Cascadable multi-string lithium battery centralized maintenance management device and method

Abstract

The invention discloses a cascading multi-string lithium battery centralized maintenance and management device and a cascading multi-string lithium battery centralized maintenance and management method. The method comprises the following steps: during power-on, detecting an electrical sequence of each maintenance management single board in a cascade link, and distributing a unique board position identity identifier for each maintenance management single board; determining the first maintenance management single board in the cascade link as a master control single board and the other maintenance management single boards as slave control single boards based on the board position identification; obtaining a maintenance mode instruction, and cooperating with each slave control single board to maintain the accessed battery pack; and when any maintenance channel detects that the battery pack is connected, automatically controlling a charging loop or a discharging loop of the channel where the battery pack is located according to the maintenance mode instruction so as to adjust the electric quantity of the battery pack to a preset target electric quantity range. The technical problems of low manual maintenance efficiency, high cost and poor consistency of the lithium battery pack in the prior art are solved, automatic, large-scale and standardized maintenance of a large number of battery packs is realized, and the service life of the battery is prolonged.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of battery management, and particularly relates to a device and method for centralized maintenance and management of lithium batteries in multiple strings. BACKGROUND

[0002] Lithium batteries are widely used in medical devices, consumer electronics, power tools and other fields due to their high energy density, low self-discharge rate and long cycle life. In many application scenarios, especially in the medical field with extremely high reliability requirements, a large number of standby lithium battery packs are usually reserved to ensure uninterrupted operation of the equipment in the event of a main power failure or battery replacement. However, the electrochemical properties of lithium batteries determine that they are not suitable for long-term storage in a fully charged or depleted state. Studies have shown that if lithium batteries are left in a state of charge (SOC) close to 100% or below 20% for a long time, it will accelerate the side reactions of the internal electrode materials, leading to irreversible capacity decay and internal resistance increase, thereby significantly shortening the service life of the battery.

[0003] To solve this problem, the industry consensus is to maintain the power of lithium batteries within the 60%-80% range for long-term storage. This requires regular and active maintenance and management of the inventory of standby battery packs, i.e., discharging batteries with too high a power and charging batteries with too low a power. In the prior art, such maintenance work relies largely on manual operation, with technicians using independent charge and discharge instruments to detect and maintain battery packs one by one. This approach has the technical problems of low efficiency, high labor costs, cumbersome operation and difficulty in ensuring consistency of all battery maintenance states. SUMMARY

[0004] The purpose of the present application is to provide a device and method for centralized maintenance and management of lithium batteries in multiple strings, aiming to solve the technical problems of low efficiency, high cost and poor consistency of manual maintenance of lithium battery packs in the prior art.

[0005] In a first aspect, the present application provides a method for centralized maintenance and management of lithium batteries in multiple strings, comprising: upon power-up, automatically detecting the electrical sequence of a plurality of maintenance and management single boards in a cascading link, and assigning each maintenance and management single board a unique board position identity; based on the board position identity, determining the first maintenance and management single board in the cascading link as a master control single board, and the remaining maintenance and management single boards as slave control single boards; obtaining a maintenance mode instruction; detecting whether any battery pack is connected to any maintenance channel of any maintenance and management single board; when detecting that a battery pack is connected, automatically controlling the charging circuit or discharging circuit of the channel where the battery pack is located according to the maintenance mode instruction, to adjust the power of the battery pack to a preset target power range.

[0006] Optionally, the automatic detection of the electrical sequence of the plurality of maintenance management boards in the cascade link comprises: constructing a series voltage division network through the cascade link; each of the maintenance management boards acquires a node voltage value in the series voltage division network through an analog-to-digital conversion unit in a master control unit thereof; and determining the electrical sequence of the maintenance management boards based on the node voltage values.

[0007] Optionally, the master control board and the slave control board communicate through a controller CAN bus; the master control board broadcasts the maintenance mode instruction to the slave control board through the controller CAN bus, and polls the state of each maintenance channel on the slave control board; and the slave control board executes the maintenance mode instruction, and returns the state information of the maintenance channel thereof to the master control board through the controller CAN bus.

[0008] Optionally, the maintenance mode instruction comprises a long-term storage mode instruction or a battery activation mode instruction; in response to the long-term storage mode instruction, the preset target power range is set to 65% to 75% of the rated capacity of the battery; and in response to the battery activation mode instruction, the discharge circuit is first activated to discharge the power of the battery pack to a discharge cutoff threshold, and then the charging circuit is activated to supplement the power of the battery pack to the preset target power range.

[0009] Optionally, the automatic control of the charging circuit or the discharge circuit of the channel where the battery pack is located comprises: periodically acquiring a real-time voltage value of the battery pack; comparing the real-time voltage value with a target voltage interval corresponding to the preset target power range; closing the charging circuit when the real-time voltage value is lower than the lower limit of the target voltage interval; and closing the discharge circuit when the real-time voltage value is higher than the upper limit of the target voltage interval.

[0010] Optionally, the method further comprises: periodically acquiring a temperature value of each of the maintenance channels to construct a system-level real-time thermal map; when there are a plurality of maintenance tasks to be executed, preferentially assigning a maintenance task with a higher heat generation to a maintenance channel with a lower temperature based on the system-level real-time thermal map; and when a temperature value of any maintenance channel exceeds a preset temperature threshold, cooperatively controlling a heat dissipation fan on the maintenance management board where the channel is located and a heat dissipation fan on an adjacent maintenance management board thereof, and / or dynamically reducing the maintenance power of the channel.

[0011] Optionally, the method further includes: obtaining a unique identifier for the battery pack and creating or calling its corresponding digital twin model based on the unique identifier; before or after executing the control of the charging circuit or discharging circuit, obtaining its online health status data by injecting a preset electrical excitation signal into the battery pack and collecting its voltage response; updating the digital twin model based on the online health status data and evaluating the health status of the battery pack and predicting its remaining service life; generating a personalized maintenance strategy based on the results of the health status evaluation and remaining service life prediction, and adjusting the preset target power range according to the personalized maintenance strategy.

[0012] Secondly, this application also provides a cascaded multi-string lithium battery centralized maintenance management device, comprising: multiple cascaded maintenance management boards, each of the maintenance management boards comprising: a power management unit for receiving power from an external power source or a power source from a previous-level maintenance management board, and supplying power to each circuit of the board; a main control unit; and at least one maintenance channel, the maintenance channel comprising a battery interface, a charging circuit controlled by the main control unit, and a discharging circuit controlled by the main control unit; wherein the main control unit is configured to perform the method described in the first aspect.

[0013] Optionally, the multiple maintenance management boards are connected via a cascading interface, which is provided with an identification voltage sampling pin for determining the electrical sequence; each maintenance management board is provided with a voltage divider resistor, and the voltage divider resistors of all maintenance management boards together form a series voltage divider network after cascading; the analog-to-digital conversion input terminal of the main control unit is connected to the identification voltage sampling pin for collecting node voltage to determine the board position identification.

[0014] Optionally, the charging circuit includes a series of lithium battery charging management chips; the discharging circuit includes a power resistor and a power switch controlled by the pulse width modulation signal of the main control unit.

