kalman filter models) and machine learning algorithms to perform high-precision, real-time dynamic estimation of the SOC and SOH of each mobile power node. The SOC estimation module comprehensively considers multiple techniques such as open-circuit voltage method, coulomb counting method, internal resistance method, and temperature compensation to improve the accuracy and robustness of the estimation. The SOH estimation module assesses the aging degree and usable life of the battery pack by analyzing parameters such as battery internal resistance changes, capacity decay curves, cycle count, and charge / discharge efficiency. The application server also maintains a node status table, which updates the unique identifier, geographical location information, current state of charge, health status, temperature, charging / discharging status, and availability status of each mobile power node in real time.

[0010] Power scheduling and balancing control algorithm: The application server has a built-in power scheduling and balancing control module, which generates optimized scheduling instructions for the distributed mobile power bank group based on the node status table and preset scheduling strategies. The power scheduling and balancing control algorithm adopts a hierarchical multi-objective optimization strategy. Its inputs include: real-time state of charge, health status, temperature, geographical location, historical usage data, and external demand forecasts for all mobile power bank nodes (e.g., mobile power bank demand forecasts for specific areas, charging pile availability information). The multi-objective optimization strategy aims to simultaneously achieve the following objectives: The goal of state-of-charge equalization is to minimize the standard deviation of the state of charge of all mobile power nodes in the system, so as to avoid overcharging or over-discharging of individual nodes and extend the overall system life.

[0011] The goal of maximizing the available capacity of the system is to maximize the number of mobile power nodes in the system that are in an available state (within the safe charging range) and their total power, while satisfying the balance constraints.

[0012] Charging / discharging efficiency optimization goal: Minimize the energy loss generated during charging and discharging, and take into account the aging degree of batteries at different nodes, prioritizing charging and discharging of batteries in good health.

[0013] Cycle life balancing target: During long-term operation, balance the number of charge and discharge cycles of each mobile power node to avoid overuse of some nodes and accelerated aging, thereby extending the overall service life of the system.

[0014] The specific execution steps of the power scheduling and balancing control module include: Data collection and preprocessing: Obtain the latest operating parameters of all online mobile power nodes from the real-time database, and update the state of charge, health status, and health rating of each node based on the output of the state of charge and health status estimation module.

[0015] Demand and resource matching: Based on the distribution of all mobile power nodes in the current system, real-time state of charge and health status, as well as the prediction of potential external demand, identify low-power nodes that need charging, high-power nodes that can provide power, and nodes that are in standby or maintenance status.

[0016] Build an optimization model: Decision variables: Define the charging state (charging, discharging, standby), target state of charge for charging / discharging, and charging / discharging rate of each mobile power node in the next scheduling cycle.

[0017] Objective function: ,in It is the charge state of node i. It is the average state of charge of all nodes.

[0018] ,in This is the available capacity of node j. It is a binary variable that indicates whether node j is available.

[0019] ,in It represents the energy loss of node k during the charging and discharging process.

[0020] ,in It is the number of iterations of node m. It represents the average number of cycles across all nodes.

[0021] Constraints: ,in and These are the lower and upper limits of the battery's safe state of charge (e.g., 10% and 95%).

[0022] as well as Ensure that the charging and discharging rates are within a safe range.

[0023] The available capacity of a node is related to its health status.

[0024] ∑ChargePower=∑DischargePower+∑SystemLoad, the total energy of the system is conserved.

[0025] CommunicationLatency≤Threshold ensures that scheduling instructions can be delivered in a timely manner.

[0026] Solving the optimization problem: Heuristic algorithms (such as genetic algorithms and particle swarm optimization algorithms) or linear / nonlinear programming solvers (such as Gurobi and CPLEX) are used to solve the above multi-objective optimization model to obtain the optimal scheduling scheme.

[0027] Generate scheduling instructions: Based on the solution results, generate specific scheduling instructions for each mobile power node. These instructions include the target state of charge, charging / discharging mode (charging, discharging, or standby), and suggested charging / discharging rates. The scheduling instructions are encoded in a simplified binary format to minimize LoRa transmission load.

[0028] Command issuance: The scheduling command is transmitted to the LoRa communication module of the target mobile power node via the application server, the LoRa network server, and the LoRa gateway.

[0029] Execution and Feedback of Scheduling Commands: After receiving a scheduling command from the LoRa communication network, the MCU of the mobile power node parses and verifies the command. If the command is valid, the MCU will adjust the charging and discharging strategy of the BMS or control external charging and discharging devices to execute the scheduling command. The execution process includes, but is not limited to: starting or stopping charging, adjusting the charging current, starting or stopping discharging, and adjusting the discharging current. During command execution, the mobile power node will monitor the battery pack status in real time and periodically send feedback information to the application server based on the execution results. The feedback information includes the current state of charge, actual charging and discharging rate, command execution status, and any abnormal situation reports. Based on the feedback information, the application server evaluates the scheduling effect and adjusts subsequent scheduling strategies as needed, forming a closed-loop control system.

[0030] Furthermore, preferably, the LoRa communication module operates in the 915MHz or 868MHz frequency band and supports the LoRaWAN protocol stack to ensure compatibility with existing LoRaWAN infrastructure. The LoRa communication module is configured as a Class A device, enabling low-power listening, briefly opening the receive window to receive downlink commands only after data transmission, and entering deep sleep mode when no commands are received.

[0031] Furthermore, preferably, the microcontroller unit (MCU) is a power-optimized 32-bit ARM Cortex-M series microprocessor with an operating frequency of tens of megahertz. It integrates low-power peripheral interfaces such as SPI, I2C, and UART, as well as ample flash memory and SRAM to support local data storage and the operation of complex state-of-charge / state-of-health algorithms. The MCU also includes a watchdog timer and multiple low-power modes to ensure system stability and battery life.

[0032] Furthermore, preferably, the battery management system (BMS) integrates a high-precision voltage acquisition chip with millivolt-level voltage measurement accuracy; a high-precision current sampling chip with milliampere-level current measurement accuracy; and a multi-channel temperature sensor interface to monitor the temperature of multiple cells in real time. The BMS also features passive or active balancing functions, working in conjunction with the MCU to initiate or adjust the balancing current upon receiving a balancing command, thereby minimizing voltage or capacity differences between cells.

[0033] Furthermore, preferably, the power scheduling and balancing control module performs data cleaning and anomaly detection before executing the multi-objective optimization algorithm. The data cleaning module is used to identify and remove erroneous data packets or abnormal sensor readings that may be generated during transmission; the anomaly detection module, based on historical data and preset thresholds, determines in real time whether there is abnormal behavior at the level of a single mobile power node or the entire system, and triggers an alarm mechanism to notify the system administrator to intervene.

[0034] Furthermore, preferably, the state of charge (SOC) and state of health (SOH) estimation module of the application server employs a deep learning model based on a Long Short-Term Memory (LSTM) network. This model uses battery voltage, current, temperature, and historical charge / discharge data as input features to predict the real-time SOC and SOH of the battery through training. The training data comes from a large number of operational mobile power nodes, and the model parameters are continuously optimized through online learning or offline periodic updates to adapt to changes in battery characteristics caused by aging and different usage environments.

[0035] Furthermore, preferably, the power scheduling and balancing control algorithm incorporates geographical location information into the optimization decision. By performing cluster analysis on the geographical distribution of mobile power nodes, the algorithm can identify areas with high power demand or excess power within a specific region. When mobile power nodes in a certain area are generally in a low-charge state, the scheduling module will prioritize instructing nearby high-charge nodes to move to that area (if they have the capability to move) or instruct charging infrastructure to tilt towards that area.

[0036] Furthermore, preferably, the application server is configured with a user interface module to provide a visual dashboard and data reporting functions. The dashboard can display key information in real time, such as the geographical distribution, state of charge distribution, health status trends, total system capacity, and charging / discharging status of all mobile power nodes. The data reporting module can generate periodic reports to evaluate system performance, energy utilization efficiency, and the effectiveness of scheduling strategies. The user interface module also allows administrators to manually intervene in scheduling strategies, such as forcing specific nodes to charge or discharge, or adjusting the weights of the objective function.

[0037] The beneficial effects of this invention are as follows: This invention discloses a multi-node power scheduling and balancing control method for mobile power supplies based on LoRa technology. By introducing LoRa communication technology, it completely solves the inherent contradictions of traditional wireless communication technologies in multi-node mobile power supply scenarios, such as high power consumption, small coverage, complex network deployment, and high operating costs. LoRa technology achieves low-power wide-area coverage, enabling large-scale distributed mobile power supply nodes to conduct real-time and reliable data interaction with extremely low self-energy consumption, thereby ensuring the system's battery life and deployment economy. By integrating an MCU and LoRa communication module into the mobile power node and deeply integrating it with the BMS, the intelligence and networking of a single node are realized, enabling it to have the ability to process local data, receive and execute remote commands, and report status in real time, laying a solid foundation for global optimization scheduling. The hierarchical multi-objective optimization scheduling algorithm proposed in this invention comprehensively considers multiple factors such as the state of charge, health status, geographical location, usage demand, and energy utilization efficiency of all mobile power nodes within the system, achieving globally optimal allocation of distributed power resources. This not only effectively solves the "weakest link" effect caused by differences in power levels between different nodes, significantly improving the overall available capacity and average lifespan of the system, but also avoids premature aging of some batteries by balancing the cycle count of each battery, thereby greatly extending the service life and return on investment of the entire mobile power group. This invention provides detailed engineering implementation specifications, including optimization of the LoRa communication protocol, selection of the MCU, data interface of the BMS, accurate estimation method for state of charge / health, and specific optimization algorithm models, ensuring the feasibility, stability, and high reliability of this technical solution. The method not only improves the overall energy management level of the mobile power system and optimizes resource allocation efficiency, but also significantly reduces operating and maintenance costs, thus providing strong technical support for the future promotion and application of large-scale, distributed mobile power supplies. Attached Figure Description

[0038] The invention will now be further described with reference to the accompanying drawings.

[0039] Figure 1 This is a flowchart illustrating the power scheduling and balancing control method of the present invention; Figure 2 This is a system block diagram of the mobile power supply multi-node power scheduling and balancing control system based on LoRa technology of the present invention; Figure 3 This is a schematic diagram of the structure of the mobile power node of the present invention.