[0015] Optionally, it further includes: an identification unit for obtaining the unique identifier of the battery pack; and each of the maintenance channels further includes a micro-current excitation and precision sampling circuit for injecting an electrical excitation signal into the battery pack and collecting its response under the control of the main control unit; the main control unit is also configured to collaborate with the cloud computing platform to perform the steps of creating or calling a digital twin model, obtaining online health status data, performing health status assessment and lifespan prediction, and generating and executing personalized maintenance strategies. Attached Figure Description

[0016] Figure 1This is a schematic diagram of the system architecture of a cascaded multi-string lithium battery centralized maintenance and management device provided in an embodiment of this application, showing the topology of the main control board and multiple slave control boards connected through power cascading, ID identification and CAN bus.

[0017] Figure 2 This is a functional block diagram of a maintenance management board provided in an embodiment of this application, showing the connection relationship and signal flow between the power management unit, main control unit, communication unit and four independent maintenance channels inside the board.

[0018] Figure 3a This is a simplified circuit diagram of the discharge circuit and voltage detection circuit in a single maintenance channel provided in an embodiment of this application, showing the connection method of the power resistor, switching transistor, voltage divider network and main control unit.

[0019] Figure 3b This is a schematic diagram of a series voltage divider network for automatically determining electrical sequence provided in an embodiment of this application, showing how the ID identification voltage node is formed when three single boards are cascaded.

[0020] Figure 4 This is a simplified circuit diagram of the charging circuit in a single maintenance channel provided in an embodiment of this application, showing the connection method between the charging management chip and its peripheral components and the battery pack.

[0021] Figure 5 This is a flowchart illustrating a centralized maintenance and management method for cascaded multi-string lithium batteries provided in an embodiment of this application. It shows a preferred implementation, which includes steps for retrieving and restoring unfinished tasks to ensure the continuity of maintenance.

[0022] Figure 6 This is a macroscopic architecture diagram of an impedance-sequence fusion network (ISFN) model for battery health status assessment and remaining service life prediction in an optional embodiment of this application. It shows the dual-stream feature extraction architecture, fusion mechanism, and final prediction output layer of the model. Detailed Implementation

[0023] To make the objectives, technical solutions, and advantages of this application clearer, the technical solutions of this application will be clearly and completely described below in conjunction with the accompanying drawings and specific embodiments. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0024] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature. In the description of this application, "multiple" means two or more, unless otherwise explicitly specified.

[0025] System Implementation Examples Reference Figure 1 This application provides a cascaded multi-cell lithium battery centralized maintenance and management device. As a whole, the device's functional entity consists of an external switching power supply and at least one, typically multiple, maintenance and management boards interconnected in a cascaded manner. In a specific implementation, the device employs a modular and scalable hardware architecture, enabling flexible configuration of the system scale according to the number of lithium battery packs to be maintained. This device solves the technical problems of poor scalability and inability to centrally manage large numbers of batteries in existing battery maintenance equipment, achieving the beneficial effects of efficient, flexible, and automated battery pack maintenance.

[0026] Reference Figure 1 This application provides a cascaded multi-cell lithium battery centralized maintenance and management device. As a whole, the device's functional entity consists of an external switching power supply and at least one, typically multiple, maintenance and management boards 100 interconnected in a cascaded manner. In a specific implementation, the device employs a modular and scalable hardware architecture, enabling flexible configuration of the system scale according to the number of lithium battery packs to be maintained.

[0027] Figure 1 This cascaded structure is clearly demonstrated, where the main control board (ID=1) is connected sequentially to subsequent slave control boards (ID=2 to ID=n) via a "power cascading & ID identification" link and a "CAN bus" link, forming a unified management system. This device solves the technical problems of poor scalability and inability to centrally manage large numbers of batteries in existing battery maintenance equipment, achieving efficient, flexible, and automated battery pack maintenance.

[0028] The device includes multiple cascaded maintenance and management boards 100. Figure 1 The diagram illustrates n maintenance management boards, labeled 1, 2, ..., n. Each maintenance management board 100 is physically an independent printed circuit board (PCB) that integrates all the hardware circuitry required to perform battery maintenance functions. This modular design simplifies system deployment and maintenance; a failure in any one board does not affect the independent operation of the others, and system functionality can be restored simply by replacing the faulty module.

[0029] like Figure 2 As shown, each maintenance management board 100 includes a power management unit, a main control unit 200, at least one maintenance channel, and a communication unit. Power is supplied to the power management unit via a cascaded input interface. This unit outputs 3.3V to power the core logic circuit and transmits a 24V high voltage to the maintenance channel. The main control unit 200, as the core, controls the start / stop of the charging circuit (IP2365) via control signal lines (e.g., I2C or GPIO) and precisely controls the conduction of the MOSFETs in the discharge circuit via PWM (Pulse Width Modulation) signals. The voltage detection circuit in the maintenance channel sends the processed battery voltage signal to the ADC module inside the main control unit 200 for digitization via the ADC acquisition line. The main control unit 200 communicates bidirectionally with the human-machine interface unit (DIP switch / LED / buzzer), receiving user settings and outputting status indicators. The main control unit 200 connects to the CAN bus via the communication unit (CAN transceiver) to exchange data with other boards.

[0030] The power management unit is responsible for converting and distributing the input power. At the device level, an external switching power supply provides the main power, such as 24V DC, to the entire cascaded system. This 24V power is first input to the power input interface of the first maintenance management board 100 (board 1). The power management unit of board 1 receives this 24V power and, on the one hand, directly supplies it to loads on the board that require high voltage operation, such as charging circuits and cooling fans; on the other hand, through an onboard high-efficiency DC-DC buck converter, it converts the 24V voltage into a stable 3.3V DC power, specifically for powering the main control unit 200, communication unit, and other logic circuits. Simultaneously, the input 24V power is also passed intact to the power input interface of board 2 through the power cascade output interface of board 1, and then from board 2 to board 3, and so on, forming a 24V power bus to power all cascaded boards. This design simplifies system wiring, requiring only one main power supply to drive the entire system.

[0031] The main control unit 200 is the control core of each maintenance and management board 100. (Refer to...) Figure 2 In a preferred embodiment, the main control unit 200 employs an STMicroelectronics STM32F103 series microcontroller (MCU). This MCU, based on the ARM Cortex-M3 core, possesses powerful processing capabilities, abundant on-chip peripheral resources (such as a multi-channel 12-bit ADC, CAN controller, multiple timers, and GPIOs), and low power consumption. The main control unit 200 is electrically connected to all other functional units on the board via its pins, forming the control center of the entire board.

[0032] The maintenance channel is a functional unit that directly interacts with the battery pack to be maintained. (Refer to...) Figure 2 Each maintenance management board 100 is exemplarily equipped with four independent maintenance channels. Each maintenance channel has the same structure and function, and can independently and in parallel maintain one battery pack. Therefore, one board can maintain four battery packs simultaneously. Each maintenance channel specifically includes a battery interface, a charging circuit, a discharging circuit, and a voltage detection circuit.

[0033] The battery interface is the port through which the battery pack physically connects to the maintenance and management board 100. In one embodiment, the interface uses a universal XH2.54-4P female connector, which offers good compatibility and ease of insertion and removal, and can withstand relatively large voltage and current.