[0040] In the diagram: 1. Battery pack; 2. Battery management system; 3. Microcontroller unit; 4. LoRa communication module; 5. Antenna; 6. LoRa gateway; 7. LoRa network server; 8. Application server; 9. Data parsing module; 10. Historical database; 11. Real-time database; 12. State of charge and health estimation module; 13. Node status table; 14. Power scheduling and balancing control module. Detailed Implementation

[0041] This invention provides a multi-node power scheduling and balancing control method for mobile power banks based on LoRa technology, constructing an efficient, reliable, and economical distributed mobile power bank management system. This method integrates low-power LoRa communication capabilities into each mobile power bank node and combines it with intelligent scheduling and balancing control algorithms on the application server to achieve real-time monitoring, status assessment, predictive scheduling, and closed-loop feedback control of a large-scale mobile power bank group, thereby significantly improving the overall energy utilization efficiency, availability, and lifespan of the system. The following will be discussed in conjunction with the attached... Figure 1-3 As shown, a specific embodiment of the present invention will be described in detail to ensure that those skilled in the art can fully understand and implement the present invention in accordance with this disclosure.

[0042] In one specific embodiment, the core of the LoRa-based mobile power supply multi-node power scheduling and balancing control method of the present invention lies in a collaborative distributed hardware system and a centralized intelligent software platform.

[0043] First, at the mobile power node level, such as Figure 3 As shown, each mobile power node (hereinafter referred to as "node") is designed as an autonomous unit integrating energy storage, status monitoring, local control, and wireless communication functions. Specifically, the core components of each node include a battery pack 1, a battery management system (BMS) 2, a microcontroller unit (MCU) 3, and a LoRa communication module 4. The battery pack 1 consists of multiple lithium-ion cells connected in series and parallel, such as standardized 18650 or 21700 ternary lithium-ion cells, to achieve a rated capacity of 20Ah to 50Ah and provide a nominal voltage output of 12V to 24V. These cells are connected by high-precision laser welding or ultrasonic welding processes and encapsulated in a shell with good heat dissipation and mechanical strength to ensure safety and durability in various mobile application scenarios. The battery pack 1 also integrates a temperature sensor to monitor the temperature distribution on the surface of the cells in real time to prevent overheating risks.

[0044] The battery management system 2 acts as the "guardian" of the battery pack 1, and its main responsibility is to perform comprehensive safety management and performance optimization of the battery pack 1.

[0045] Specific functions include: high-precision single-cell voltage monitoring to ensure that the voltage of all series-connected cells is within a safe range, such as 2.5V to 4.2V; Overcharge protection: When the voltage of any cell exceeds the set upper limit (e.g., 4.25V), the charging path is immediately cut off; Over-discharge protection: When the voltage of any cell falls below the set lower limit (e.g., 2.5V), the discharge path is immediately cut off. Overcurrent protection monitors the charging and discharging current. When the current continuously exceeds a preset threshold (such as 50A), the circuit is quickly disconnected to prevent damage. Short circuit protection: responds to external short circuit events in a very short time and quickly cuts off the current; Over-temperature protection: when the temperature of the battery pack or any cell exceeds the safe range (e.g., 60°C), the charging and discharging operation is suspended. And crucially, cell voltage balancing management is also included. This can be achieved through passive or active balancing techniques. Passive balancing uses parallel resistors to dissipate the charge of high-voltage cells, while active balancing uses energy transfer circuits to transfer energy from high-voltage cells to low-voltage cells, minimizing voltage imbalance between cells and thus extending the overall lifespan and usable capacity of the battery pack. The battery management system 2 typically integrates a dedicated battery management IC, such as Analog Devices' LTC681X series or TI's BQ769x0 series chips. These chips provide accurate acquisition of multi-channel voltage, current, and temperature data and communicate with an external MCU via high-speed serial interfaces such as UART or SPI. Through this interface, the MCU can obtain real-time operating parameters of the battery pack, including but not limited to: the total voltage of the battery pack, the independent voltage of each series-connected cell, real-time charging and discharging current, multiple temperature readings of the battery pack, the number of charge and discharge cycles since the BMS was started, and the remaining capacity and current state of charge estimated by the BMS's internal algorithm.

[0046] The microcontroller unit 3 is the "brain" of the node, responsible for processing local logic, controlling the charging and discharging path, and managing the LoRa communication module 4.

[0047] In a preferred embodiment, the microcontroller unit 3 is a power-optimized 32-bit ARM Cortex-M series microprocessor, such as STMicroelectronics' STM32L4 series, whose operating frequency can be flexibly adjusted within the tens of megahertz range (e.g., from 4MHz to 80MHz), and integrates low-power peripheral interfaces such as SPI, I2C, and UART for data communication with the battery management system 2. Furthermore, the microcontroller unit 3 has sufficient flash memory (e.g., 256KB to 1MB) and SRAM (e.g., 64KB to 256KB) to support local sensor data storage, partially local operation of complex state-of-charge / state-of-health estimation algorithms (such as coulomb counting accumulation and simple temperature compensation), and the operation of the LoRa communication protocol stack. To ensure system stability and battery life, the microcontroller unit 3 integrates a watchdog timer to prevent program crashes and supports multiple low-power modes during inactivity, such as stop mode, standby mode, or sleep mode, thereby reducing the node's static power consumption to the microampere level. The microcontroller unit 3 also controls the charging and discharging relays or solid-state switches inside the node to precisely cut off or connect the charging or discharging circuit in response to the safety instructions of the battery management system 2 or the scheduling instructions of the application server.

[0048] The LoRa communication module 4 serves as a bridge for long-distance, low-power communication between the node and the outside world. Connected to the SPI interface of the microcontroller unit 3 via a high-efficiency PCB antenna or an external rod antenna 5, the LoRa communication module 4 encapsulates the operating parameter data collected and processed by the microcontroller unit 3 into LoRa data packets. These data packets are periodically transmitted to the LoRa gateway 6 in a low-power, long-distance wireless manner. The LoRa communication module 4 typically uses Semtech's SX127X or SX126X series chips, whose core technology is Chirp Spread Spectrum (CSS), which gives LoRa excellent anti-interference capabilities and ultra-long communication distances. The module operates in unlicensed Industrial, Scientific, and Medical (ISM) bands, such as 868MHz in Europe or 915MHz in North America.

[0049] To further optimize energy consumption and communication reliability, the module supports an adaptive data rate (ADR) mechanism, which dynamically adjusts the transmit power (TxPower), spreading factor (SF), and bandwidth (BW) based on link quality (signal-to-noise ratio SNR, received signal strength RSS). For example, when link quality is good, the ADR mechanism can automatically reduce the spreading factor and transmit power, increase the data rate, shorten the transmission time, and thus reduce power consumption. When link quality is poor, the spreading factor and transmission power are increased, sacrificing data rate to enhance signal penetration and reception reliability. The structure of the data packets is carefully designed to minimize transmission load while ensuring information integrity. It typically includes a unique node identifier (e.g., a 32-bit or 64-bit sequence number), a data type identifier (indicating whether the data packet contains state of charge, health status, temperature, or feedback information), a precise timestamp, an encryption checksum (e.g., a digest encrypted with CRC16 or AES-128), and specific operating parameter data (e.g., compressed state of charge values, voltage and current data accurate to one decimal place).

[0050] Preferably, the LoRa communication module 4 strictly adheres to the LoRaWAN protocol stack to ensure compatibility and interoperability with existing LoRaWAN infrastructure. It is configured as a Class A device to achieve ultra-low power listening: after data transmission is complete, it only activates the receive mode for a preset short window period (e.g., 1 or 2 seconds) to receive possible downlink scheduling commands. Once the receive window closes and no commands arrive, it immediately enters deep sleep mode, thereby reducing average current consumption to the microampere level.

[0051] Secondly, regarding the construction of LoRa communication networks and data transmission, such as Figure 2As shown, this invention constructs a multi-layered communication architecture. The core components of the LoRa communication network include at least one LoRa gateway 6, one LoRa network server 7, and one application server 8. The LoRa gateway 6 is deployed within the coverage area of ​​distributed mobile power nodes, serving as a relay station between the nodes and the backend network. The LoRa gateway 6 typically integrates a baseband concentrator chip such as Semtech's SX1301 or SX1302, making it a multi-channel receiver capable of simultaneously receiving data transmissions from multiple mobile power nodes at different frequencies (e.g., 8 uplink channels) and different spreading factors, significantly improving network capacity. After receiving LoRa data packets, the gateway forwards them to the LoRa network server 7 through various backhaul methods. These backhaul methods include high-speed Ethernet connections (suitable for wired network coverage areas), cellular networks (e.g., 4G or 5G modules, suitable for vast mobile areas or areas without wired networks, providing high bandwidth and reliability), or satellite links (suitable for extremely remote or disaster scenarios, providing global coverage despite higher costs).

[0052] The LoRa network server 7 acts as the "manager" of the LoRa network, encompassing preliminary processing of downlink data and management of network parameters. It is responsible for deduplicating (filtering out redundant packets caused by multiple receptions), decrypting (decrypting the data packet payload using a preset AES key), verifying integrity (verifying data integrity through CRC or MAC checksums), and calibrating timestamps (uniformly calibrating the timestamps reported by the gateway based on the NTP protocol to eliminate clock drift) the verified valid data packets to the application server 8. In addition, the LoRa network server 7 also plays the role of a network parameter manager, including but not limited to: dynamically adjusting the ADR parameters of the mobile power node according to link quality indicators (such as SNR, RSSI) to optimize the transmission strategy of each node; intelligently allocating transmission time slots (e.g., through a simple ALOHA protocol or a more complex TDMA-like mechanism) to avoid air collisions between nodes; and handling network access authentication to ensure that only legitimate mobile power nodes can join the network and transmit data, for example through LoRaWAN's OTAA (Over-The-Air Activation) or ABP (Activation By Personalization) mechanisms.

[0053] Furthermore, in terms of data processing and state modeling for application servers, such as Figure 2As shown, the application server 8 is the "intelligent center" of the entire system, responsible for receiving, storing, and analyzing the mobile power node operating parameter data from the LoRa network server 7, and performing advanced state estimation and intelligent decision-making based on this data. The application server 8 can be deployed on cloud servers (such as AWS, Azure, Alibaba Cloud) or local high-performance server clusters to provide sufficient computing and storage resources.