[0034] Reference Figure 3a The charging circuit, controlled by the main control unit 200, is used to replenish energy to battery packs with low power. In one specific implementation, the core component of this circuit is an Injoinic IP2365 multi-cell lithium battery fast charging management chip. This chip supports a wide voltage input (e.g., provided by stepping down 24V), can manage the charging process of 1 to 4 cells of lithium batteries, and can provide a charging current of up to 3A. The main control unit 200 communicates with the IP2365 chip through a specific GPIO or I2C interface, and can control the start and stop of the charging process and may configure charging parameters.

[0035] The charging circuit, controlled by the main control unit 200, is used to replenish energy to battery packs with low power. (Refer to...) Figure 4This diagram illustrates the core circuitry of the charging loop, which is essentially a high-efficiency synchronous buck converter. An external 24V DC power supply serves as the input, first passing through an input filter capacitor Cin (10μF in the example) to stabilize the input voltage, and then being sent to the VIN pin of the IP2365 charging management chip. The chip's internal switching regulator outputs a square wave voltage at an extremely high frequency through its SW (switch) pin. This square wave voltage drives an LC low-pass filter consisting of a power inductor (4.7μH in the example) and an output filter capacitor Cout (22μF in the example). This LC filter smoothly converts the high-frequency square wave voltage into a stable DC voltage, the output node of which is connected to the positive terminal of the lithium-ion battery pack, providing it with charging energy. The BAT pin of the IP2365 chip is also connected to this output node for accurately sampling the real-time voltage of the battery, serving as feedback for internal charging strategy control (e.g., constant current-constant voltage mode switching). The negative terminal of the lithium-ion battery pack is connected to the circuit's common ground (GND). The main control unit 200 communicates with the IP2365 chip through specific control signals (not shown in the figure) and can control the start and stop of the entire charging process.

[0036] The discharge circuit is controlled by the main control unit 200 and is used to discharge energy from the battery pack with excessive discharge. (Refer to...) Figure 3a The diagram illustrates the detailed connections of the discharge circuit. The positive terminal of the lithium battery is connected to a high-power discharge resistor, the other end of which is connected to the drain of an N-channel MOSFET. The source of the MOSFET is grounded, and its gate is grounded through a pull-down resistor and connected to the PWM control pin of the main control unit 200. When discharge is required, the main control unit outputs a high-level PWM signal, turning on the MOSFET and forming a complete circuit of "battery positive terminal -> discharge resistor -> MOSFET -> ground". The battery energy is dissipated as heat through the resistor.

[0037] The voltage detection circuit is composed of a precision resistor voltage divider network. For example... Figure 3a As shown, this network is used to safely provide the real-time voltage of the battery pack to the main control unit for measurement. It consists of two precision resistors, R1 (100kΩ in the example) and R2 (20kΩ in the example), connected in series. This resistor string is connected between the positive and negative terminals of the lithium battery. The connection point of R1 and R2, i.e., the voltage divider point, is connected to the ADC input pin of the main control unit (MCU ADC). Since the input voltage range of the ADC is limited (e.g., 0-3.3V), this voltage divider network can proportionally reduce the battery voltage (e.g., 3.0V-16.8V) to a safe range, thereby achieving accurate and safe acquisition of the battery voltage.

[0038] The communication unit is used to realize information interaction between multiple maintenance and management boards 100. In a preferred embodiment, communication adopts Controller Area Network (CAN) bus technology. Each board integrates a CAN transceiver chip (such as TJA1050). The CAN controller integrated inside the main control unit 200 is responsible for data packaging and parsing at the protocol layer, and its TX and RX pins are connected to the CAN transceiver; the CAN transceiver is responsible for physical layer level conversion, converting the MCU's logic signals into differential signals (CAN_H and CAN_L) on the CAN bus. The CAN_H and CAN_L of all boards are connected in parallel through a data cascading interface to form a CAN bus, realizing the interconnection between all main control units 200.

[0039] In addition, the maintenance management board 100, which is designated as the main control board, is also equipped with a human-machine interface unit. Referring to Figure 3, this unit may include a DIP switch (labeled "three months / six months") for setting the maintenance mode, a button for starting and stopping tasks (labeled "power on / off"), a buzzer for audible prompts, and multiple LEDs for status indication. Users can control the entire cascaded system simply by operating the main control board.

[0040] Method Implementation Examples Reference Figure 2 This application also provides a centralized maintenance and management method for cascaded multi-string lithium batteries based on the above-mentioned device. The following will use a system comprising three cascaded maintenance and management boards as an example to illustrate each step of the method in detail.

[0041] S100: Upon power-up, automatically detect the electrical sequence of multiple maintenance management boards in the cascaded link and assign a unique board position identifier to each maintenance management board.

[0042] In this embodiment, this step is accomplished through a hardware self-identification mechanism that requires no manual intervention. Its core is to utilize cascaded links to construct a series voltage divider network to determine the electrical order of each board in the cascaded topology, rather than its absolute physical spatial location. Each maintenance management board 100 has a pre-set precision voltage divider resistor for ID identification and an identification voltage sampling pin (ID Pin) defined on its cascade interface. When multiple boards are connected via cascaded cables, their ID Pins and voltage divider resistors are automatically connected in series.

[0043] The core of this step is to construct a series voltage divider network using cascaded links. (Refer to...) Figure 3bThis diagram details a voltage divider network schematic with three cascaded boards. On the first board (the starting point of the electrical link), a 3.3V reference power supply is connected to Node 1 via a pull-up resistor R_up (example value 1kΩ). Node 1 is the ID sampling point of the first board, and its voltage is sampled by ADC1 on this board. Node 1 is also connected to the second board via a cascaded cable. Inside the second board, a voltage divider resistor R_div (example value 10kΩ) forms a voltage divider with the voltage from the previous stage, resulting in Node 2, whose voltage is sampled by ADC2 on the second board. Similarly, Node 2 is connected to the third board via a cascaded cable, and its voltage is divided by another R_div resistor inside the third board to form Node 3, which is sampled by ADC3. Thus, from Node 1 to Node 3, the voltage decreases sequentially due to the staged voltage division. Each maintenance and management board 100's main control unit 200 can determine its own electrical sequence by collecting the voltage values ​​of its respective nodes (Node 1, 2, 3) and assign itself a unique board ID.

[0044] Specifically, such as Figure 3b As shown, on the first board (the starting point of the electrical link), its ID pin is connected to a 3.3V reference power supply via a pull-up resistor. The voltage of this ID pin is then passed as input to the ID pin of the second board. Inside the second board, its ID pin passes through a voltage divider resistor to ground before being passed to the third board. Similarly, subsequent boards repeat this structure. In this way, from the first board to the last board, the voltage on the ID pin decreases sequentially due to the step-by-step voltage division, forming a stepped voltage gradient.

[0045] Each maintenance and management board 100's main control unit 200 uses its built-in 12-bit analog-to-digital converter (ADC) to acquire the voltage value of its own board's ID pin. Because each board has a different electrical sequence, the voltage values ​​they acquire are unique and distinguishable. The firmware of the main control unit 200 pre-defines a voltage range-ID mapping table. During power-on initialization, the main control unit 200 reads the digital value converted by the ADC, then consults the mapping table to determine its board number and assigns itself a unique board ID, such as ID=1, ID=2, ID=3.