[0054] The application server 8 integrates a data parsing module 9, whose primary task is to perform deep parsing of the received raw LoRa data packets and extract the core operating parameters of the mobile power node. This module is responsible for converting binary or compressed data into structured, readable formats (such as JSON and Protocol Buffers), and performing preliminary data type conversion and unit unification. For example, it converts raw ADC readings into voltage values ​​and current sensor readings into ampere values.

[0055] To store and manage massive amounts of node data, the application server 8 also includes a historical database 10 and a real-time database 11. The historical database 10 typically uses a relational database (such as PostgreSQL or MySQL) or a time-series database (such as InfluxDB or TimescaleDB) to store historical operational data over long periods, such as hourly state-of-charge curves, daily charge / discharge cycle counts, and monthly health status decay trends. This data is used for offline analysis, model training, and trend prediction. The real-time database 11 typically uses an in-memory database (such as Redis) or a high-speed NoSQL database (such as MongoDB) to store the latest and most accurate real-time data for all online mobile power nodes at the current moment, including the latest state of charge, health status, temperature, and charge / discharge status, so that the power scheduling and balancing control module 14 can quickly obtain the latest status for decision-making.

[0056] The application server 8 is configured with a state of charge and health estimation module 12, which is a key component of the present invention for realizing intelligent scheduling. It is used to perform high-precision, real-time dynamic estimation of the state of charge and health of each mobile power node based on received operating parameter data, combined with advanced battery models and machine learning algorithms.

[0057] For state of charge estimation, this module integrates multiple techniques to improve the accuracy and robustness of the estimation: Open-circuit voltage method (OCV method): This method quickly obtains the initial state of charge (SOC) value or corrects long-term coulomb counting errors by measuring the battery's terminal voltage in a static state and consulting a pre-calibrated OCV-SOC lookup table. This method requires a sufficiently long static period for the battery and is therefore mainly used after prolonged standby at the nodes or as an auxiliary correction.

[0058] Coulomb counting method: This method calculates the change in state of charge by accurately integrating the charging and discharging current and accumulating the amount of charge entering and leaving the battery. Specifically, the MCU collects the charging and discharging current at sampling periods of 100ms or even shorter and uploads it to the application server, where the state of charge estimation module performs cumulative calculations. To overcome the long-term cumulative error problem of coulomb counting, this invention combines calibration information provided by the BMS with the OCV method for periodic correction.

[0059] Internal resistance method: The internal resistance of a battery is correlated with its state of charge. By measuring the change in AC or DC internal resistance, the range of the state of charge can be determined.

[0060] Temperature compensation: Both the battery capacity and the OCV curve are affected by temperature. The state of charge estimation module will dynamically compensate the capacity parameters in the OCV-state of charge curve and coulomb count based on the real-time temperature data reported by the node, so as to eliminate the interference of temperature changes on the state of charge estimation.

[0061] Kalman Filter Model: To achieve high-precision and robust state-of-charge (POC) estimation under dynamic conditions, this invention employs either an Extended Kalman Filter (EKF) or an Unscented Kalman Filter (UKF) model. These models combine the battery equivalent circuit model (e.g., an RC network connected in series with an open-circuit voltage source) with the Coulomb counting method. Utilizing real-time voltage, current, and temperature as observations, they iteratively update the state-of-charge equation and covariance matrix to predict the POC in real time and correct measurement noise and model errors. Especially in scenarios with drastic fluctuations in charging and discharging current, their performance far surpasses that of single methods.

[0062] For health status estimation, this module assesses the aging level and usable life of the battery pack by analyzing the following parameters: Internal resistance variation analysis: As the battery ages, its internal resistance gradually increases. The health status estimation module monitors the long-term trend of the battery's internal resistance and compares it with the baseline internal resistance at the time of manufacture to assess the degree of aging.

[0063] Capacity decay curve: A capacity decay curve is constructed by monitoring the ratio of the actual usable capacity to the nominal capacity during each full charge-discharge cycle. By fitting this curve to a pre-defined capacity decay model, the remaining battery life can be predicted.

[0064] Charge / discharge cycle count: The number of cycles reported by the BME is an important indicator for health status assessment, as battery life is typically measured in cycles. The health module tracks the cumulative number of cycles at each node and combines this data with data provided by the manufacturer to assess health status degradation.

[0065] Charge / discharge efficiency: As batteries age, charge / discharge efficiency may decrease. By comparing the input energy during charging and the output energy during discharging, efficiency can be calculated and used as an auxiliary indicator of battery health.

[0066] Machine Learning Algorithm: Further, preferably, the state of charge (SOC) and health state estimation module employs a deep learning model based on Long Short-Term Memory (LSTM) networks. This LSTM model uses battery voltage, current, temperature, and historical charge / discharge data (e.g., average charge / discharge current and cycle increment over the past 24 hours) as input features. The unique gating mechanism of LSTM enables it to effectively handle long-term dependencies in time-series data, thereby more accurately capturing the complex nonlinear aging characteristics of the battery. The model is trained on an application server using real data collected from a large number of actually operating mobile power supply nodes. This data includes voltage, current, and temperature curves of the battery under different usage intensities and ambient temperatures, along with corresponding real SOC / health state labels. The training process employs an optimizer (e.g., Adam) and a loss function (e.g., mean squared error), continuously optimizing model parameters through online learning (periodic mini-batch data updates) or offline periodic updates to adapt to changes in battery characteristics caused by aging and different usage environments, thus ensuring the accuracy and adaptability of the health state estimation.

[0067] The application server 8 also maintains a node status table 13, a crucial in-memory data structure that updates the latest status information of each mobile power node in real time. This table typically contains the following fields: a unique identifier for the node (MAC address or serial number), geographic location information (GPS coordinates, Wi-Fi / cellular base station positioning, if the node has mobility capabilities), current state of charge (provided by the state of charge estimation module), health status (provided by the health status estimation module), battery pack temperature, current charging / discharging status (charging, discharging, standby), and availability status (e.g., online, offline, faulty, under maintenance). This table is updated at a very high frequency (e.g., every 10–30 seconds) to provide the most timely data support for subsequent power scheduling and balancing control.

[0068] Next, regarding the power scheduling and balancing control algorithm, the application server 8 has a built-in power scheduling and balancing control module 14, which serves as the decision-making core of the entire system. This module generates optimized scheduling instructions for the distributed mobile power bank group based on the node status table 13 and a preset complex scheduling strategy. The power scheduling and balancing control algorithm employs a hierarchical multi-objective optimization strategy, with rich and comprehensive inputs, including: real-time state of charge, health status, temperature, precise geographical location, historical usage data (such as average daily charge / discharge cycles, average charging time, and typical depth of discharge) of all online mobile power bank nodes, as well as demand predictions from external systems or artificial intelligence models (e.g., demand predictions for mobile power banks in a specific area over the next few hours, real-time availability information of nearby charging stations, traffic congestion conditions, etc.).

[0069] The multi-objective optimization strategy aims to simultaneously achieve the following four core objectives in order to optimize the overall system performance: State of Charge (SOC) equalization goal: To minimize the standard deviation of the SOC of all online mobile power nodes within the system. This means avoiding prolonged overcharging (above the safety limit, accelerating aging) or over-discharging (below the safety limit, damaging the battery) of some nodes, ensuring that all nodes operate within a relatively narrow SOC range, thereby significantly extending the average lifespan of the batteries throughout the system.

[0070] The goal of maximizing system available capacity is to maximize the number of mobile power nodes in the system that are in an available state (i.e., the state of charge is maintained within a safe range, such as 10% to 95%) and the total amount of electricity they can provide, while strictly satisfying the state of charge balance constraint. This is directly related to the system's external service capability.

[0071] Charging / Discharging Efficiency Optimization Goal: This goal aims to minimize energy loss during charging and discharging. The optimization algorithm comprehensively considers the battery aging level (health status) at different nodes, prioritizing charging and discharging of batteries with good health and low internal resistance, as these batteries have higher charging and discharging efficiency and generate less heat. Simultaneously, the algorithm avoids high C-rate charging and discharging, selecting an appropriate current to reduce Joule heat loss.

[0072] Cycle life balancing goal: During long-term operation, this invention aims to balance the number of charge-discharge cycles of each mobile power bank node. Through intelligent scheduling, it avoids overuse of some nodes leading to accelerated aging, while other nodes remain idle for extended periods. This strategy effectively extends the overall service life and return on investment of the entire mobile power bank system.

[0073] The specific execution steps of the power scheduling and balancing control module 14 include: Data Collection and Preprocessing: This module first obtains the latest operating parameters of all currently online mobile power supply nodes from the real-time database 11, and updates the state of charge, health status, and health rating (e.g., A, B, C levels, or a health status score of 0-100) of each node in real time based on the output of the state of charge and health status estimation module 12. During this stage, rigorous data cleaning and anomaly detection are performed. The data cleaning module identifies and removes erroneous data packets, duplicate data, or abnormal sensor readings that may occur during transmission (e.g., by setting thresholds or using statistical methods such as moving averages or median filtering). The anomaly detection module, based on historical data and preset statistical or machine learning models, determines in real time whether there are any abnormal behaviors at the individual mobile power supply node or the entire system level (e.g., drastic fluctuations in state of charge within a short period, abnormally high battery temperature, charging / discharging current exceeding the normal range, or prolonged communication interruption), and immediately triggers an alarm mechanism to notify the system administrator for manual intervention or automatic execution of emergency safety strategies.

[0074] Demand and resource matching: Based on the precise geographical distribution, real-time state of charge, and health status of all mobile power nodes in the current system, as well as potential demand predictions from external inputs (e.g., analyzing historical rental data and factors such as weather and holidays through machine learning models to predict the mobile power demand in a specific area within the next few hours), the module identifies low-power nodes that need charging, high-power nodes that can provide power, and nodes in standby or maintenance status. Furthermore, preferably, the power scheduling and balancing control algorithm incorporates geographical location information into the optimization decision. By performing cluster analysis on the geographical distribution of mobile power nodes (e.g., K-means clustering or DBSCAN clustering), the algorithm can identify "hotspot" areas with high power demand or excess power within a specific region. When mobile power nodes in a certain area are generally in a low-charge state, the scheduling module will prioritize instructing nearby high-charge nodes to move to that area (if the mobile power nodes have the ability to move automatically or be scheduled, such as by integrating an autonomous driving module), or instruct charging infrastructure (e.g., mobile charging vehicles or fixed charging piles) to tilt towards that area to meet local demand.

[0075] Building an optimization model: This is a multi-objective mathematical programming problem, and building the model is crucial.