[0046] For example, assume the ADC reference voltage of the main control unit 200 The voltage is 3.3V (dimensionless), the ADC has 12-bit precision, and its output digital value range is 0 to 4095 (dimensionless). The first board's pull-up resistor... Set to 1kΩ, the voltage divider resistor to ground for each board. The ADC is set to 10kΩ. In an ideal model, the ADC reading of the first board is close to 4095, the second board is close to 2048, and the third board is close to 1365.

[0047] However, in actual physical implementation, the ideal model described above is subject to various non-ideal factors, which are crucial to the robustness of this self-identification mechanism. These factors mainly include: 1) Component tolerances: The voltage divider resistors themselves have manufacturing tolerances of ±1% or even higher, which directly lead to voltage drift at each node. 2) Connector and cable impedance: The cascaded cables and connectors between boards have non-negligible contact and line resistances, which introduce additional series voltage drops, especially when the cascaded link is long, the voltage of the end board will be lower than the theoretical value. 3) ADC sampling error: The ADC of the main control unit 200 itself has quantization errors, nonlinear errors, and reference voltage drift. To address these complex real-world challenges, the core design of this solution does not rely on precise voltage points, but instead uses a wide voltage window with sufficient guard intervals to determine the ID.

[0048] Specifically, the voltage-ID mapping table in the firmware is designed as a series of non-overlapping voltage ranges. For example, the full scale of the ADC corresponds to 3.3V.

[0049] The threshold range for ID=1 is set to [3.00V, 3.30V].

[0050] The threshold range for ID=2 is set to [2.20V, 2.80V].

[0051] The threshold range for ID=3 is set to [1.60V, 2.00V].

[0052] It can be seen that there is a clear "guard band" between the judgment intervals of adjacent IDs. For example, there is a guard band of (2.80V, 3.00V) between ID=1 and ID=2. As long as the total voltage offset caused by all non-ideal factors does not exceed the width of this guard band, the accuracy of ID recognition can be guaranteed. S200: Based on the board position identification, determine the first maintenance management board in the cascaded link as the master control board, and the remaining maintenance management boards as slave control boards.

[0053] After S100 is completed, each maintenance and management board 100 has obtained its own unique Board ID. Then, the master control unit 200 of all boards will execute a preset arbitration logic: the board with Board ID 1 will be defined as the system's unique master board, while all other boards with IDs (ID>1) will automatically enter slave board mode.

[0054] The main control board is responsible for human-computer interaction, task scheduling, global status monitoring, and communication with the host computer. When the human-computer interaction units (buttons, DIP switches) on the board are activated, user commands are input only through this board.

[0055] The slave control board disables its local human-machine interaction function. Its role is to receive and execute instructions from the master control board and report the status of the maintenance channel it is responsible for to the master control board.

[0056] The process of establishing this master-slave architecture is fully automated. Subsequently, all boards initialize their CAN communication units. The master control board periodically broadcasts heartbeat packets or polling commands via the CAN bus to confirm that all slave control boards are online and establish communication with them.

[0057] For example, the master control board (ID=1) constructs a CAN message with a frame ID that can be defined as 0x100 (broadcast address or master-specific address). The data field contains an instruction code to query the status of all slave boards. This message is sent to the CAN bus. The CAN controllers of slave control boards 2 (ID=2) and 3 (ID=3) are both configured with filters and will receive this message. Upon receiving it, they will each construct a response message. For example, the response message frame ID of slave control board 2 can be defined as 0x201 (including its own ID), and its data field contains its current status (such as "idle"); the response message frame ID of slave control board 3 can be 0x301. By listening to these specific response frame IDs, the master control board can confirm the online status of the slave nodes and collect their initial information, completing the topology discovery and communication link establishment of the entire cascaded system.

[0058] S210: After power-on, retrieve and resume any unfinished maintenance tasks.

[0059] After determining the master and slave board roles and establishing a communication link in the S200, the master control board immediately executes a crucial initialization step: retrieving and resuming any maintenance tasks interrupted due to unexpected power outages. This step is designed to ensure the entire maintenance management device has task continuity and state memory capabilities. It solves the technical problem of losing maintenance progress due to grid fluctuations or unexpected power outages in long-term, unattended maintenance scenarios, requiring a reassessment of all batteries. This achieves the beneficial effect of improving system reliability and maintenance efficiency.

[0060] In a specific implementation, each maintenance management board 100's main control unit 200 integrates a non-volatile memory, such as a 256Kbit EEPROM chip or the Flash data storage area integrated within the STM32F103 microcontroller. This non-volatile memory is used to persistently store key task parameters for each maintenance channel. The main control board is responsible for coordinating the task status storage of the entire system.

[0061] When any channel (whether on the master control board or the slave control board) begins to execute a maintenance task (i.e., after entering step S500), the master control unit 200 of its board immediately creates a task record for that channel in non-volatile memory. This task record is a structured dataset containing all the information required to recover the task. This dataset includes at least: a task status identifier (e.g., 0x01 represents charging, 0x02 represents discharging), the real-time battery voltage value (accurate to millivolts), the task execution time (accurate to seconds), and the maintenance mode (e.g., long-term storage mode or activation mode).

[0062] During maintenance tasks, the main control unit 200 responsible for the channel refreshes the latest task parameters to non-volatile memory at a preset period (e.g., every 30 seconds). This periodic saving strategy strikes a balance between system reliability and memory write / erase lifespan, ensuring that even in the event of a sudden power outage, the lost maintenance progress is at most 30 seconds.

[0063] Upon power-up, during step S210, the master control board (ID=1) first reads the task records from its local non-volatile memory. Simultaneously, it sends a "read task records" command to all identified slave control boards via the CAN bus. Upon receiving the command, each slave control board reads its own local non-volatile memory and packages all task records with a "in progress" status into a CAN message, sending it back to the master control board.

[0064] After collecting all interrupt task information from the entire cascaded system, the master control board reconstructs the task queue based on this information. It sends a "restore task" command to the corresponding slave control board (via Board ID) and channel number. This command contains complete parameters such as the battery voltage and task status saved last time before the power outage. Upon receiving the command, the board immediately initiates the maintenance process for the corresponding channel, but not from the beginning. Instead, it directly enters the S550 step (a special recovery step), seamlessly continuing charging or discharging operations based on the received parameters until the target voltage range is reached.

[0065] For example, suppose a 4S battery pack located on channel 2 of slave control board 3 (ID=3) is performing a discharge operation in "long-term storage mode" before power failure. Its target voltage range is [15.20V, 15.60V], and the real-time voltage before power failure is 15.85V. Five seconds before power failure, the master control unit 200 of board 3 has written a task record to its EEPROM, which is: {Channel: 2, Status: 0x02 (Discharging), Voltage: 15.85V, ElapsedTime:125s, Mode: Storage}. After the system is powered on again, master control board 1 queries the record and sends a command to board 3: "Resume channel 2 task, current voltage 15.85V, target range [15.20V, 15.60V], continue discharging." Channel 2 of board 3 then activates the discharge circuit, continuing to reduce the voltage from 15.85V, instead of re-judging from the battery's actual current voltage (which may have slightly increased due to rest), thus ensuring the continuity of the maintenance logic.

[0066] S300: Obtain a maintenance mode command.

[0067] After the system completes initialization and enters standby mode, it needs to acquire a clear maintenance mode instruction, which will serve as the global policy benchmark for all subsequent maintenance tasks. The acquisition method for this instruction is flexible and can adapt to different application scenarios and levels of automation. In a specific implementation, the acquisition method for this maintenance mode instruction includes, but is not limited to, local human-machine interaction input, remote instruction issuance from a host computer, or adaptive policy selection based on battery type.

[0068] In a preferred embodiment, the maintenance mode command is directly input by the user through a human-machine interface unit located on the main control board. Referring to Figure 3, a two-position DIP switch (labeled "three months / six months") is provided on the PCB of the main control board, and its switch state is read by the GPIO pin of the main control unit 200. When the user closes the first switch, the main control unit 200 interprets the maintenance mode command as "long-term storage mode". In this mode, the preset target capacity range is set to 65% to 75% of the battery's rated capacity, which corresponds to a voltage range most suitable for long-term healthy battery storage. When the user closes the second switch, the command is interpreted as "battery activation mode". In this mode, the system will perform a complete deep discharge-charge cycle on the connected battery, that is, first discharge to the discharge cutoff threshold, and then charge to the aforementioned long-term storage target capacity range to restore the battery's electrochemical activity.

[0069] In another optional embodiment, the maintenance mode command is remotely issued by a host computer or central control server connected to the device's CAN bus. In this configuration, the entire maintenance device acts as an execution node, and its local human-machine interface unit can be disabled. The host computer can dynamically send CAN messages to the main control board based on the global battery inventory management strategy. For example, the host computer can send a CAN message with frame ID 0x101, whose data field contains instruction codes, such as 0x01 representing "switch to long-term storage mode" and 0x02 representing "switch to activation mode." After the main control board's CAN communication unit receives and parses the message, it can update the global maintenance mode command. This approach is particularly suitable for large-scale, centralized battery management centers.

[0070] In another alternative embodiment, the system can even implement adaptive mode selection. For example, the battery pack itself can carry identification information (such as an identification resistor with a specific resistance value). When the maintenance channel detects a battery connection, it measures this resistance value via a dedicated pin to identify the battery model or a preset maintenance strategy. The main control unit 200 can have a pre-set mapping table of "battery model - maintenance mode". For example, when the detection resistance is 10kΩ, the mapping table is consulted to determine that it is a model A battery, and the "long-term storage mode" should be used; when the resistance is 20kΩ, it is determined that it is a model B battery that has not been used for a long time, and the "battery activation mode" should be automatically used. This approach achieves the highest degree of automation and reduces the possibility of human error.

[0071] Regardless of the acquisition method, the master control board will store the final maintenance mode instruction in its internal RAM and broadcast the instruction to all slave control boards via the CAN bus, ensuring a high degree of consistency in the working strategy of the entire cascaded system.

[0072] S400: Detects whether a battery pack is connected to any maintenance channel on any maintenance management board.

[0073] After the maintenance mode is set, the entire system enters an automatic detection and waiting state. At this time, the master control units 200 of all maintenance management boards (including master and slave control) will independently, in parallel and periodically scan the four maintenance channels they are responsible for.

[0074] The core mechanism of the detection is the use of a voltage detection circuit. The voltage detection circuit of each maintenance channel is always connected to the positive and negative terminals of the battery interface. The main control unit 200 reads the voltage division values ​​of all four channels through its ADC module at a fixed frequency (e.g., every 100 milliseconds).

[0075] When no battery pack is inserted into a channel, its battery interface is in an open circuit state, and the voltage measured on the voltage detection circuit is theoretically 0V, or a very low floating voltage. The firmware of the main control unit 200 sets a battery presence threshold voltage. For example, the voltage corresponding to a single battery cell is 2.5V.

[0076] When a battery pack is inserted into the battery interface of a certain channel, the actual voltage of the battery pack will immediately appear on the voltage detection circuit of that channel. After the ADC of the main control unit 200 acquires this voltage value, it calculates that it is greater than... To prevent false alarms caused by contact bounce during insertion and removal, the firmware also includes software debouncing logic. For example, the main control unit 200 must detect a voltage higher than [a certain value] five consecutive times (i.e., within 500 milliseconds). Only then can it be confirmed that the channel has a stable battery pack connection.

[0077] Once access is confirmed, the main control unit 200 records the channel number and access event, and illuminates the corresponding status indicator LED (e.g., changing from off to solid blue, indicating "identified, ready to go"). If the board is a slave board, it immediately sends a status update message to the main control board via the CAN bus, containing: "Board ID=X, Channel Y, Event=Battery Inserted, Voltage=Z.ZZ V". Upon receiving this information, the main control board updates its maintained global status table.

[0078] S500: When a battery pack is detected to be connected, the charging circuit or discharging circuit of the channel where the battery pack is located is automatically controlled according to the maintenance mode instruction to adjust the power of the battery pack to a preset target power range.

[0079] After the S400 confirms battery connection, the corresponding maintenance channel immediately enters the automatic maintenance process. This process is a closed-loop control process, consisting of the following sub-steps.

[0080] S510: Periodically collect the real-time voltage value of the battery pack.

[0081] The main control unit 200 continuously monitors the battery pack's terminal voltage by passing through the voltage detection loop of this channel at a higher frequency (e.g., 10 times per second).

[0082] For example, suppose the battery to be maintained is a 4-cell (4S) lithium-ion battery pack with a full-charge voltage of 16.8V and a discharge cut-off voltage of 12.0V. The voltage sensing circuit consists of a 100kΩ resistor. and a 20kΩ resistor Series connection. ADC reference voltage. It is 3.3V.

[0083] When the real-time voltage of the battery pack When the voltage is 15.0V, the voltage at the ADC input terminal for .

[0084] The digital value read by the 12-bit ADC of the main control unit 200 for .

[0085] The algorithm inside the main control unit 200 performs reverse calculations, from... Calculate the actual battery voltage: In this way, the main control unit 200 can accurately and in real time monitor the battery pack's power status.

[0086] S520: Compare the real-time voltage value with a target voltage range corresponding to the preset target power range.

[0087] Based on the maintenance mode set in S300, the main control unit 200 will determine a target voltage range. .

[0088] In "long-term storage mode," the target charge range is 65%-75%. For a 4S battery pack, this likely corresponds to a voltage range of 15.2V to 15.6V. Therefore, , .

[0089] The main control unit 200 will use the real-time voltage measured in S510 Compare with this interval.

[0090] S530: Based on the comparison results, close the charging circuit or the discharging circuit.

[0091] The logical judgment for this step is as follows: when When the battery level is low, it indicates that the battery needs to be charged. The main control unit 200 will send a command to the charging management chip IP2365 to start charging. At the same time, the status indicator LED of this channel will change to a specific mode, such as a red breathing flash, indicating "charging in progress".

[0092] when When the battery level is high, it indicates that the battery needs to be discharged. The main control unit 200 will output a high-level signal through its GPIO pin to drive the gate of the power MOSFET, turning it on. The battery pack current begins to flow through the high-power discharge resistor, discharging. At the same time, the status LED will change to another mode, such as a green breathing flash, indicating "discharging".