[0076] Decision variables: For each mobile power node i in the system, define its decision variables for the next scheduling cycle. These variables are typically discrete or continuous, including: : Represents the charging state of node i, which can be an enumeration type (charging, discharging, standby) or represented by a set of binary variables. .

[0077] State of charge_target_i: The target state of charge for charging / discharging node i (a continuous variable, such as 0.1 to 0.95).

[0078] Rate_charge_i: The suggested charging rate for node i (a continuous variable, such as C-rate, in A).

[0079] Rate_discharge_i: The suggested discharge rate for node i (a continuous variable, such as C-rate, in A).

[0080] Objective function: The goal of the algorithm is to simultaneously optimize the following four interrelated functions, typically solved using weighted summation or Pareto optimization methods: State of charge equilibrium: ,in S is the real-time charge state of node i. This represents the average state of charge (SOC) of all online nodes. This term minimizes the squared difference between the SOC of all nodes and the average SOC to achieve high equilibrium.

[0081] Maximize system available capacity: ,in It is the current available capacity of node j (based on health status and current temperature correction). This is a binary variable indicating whether node j is in an available state (if the charge state is within the safe range and there are no faults, then...). =1, otherwise =0).

[0082] Optimization of charge and discharge efficiency: ,in It is the energy loss of node k during the next scheduling cycle due to factors such as internal resistance and temperature. According to calculate, It is the internal resistance of node k in its current charged state and healthy state.

[0083] Cycle life balancing: ,in It is the cumulative number of charge-discharge cycles at node m. This is the average number of cycles across all nodes. This term aims to minimize the deviation of the cycle count of each node from the average.

[0084] Constraints: Ensure the scheduling scheme is feasible within physical and security limitations: Safe range of state of charge: ,in and These are the lower and upper limits of the battery's safe state of charge, typically set at 10% and 95% to extend battery life.

[0085] Charge / discharge rate limit: as well as Ensure that the charge and discharge rates are within the battery's safe tolerance range (e.g., 0.2C to 1.0C) to prevent excessively fast charging and discharging from causing the battery to overheat, shorten its lifespan, or cause safety accidents.

[0086] Capacity is related to health status: The actual available capacity of a node decreases as its health status decays, and f is a capacity decay function based on health status.

[0087] The total energy of the system is conserved: ∑ChargePower=∑DischargePower+∑SystemLoad, ensuring that at the system level, the total power of charging and discharging, plus the system's own losses (if any) and external power supply requirements are kept in balance.

[0088] Communication latency threshold: CommunicationLatency≤Threshold, ensuring that the communication latency during the entire process from the application server issuing the scheduling instruction to the mobile power node receiving and starting execution is within an acceptable threshold (e.g., less than 5 seconds), to guarantee the timeliness and effectiveness of the scheduling instruction.

[0089] Solving optimization problems: For the above multi-objective optimization model, the module uses heuristic algorithms or exact solvers to solve the problems.

[0090] Heuristic algorithms: When the number of nodes is large and the problem complexity is high, genetic algorithms (GA) or particle swarm optimization (PSO) algorithms can be used. For example, in a genetic algorithm, each potential scheduling scheme is encoded as a "chromosome," and through operations such as selection, crossover, and mutation, a better scheduling scheme is iteratively evolved. The merits of each scheme are evaluated through a comprehensive fitness function.

[0091] Linear / Nonlinear Programming Solvers: For small to medium-sized systems or where higher computational overhead is permissible, specialized commercial solvers such as Gurobi, CPLEX, or the open-source SciPy optimization library can be used. These solvers can find global or local optima, but may have specific requirements for model types (linear, nonlinear, integer, etc.).

[0092] Generate scheduling instructions: Based on the optimal scheduling scheme output by the solver, generate specific scheduling instructions for each mobile power node. These instructions typically include the target state of charge (the state of charge the node should strive to achieve), the charging / discharging mode (explicitly informing the node whether to enter charging, discharging, or standby mode), and a suggested charging / discharging rate (e.g., a specific current value or C-rate). To minimize the LoRa transmission load and ensure transmission efficiency, the scheduling instructions are encoded in a concise binary format, using bit fields to compactly store each parameter.

[0093] Command issuance: The application server transmits the generated binary scheduling command to the LoRa communication module 4 of the target mobile power node through the LoRa network server 7 and LoRa gateway 6. The downlink communication mechanism of the LoRaWAN protocol (such as the RX1 / RX2 receive window of Class A devices, or the continuous listening mode of Class C devices) ensures the reliable delivery of the command.