[0093] when When the battery level is within the ideal storage range, the maintenance task is complete. The main control unit 200 will ensure that both the charging and discharging circuits are disconnected and drive the status LED to turn solid green, indicating "maintenance complete". At the same time, the buzzer on the board may beep briefly several times to alert the user.

[0094] For example, suppose a newly connected 4S battery pack has an initial voltage of The voltage is 16.2V. The system is currently in "long-term storage mode" with a target range of [15.2V, 15.6V].

[0095] The main control unit 200 measured 16.2V and discovered... Therefore, it immediately initiates the discharge circuit. Assuming the discharge resistance is 10Ω, the discharge current at this time is approximately The discharge power is The main control unit 200 will continuously monitor the voltage as the discharge progresses. It will decrease slowly. When detected When the voltage drops to 15.6V or slightly below that value (considering control hysteresis), the main control unit 200 will immediately shut down the discharge circuit, and the maintenance will be completed.

[0096] For example, another connected battery pack had an initial voltage of 14.0V. The main control unit 200 detected this. The charging circuit is then activated. The IP2365 chip begins charging the battery in constant current mode (e.g., 2A). As charging progresses, the voltage gradually increases. When the voltage reaches 15.6V, the main control unit 200 stops charging, and maintenance is complete.

[0097] Throughout the maintenance process, the single-board unit reports its channel status (such as "charging," "discharging," "completed," and "fault") and key data (real-time voltage, current, temperature, etc.) to the main control board in real time via the CAN bus. Users can view the maintenance status of all batteries in the cascaded system on the main control board, and can remove batteries that have completed maintenance at any time, or insert new batteries to be maintained. The system will automatically recognize and start a new task, achieving a high degree of automation and parallel processing capabilities. When faults such as charging timeout or abnormal temperature occur, the system will automatically stop the maintenance of the corresponding channel and alert the user with audible and visual alarms.

[0098] To further improve maintenance efficiency and maximize the cycle life of the battery pack, this application also provides an adaptive maintenance method based on thermal collaborative optimization. This method, building upon the hardware architecture and master-slave communication mechanism constructed in the above embodiments, introduces global perception and collaborative control of the system's thermodynamic state.

[0099] This method aims to address a deeper technical problem: in large-scale, high-density battery maintenance scenarios, high-power charge-discharge operations (especially battery activation modes) generate significant heat. The accumulation of localized heat not only accelerates the chemical aging of the battery itself but also affects adjacent batteries on the same board through heat conduction and convection, forming "hot spots" that limit the parallel processing capability and maintenance speed of the entire system. This method employs a proactive, globally coordinated thermal management strategy, elevating the maintenance process from simple electrical state management to a collaborative optimization of electro-thermal states, thereby maximizing system maintenance throughput while ensuring battery safety and longevity.

[0100] To implement this method, in addition to the aforementioned circuit, each maintenance management board 100 needs to have at least one temperature sensor (e.g., an NTC thermistor) installed near the battery interface of each maintenance channel, and its signal is connected to the ADC channel of the main control unit 200.

[0101] This adaptive maintenance method can be executed by the main control board and includes the following steps: S610: Build and dynamically update system-level real-time heatmaps.

[0102] After establishing communication with all slave control boards, the master control board (ID=1) periodically (e.g., every 5 seconds) sends a "report thermal status" command to all boards (including itself). Upon receiving the command, the master control unit 200 of each board immediately collects the readings of the temperature sensors near all maintenance channels under its responsibility. Subsequently, each slave control board sends a message containing its own Board ID and the temperature values ​​of each channel back to the master control board via the CAN bus.

[0103] After collecting all temperature data, the main control board builds and updates a system-level "thermal map" in its memory in real time. This thermal map is a data structure that records the current temperature of each battery channel in the entire cascaded device.

[0104] For example, for a system containing 3 boards (4 channels per board), its heatmap at a certain moment might look like this: ThermalMap = {B1:[28.5, 29.1, 28.8, 28.6], B2:[34.1, 35.2, 34.8, 33.9],B3:[27.9, 28.2, 28.1, 28.3]} (unit: degrees Celsius). This heatmap visually shows that the second board (possibly located in the middle of the rack, where ventilation is poor) is currently the hot spot area of ​​the system.

[0105] S620: Priority scheduling of tasks based on heatmaps.

[0106] When a new maintenance task arises (for example, a battery requiring high-power discharge is inserted), the main control board does not simply follow the "first come, first served" principle. Instead, it performs intelligent task scheduling based on the real-time heat map built by the S610. The core principle of scheduling is to prioritize tasks that generate high heat (such as discharging or high-current charging) and assign them to the channels and boards with the lowest current temperature.

[0107] For example, suppose there are two "activation mode" discharge tasks to be executed in the system. The main control board, after reviewing the aforementioned heat map, finds that the average temperature of the third board (B3) (approximately 28.1°C) is significantly lower than the average temperature of the second board (B2) (approximately 34.5°C). Therefore, even if a channel on B2 becomes available first, the main control board will choose to wait or prioritize allocating these two high-heat-generating tasks to the available channel on B3 to avoid exacerbating the hotspot problem on B2, thereby maintaining the thermal balance of the entire system.

[0108] S630: Performs cooperative fan control and dynamic power throttling.

[0109] Based on the heat map, the main control board performs unified and coordinated control of the heat dissipation resources (fans on each board) and task power of the entire system.

[0110] When the main control board detects that the temperature of any channel on a board (e.g., B2) exceeds a preset first-level threshold (e.g., 35°C), it not only sends a command to B2 to increase its fan speed to 100%, but also sends commands to B2's physically adjacent boards (B1 and B3) to run their fans at an auxiliary speed (e.g., 30%). This is not to cool B1 and B3, but to actively create airflow around B2, forming a "heat dissipation channel," thereby more effectively helping the hot board B2 dissipate heat. This solves the problem of low heat dissipation efficiency of a single fan due to airflow obstruction.

[0111] If the temperature in the hotspot area continues to rise after the coordinated fan control is activated and exceeds the more dangerous secondary threshold (e.g., 40°C), or if the average temperature of the entire system exceeds the environmental safety limit, the main control board will activate the dynamic power throttling mechanism. It will send instructions to the boards performing high-heat-generating tasks (especially those in the hotspot area), requiring them to temporarily and proportionally reduce the task power.

[0112] For example, the main control board sends a command to B2: "Reduce the discharge current of all currently executing discharge tasks (achieved by adjusting the PWM duty cycle) by 50%." This will reduce heat generation power by approximately 75%, thus buying the system valuable time to cool down. Once the temperature returns to a safe range, the main control board will then instruct it to resume full-power operation.

[0113] Through the closed-loop collaborative control of S610 to S630, this method integrates the previously independent maintenance boards into an intelligent organism with environmental awareness and adaptive adjustment capabilities. It no longer passively responds to over-temperature faults, but actively and predictively manages the thermal distribution of the entire system, enabling it to complete the maintenance tasks of large-scale battery packs under any environmental conditions with near-maximum efficiency while ensuring absolute safety and battery longevity.

[0114] In an optional implementation, this application also provides a predictive maintenance method based on digital twins and online health assessment. In a specific implementation, this method employs an edge-cloud collaborative computing architecture, an online electrochemical impedance spectroscopy feature acquisition mechanism, and an impedance-sequence fusion network model specifically constructed for this application. This enables the online assessment of the health status of the battery pack and prediction of its remaining service life during routine electrical maintenance. This method solves the technical problem that existing maintenance strategies cannot identify individual battery health differences and can only perform passive state restoration, achieving the beneficial effect of refined and forward-looking full life-cycle health management of battery assets.

[0115] The hardware foundation upon which this method relies is the maintenance management board 100 and the cascaded system comprised of it described in the preceding embodiments. Based on this, the method implements complex predictive maintenance functions through an edge-cloud collaborative computing architecture. This architecture clearly defines the functional division between the field hardware devices and the backend computing platform.

[0116] The main control unit 200 (STM32F103) on the maintenance management board 100 is defined as an edge sensing and execution node in this architecture. Its core function is limited to performing physical layer interaction tasks with the highest real-time requirements. These tasks include high-frequency generation of minute current perturbation signals for electrochemical impedance spectroscopy measurements and high-speed synchronous sampling of battery voltage response. In one embodiment, the main control unit 200 uses its internal advanced timer and digital-to-analog converter to generate excitation signals of specific waveforms and uses its analog-to-digital converter supported by a direct memory access controller to temporarily store the acquired raw time-domain sampling sequence in its on-chip random access memory. After acquisition, the main control unit 200 performs preliminary data alignment and packaging on the raw time-domain sampling sequence and sends this message, containing the raw dataset, battery unique identifier, and timestamp, to the main control board via the controller area network bus.

[0117] Correspondingly, a host computer or cloud platform connected to the cascaded system via a controller area network (CLAN) bus is defined as the cloud-based digital twin and intelligent decision-making center. This center is responsible for executing all computationally intensive and storage-intensive tasks. These tasks include performing Fast Fourier Transform (FFT) operations on raw time-domain datasets received from edge nodes to generate frequency-domain impedance fingerprints, managing the database of the full lifecycle digital twin model for each battery, and running machine learning models for health status assessment and remaining lifespan prediction. After completing the calculations, the center generates specific, personalized maintenance strategy instructions and distributes them to the main control board via the CLAN bus for execution. This collaborative architecture separates real-time physical interaction from complex intelligent analysis, enabling predictive health management on existing management devices.

[0118] This predictive maintenance method can be executed by the edge sensing and execution node in collaboration with the cloud-based digital twin and intelligent decision-making center, and its specific steps include S710, S720, S730 and S740.

[0119] S710: Performs battery identification and creates or wakes up its digital twin model.

[0120] When a battery pack is detected entering a maintenance channel, the system first obtains a unique, permanent identification ID (UID) for that battery pack. In one embodiment, this ID is obtained by reading a QR code on the battery casing using a miniature image sensor positioned next to the maintenance channel. In another embodiment, the ID is obtained by interacting with an electronic tag built into the battery using a near-field communication (NFC) reader. After obtaining the ID, the main control board reports it to the cloud center. The cloud center retrieves the ID from its database. If the retrieval fails, the battery is determined to be entering for the first time, and a completely new digital twin model is created for it. This model is initialized with the battery's nominal parameters and an initial health assessment model based on statistical data from batteries of the same model. If the retrieval is successful, the existing digital twin model of the battery is activated, and all its historical data and the latest health assessment model are loaded.

[0121] S720: Performs online electrochemical impedance spectroscopy feature acquisition.

[0122] After starting or completing a routine charge-discharge maintenance task, the main control unit 200 will drive an additional micro-current excitation and precision sampling circuit to perform an online health status assessment of the battery.

[0123] The circuit, as a functional unit, is connected in parallel across the two ends of the battery pack under test. The detailed electrical connections of its internal components are organized into a current excitation path and a precision sampling path.

[0124] The micro-current excitation and precision sampling circuit includes a current excitation path configured to inject a precisely controllable excitation current into the battery pack. In one specific implementation, a digital-to-analog converter output pin of the main control unit 200 is electrically connected to the non-inverting input of an operational amplifier. The output of the operational amplifier is electrically connected to the gate of an N-channel power MOSFET. The drain of the power MOSFET is electrically connected to the circuit's common ground, and its source is electrically connected to the first terminal of a precision current sampling resistor. The second terminal of the precision current sampling resistor is electrically connected to the positive terminal of the battery pack under test. To construct a voltage-controlled constant current source, a node connected between the source of the power MOSFET and the first terminal of the precision current sampling resistor is fed back to the inverting input of the operational amplifier. The negative terminal of the battery pack under test is electrically connected to the circuit's common ground.

[0125] The micro-current excitation and precision sampling circuit further includes a precision sampling path configured to simultaneously measure the actual current waveform flowing through the battery pack and the voltage response waveform across the battery pack. In a specific implementation, the precision sampling path includes a current sampling branch and a voltage sampling branch. In the current sampling branch, the first and second terminals of the precision current sampling resistor are electrically connected to the non-inverting and inverting input terminals of a first high-precision differential amplifier, respectively. The single-ended output pin of the first differential amplifier is electrically connected to a first analog-to-digital converter channel of the main control unit 200 dedicated to current sampling. In the voltage sampling branch, the positive and negative terminals of the battery pack under test are electrically connected to the non-inverting and inverting input terminals of a second high-precision differential amplifier, respectively. The single-ended output pin of the second differential amplifier is electrically connected to a second analog-to-digital converter channel of the main control unit 200 dedicated to voltage sampling.

[0126] For example, the main control unit 200 can sequentially generate control signals with frequencies of 1Hz, 10Hz, 100Hz, and 1kHz, while the corresponding voltage-controlled constant current source generates a sinusoidal current excitation with an amplitude of 50mA (RMS). During this period, the two ADC channels synchronously acquire data at a sampling rate of 10kSPS. After acquisition, a data packet containing the original time-domain waveform data of the four frequency bands is sent to the cloud center for subsequent impedance fingerprint calculation.

[0127] S730: Update the digital twin model, conduct SoH assessment and lifetime prediction.

[0128] After receiving the raw time-domain data, the cloud center will invoke an Impedance-Sequence Fusion Network (ISFN) model specifically constructed for this application to assess battery health status and predict remaining lifespan. The ISFN model is a two-stream deep learning network that uses parallel processing of an impedance feature extraction stream and a sequence feature extraction stream, combined with a subsequent fusion and prediction head, to achieve a comprehensive analysis of the battery's current microstate and long-term aging trend.

[0129] Reference Figure 6 The figure illustrates the macroscopic architecture of the ISFN model. The model's input consists of two independent tensors: one is the current impedance fingerprint tensor calculated from the original time-domain data using a Fast Fourier Transform. The other is the historical state sequence tensor extracted from the digital twin model. .

[0130] The impedance feature extraction stream consists of a series of one-dimensional convolutional neural network blocks, which are responsible for extracting an impedance embedding vector characterizing the current internal electrochemical state of the battery from the current impedance fingerprint tensor. The core of the sequence feature extraction stream is a gated recurrent unit network, which is responsible for capturing the long-term aging pattern of the battery from the historical state sequence tensor and generating a sequence embedding vector. .