[0094] Finally, regarding the execution and feedback of scheduling instructions, when the mobile power node receives a scheduling instruction from the LoRa communication network, the microcontroller unit 3 will parse and verify the instruction. Verification includes the instruction's legality (e.g., whether the node ID matches, whether the checksum is correct), validity (e.g., whether the timestamp has expired), and security (e.g., whether the target state of charge or charging / discharging rate exceeds a safe range). If the instruction is legal and valid, the microcontroller unit 3 will adjust the internal parameters of the battery management system 2 or control external charging / discharging devices (e.g., start or stop the built-in DC-DC converter, switch charging relays) via the SPI or I2C interface to accurately execute the scheduling instruction. The execution process includes, but is not limited to: starting or stopping charging, adjusting the charging current or charging voltage according to the instruction, starting or stopping discharging, and adjusting the discharging current or discharging load according to the instruction. During instruction execution, the mobile power node will monitor the battery pack status in real time and periodically send feedback information to the application server 8 based on the execution results. The feedback information typically includes the current real-time state of charge, the actual observed charge / discharge rate, the command execution status (success, failure, in progress), and any abnormal situation reports (such as over-temperature, over-voltage, communication interruption, etc.). Based on this real-time feedback information, the application server 8 evaluates the scheduling effect (e.g., whether the target state of charge has been achieved, whether the standard deviation of the state of charge has decreased), and dynamically adjusts subsequent scheduling strategies and optimization model parameters according to the evaluation results and the current system state, thereby forming an efficient closed-loop control system to ensure continuous optimization and stable operation of the system.

[0095] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely illustrative of the principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the present invention as claimed. The scope of protection of the present invention is defined by the appended claims and their equivalents.

Claims

1.A method for mobile power supply multi-node power scheduling and equalization control based on LoRa technology, characterized in that, Comprise the following steps: S1, mobile power node construction and data collection: build multiple mobile power nodes, each node contains battery pack (1), battery management system (2), microcontroller unit (3) and LoRa communication module (4); microcontroller unit (3) communicates with battery management system (2), real-time acquisition of battery pack (1) operating parameters and local processing, control the on-off of charge-discharge path; LoRa communication module (4) encapsulates the processed data into LoRa data packet through antenna (5), periodically sends to LoRa gateway (6), uses spread spectrum modulation technology, works in unlicensed ISM frequency band, supports adaptive data rate mechanism; S2, LoRa communication network construction and data transmission: build a communication network containing at least one LoRa gateway (6), LoRa network server (7) and application server (8); LoRa gateway (6) receives data packet and returns to LoRa network server (7), network server (7) processes data packet and forwards to application server (8), while managing network parameters; S3, application server data processing and state modeling: application server (8) stores and parses data packet, estimates the state of charge and health of each node in real time through state of charge and health estimation module (12) combined with battery model and machine learning algorithm, maintains node state table (13) and updates in real time; S4, power scheduling and balance control algorithm: power scheduling and balance control module (14) of application server (8) adopts hierarchical multi-objective optimization strategy, combines real-time state of node and external demand prediction to generate optimized scheduling instruction, the strategy takes minimizing standard deviation of state of charge, maximizing available node number and total power, minimizing energy loss and balancing charge-discharge cycle number as target; S5, scheduling instruction execution and feedback: after receiving the scheduling instruction, the mobile power node verifies it through the microcontroller unit (3), and executes the charge-discharge adjustment through the battery management system (2); the node monitors the state in real time and feeds back to the application server (8) periodically, the server evaluates the scheduling effect and dynamically adjusts the strategy to form a closed-loop control. 2.The method of claim 1, wherein, LoRa communication module (4) works in 915MHz or 868MHz frequency band, complies with LoRaWAN protocol stack, is configured as ClassA device, and only starts receiving mode in a short preset window period after data transmission to realize low power consumption. 3.The method of claim 1, wherein, Microcontroller unit (3) integrates SPI, I2C and UART low-power peripheral interfaces. 4.The method of claim 1, wherein, Battery management system (2) integrates millivolt-level precision voltage acquisition chip, milliamper-level precision current sampling chip and multi-channel temperature sensor interface, has passive or active balancing function, and can start balancing current cooperatively with microcontroller unit (3). 5.The method of claim 1, wherein, State of charge and health estimation module (12) adopts long short-term memory network deep learning model, takes battery voltage, current, temperature and historical charge-discharge data as input features, trains through actual running data, and updates model parameters through online learning or offline periodically. 6.The method of claim 1, wherein, The power scheduling and balancing control module (14) performs data cleaning and anomaly detection before executing the optimization algorithm: the data cleaning module eliminates errors, duplicates and abnormal data, and the anomaly detection module judges abnormal behavior based on historical data and preset thresholds and triggers an alarm. 7.The method of claim 1, wherein, The power scheduling and balancing control algorithm incorporates geographic location information, identifies regional power supply and demand status through cluster analysis, and when the node in a certain region is generally in a low state of charge, preferentially instructs nearby high state of charge nodes to move or tilt the charging infrastructure towards the region. 8.The method of claim 6, wherein, The power scheduling and balancing control module (14) uses heuristic algorithms such as genetic algorithms and particle swarm optimization algorithms or planning solvers such as Gurobi and CPLEX to solve the optimization model; the scheduling instructions are encoded in binary format, and the target state of charge, charging and discharging mode and rate parameters are stored in bit fields. 9.The method of claim 1, wherein, The mobile power supply node feedback information includes the current state of charge, actual charging and discharging rate, instruction execution status and abnormal reports such as over-temperature, over-voltage and communication interruption; the application server (8) evaluates the scheduling effect in real time based on the feedback, dynamically adjusts the scheduling strategy and model parameters. 10.The method of claim 1, wherein, The application server (8) configures a user interface module, provides a visual dashboard to display node geographic distribution, state of charge distribution and other information, generates periodic performance reports, and supports manual intervention in scheduling strategy and adjustment of optimization target weights.

Images