[0131] Subsequently, a fusion and prediction head concatenates the impedance embedding vector with the sequence embedding vector to form a fused feature vector. The fused feature vector is then fed into a multilayer perceptron, whose output layer contains two neurons that output an assessment value representing the current health status (SoH) and a prediction value representing the remaining useful life (RUL).

[0132] The ISFN model described in this application is trained in a supervised manner on a dataset constructed through accelerated aging experiments on a large number of identical batteries throughout their entire lifecycle. During the dataset construction phase, high-precision offline capacity calibration and EIS measurements were performed at each aging node to obtain the SoH and RUL ground truth labels for model training. The model training process employs a weighted mean squared error loss function and the Adam optimizer, iteratively optimizing the model parameters through mini-batch stochastic gradient descent until the model converges.

[0133] S730: Generates and executes personalized adaptive maintenance strategies.

[0134] Based on the SoH (Sort of Health) assessment value and RUL (Round Usage Limit) prediction value output by the ISFN model in the S730, the cloud center generates a personalized maintenance strategy instruction for the current battery. This instruction replaces the uniform maintenance target, enabling differentiated management of individual batteries.

[0135] For example, For a healthy battery with an evaluation result of SoH=98.2% and RUL=850 cycles, the maintenance instruction generated by the system may be: "Execute long-term storage mode, target SOC range [68%, 72%]".

[0136] For a sub-healthy battery with an evaluation result of SoH=86.5% and RUL=120 cycles, the maintenance instruction generated by the system may be: "Execute conservative storage mode, target SOC range [58%, 62%], and increase maintenance detection frequency."

[0137] For a battery with an evaluation result of SoH=79.1% (below the scrap threshold of 80%), the maintenance instruction generated by the system may be: "Maintenance prohibited, marked as 'pending scrapping', and an alert sent to the system administrator."

[0138] The cloud center sends the generated instructions to the main control board, which then coordinates with the corresponding maintenance channel to execute them precisely. Through the cyclical execution from S710 to S740, this method constructs a dynamically updated digital twin for each battery, transforming maintenance from reactive intervention to data-driven, proactive health management.

[0139] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit described above can be implemented in hardware.

[0140] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any changes or substitutions within the technical scope disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A method for centralized maintenance management of a cascade of multiple strings of lithium batteries, characterized in that, The application comprises: Upon power-up, automatically detecting the electrical sequence of multiple maintenance management boards in a cascaded link, and assigning a unique board position identity to each of the maintenance management boards; Based on the board position identity, determining the first maintenance management board in the cascaded link as a master board, and the rest of the maintenance management boards as slave boards; Obtaining a maintenance mode instruction; Detecting whether a battery pack is connected to any maintenance channel on any maintenance management board; When detecting that a battery pack is connected, automatically controlling the charging circuit or discharging circuit of the channel where the battery pack is located according to the maintenance mode instruction, so as to adjust the power of the battery pack to a preset target power range.

2. The method of claim 1, wherein, The automatic detection of the electrical sequence of multiple maintenance management boards in a cascaded link comprises: Building a series voltage division network through the cascaded link; Each of the maintenance management boards collects the node voltage value in the series voltage division network through the analog-to-digital conversion unit in its master unit; Based on the node voltage value, the electrical sequence of the maintenance management board is determined.

3. The method of claim 1, wherein, The master board and the slave boards communicate through a controller area network bus; The master board broadcasts the maintenance mode instruction to the slave boards through the controller area network bus, and polls the state of each maintenance channel on the slave boards; The slave boards execute the maintenance mode instruction, and report the state information of their maintenance channels to the master board through the controller area network bus.

4. The method of claim 1, wherein, The maintenance mode instruction includes a long-term storage mode instruction or a battery activation mode instruction; In response to the long-term storage mode instruction, the preset target power range is set to 65% to 75% of the rated capacity of the battery; In response to the battery activation mode instruction, first activate the discharging circuit to discharge the power of the battery pack to a discharge cutoff threshold, and then activate the charging circuit to supplement the power of the battery pack to the preset target power range.

5. The method of claim 1, wherein, The automatic control of the charging circuit or discharging circuit of the channel where the battery pack is located comprises: Periodically collect the real-time voltage value of the battery pack; Compare the real-time voltage value with a target voltage interval corresponding to the preset target power range; When the real-time voltage value is lower than the lower limit of the target voltage interval, close the charging circuit; When the real-time voltage value is higher than the upper limit of the target voltage interval, close the discharging circuit.

6. The method of claim 1, wherein, Further comprising: Periodically collect the temperature value of each maintenance channel to build a system-level real-time thermal map; When there are multiple maintenance tasks to be executed, based on the system-level real-time thermal map, preferentially assign maintenance tasks with higher heat generation to maintenance channels with lower temperature; And When the temperature value of any maintenance channel exceeds a preset temperature threshold, cooperatively control the cooling fans on the maintenance management board where the channel is located and its adjacent maintenance management boards, and / or dynamically reduce the maintenance power of the channel.

7. The method of claim 1, wherein, Further comprising: Obtain a unique identity of the battery pack, and create or call its corresponding digital twin model based on the unique identity; Before or after the control of the charging or discharging loop, the online health state data of the battery pack is acquired by injecting a preset electrical excitation signal into the battery pack and collecting its voltage response; Based on the online health state data, the digital twin model is updated and the health state of the battery pack is evaluated and its remaining useful life is predicted; And Based on the results of the health state evaluation and the remaining useful life prediction, a personalized maintenance strategy is generated, and the preset target power range is adjusted according to the personalized maintenance strategy.

8. A device for centralized maintenance management of a cascade of multiple strings of lithium batteries, characterized in that, Comprise: A plurality of cascadable maintenance management single boards, each of which comprises: A power management unit for receiving external power or power from the previous maintenance management single board and supplying power to the circuits of the single board; A main control unit; At least one maintenance channel, the maintenance channel comprising a battery interface, a charging loop controlled by the main control unit, and a discharging loop controlled by the main control unit; The main control unit is configured to perform the method of any one of claims 1 to 7.

9. The apparatus of claim 8, wherein, The plurality of maintenance management single boards are connected through a cascading interface, and an identity voltage sampling pin for determining the electrical sequence is arranged on the cascading interface; Each of the maintenance management single boards is provided with a voltage dividing resistor, and the voltage dividing resistors of all the maintenance management single boards form a series voltage dividing network after cascading; The analog-to-digital conversion input end of the main control unit is connected with the identity voltage sampling pin, for collecting node voltage to determine the board position identity.

10. The apparatus of claim 8, wherein, Also comprise: An identity recognition unit for acquiring the unique identity of the battery pack; and Each of the maintenance channels further comprises a micro-current excitation and precise sampling circuit for injecting an electrical excitation signal into the battery pack and collecting its response under the control of the main control unit; The main control unit is further configured to perform the steps of creating or calling a digital twin model, acquiring online health state data, performing health state evaluation and life prediction, and generating and executing a personalized maintenance strategy with a cloud computing platform as claimed in claim 7.

Images