A multi-cluster battery dynamic grid-connected equalization method and system for a domain management architecture

Abstract

The application discloses a kind of multi-cluster battery dynamic grid-connected equalization methods and systems for domain management architecture, belong to multi-cluster battery grid-connected equalization control field, including domain control unit, cluster control unit, acquisition safety unit and man-machine interaction extension unit.The application realizes the global collaborative equalization of multi-cluster battery by domain management architecture, combines AI prediction equalization and dynamic adjustment strategy, judges the development trend of difference between clusters in advance, reduces the SOC difference between clusters, prolongs the cruising range of ship, and relies on the sampling accuracy of single battery voltage of BCU unit and passive equalization function, ensures the accuracy of intra-cluster equalization, prolongs the overall service life of battery pack, to improve the equalization accuracy and cruising ability of multi-cluster battery, through multi-physical field coupling optimization, fusion temperature, vibration, electromagnetic interference and other multi-dimensional environmental data, make equalization strategy dynamically adapt to complex working conditions of ship, improve the average trouble-free time of ship in harsh sailing environment, ensure long-term stable operation.

Description

Technical Field

[0001] This invention relates to the field of grid-connected equalization control technology for multi-cluster batteries, and in particular to a dynamic grid-connected equalization method and system for multi-cluster batteries oriented towards a domain management architecture. Background Technology

[0002] With the rapid development of pure battery-powered and hybrid-powered ships, the power battery pack, as the core power source, directly affects the navigation reliability of ships due to its operational stability, range, and safety. Ships face complex and unique navigation environments, including harsh conditions such as salt spray, vibration, and severe temperature and humidity fluctuations. Furthermore, they often adopt a parallel power supply mode with multiple battery clusters to meet high power demands. This places stringent requirements on the balanced control and environmental adaptability of the battery management system (BMS).

[0003] Existing multi-cluster battery management technologies for ships lack a domain-level collaborative control architecture, making it impossible to achieve global dynamic balancing and scheduling of multiple battery clusters. This leads to the accumulation of differences in remaining charge (SOC) and state of health (SOH) among the individual battery clusters, which not only reduces the overall range of the battery pack but also accelerates the aging of individual battery clusters. Furthermore, traditional BMS balancing strategies are static and do not dynamically adjust according to the ship's operating conditions (such as light load, heavy load, and charging while docked). They rely solely on real-time status data for passive balancing, lacking the ability to predict the development trend of differences between clusters, resulting in a delayed response. At the same time, traditional BMS does not consider the multi-physics field coupling effects of temperature, vibration, and electromagnetic fields in the ship's environment. The balancing strategy is susceptible to interference from complex environments, and is prone to decreased balancing accuracy or equipment failure in environments with superimposed salt spray, vibration, and electromagnetic interference. Summary of the Invention

[0004] This invention addresses the problems of existing shipboard multi-cluster battery management technologies lacking a domain-level collaborative control architecture and being unable to adapt to multi-physics coupling environments. It proposes a dynamic grid-connected balancing method and system for multi-cluster batteries oriented towards a domain management architecture.

[0005] To achieve the above objectives, the present invention adopts the following technical solution: A multi-cluster battery dynamic grid-connected equalization system oriented towards a domain management architecture includes a domain control unit, a cluster control unit, a data acquisition and security unit, and a human-machine interaction extension unit; The domain control unit includes an AI prediction processing module and a multiphysics coupling calculation module, which serve as the control core of the system. The AI ​​prediction processing module adopts an embedded LSTM neural network model and is responsible for predicting the inter-cluster difference trend based on historical data. The multiphysics coupling calculation module is responsible for fusing temperature, vibration, and electromagnetic data for coupled analysis. The cluster control unit has an isolated CAN communication interface and a vibration sensor signal acquisition interface. The cluster control unit is responsible for executing domain control unit commands, monitoring battery status within the cluster, and equalization control. The safety acquisition unit includes a data acquisition module, an independent temperature monitoring module, a safety detection module, and a multi-physics field auxiliary acquisition module. The data acquisition module has a passive equalization function and uploads data to the cluster control unit through isolated CAN communication. The independent temperature monitoring module provides accurate temperature data for the equalization strategy. The safety detection module is linked with the cluster control unit through CAN communication. The multi-physics field auxiliary acquisition module includes an integrated vibration sensor and an electromagnetic interference sensor, which respectively collect battery compartment vibration data and electromagnetic interference data and upload them synchronously to the cluster control unit. The human-machine interaction extension unit includes a human-machine interaction module and an extension module. The human-machine interaction module is equipped with an interactive display, and the extension module consists of an emergency stop input and an emergency stop output for responding to ship emergency control.

[0006] As a preferred embodiment, the domain control unit is equipped with 4 isolated CAN communication channels and 4 isolated 485 communication channels, supporting real-time interaction of multi-cluster battery status data and multi-physical field data. The domain control unit also has 4 DI inputs and 4 DO outputs, which can realize emergency stop signal detection and status indication.

[0007] As a preferred embodiment, each battery cluster is equipped with a cluster controller, which has 3-channel isolated CAN communication to establish communication with the domain control unit and the data acquisition module. The vibration sensor signal acquisition interface is connected to the vibration data of the ship's battery compartment.

[0008] As a preferred embodiment, the data acquisition module supports voltage acquisition of 4 to 36 battery cells and temperature monitoring of 1 to 36 channels, the independent temperature monitoring support module supports temperature monitoring of 1 to 48 channels, and the safety detection module integrates 2 liquid sensors, 1 smoke sensor, 1 combustible gas sensor and electromagnetic interference intensity detection module.

[0009] As a preferred embodiment, the interactive display supports Chinese and English bilingual switching, balanced status display, visualization of AI prediction results, parameter configuration, historical data export, and alarm prompts.

[0010] A method for dynamic grid-connected balancing of multi-cluster batteries oriented towards a domain management architecture includes the following steps: S1: Acquires status data and multi-physics data. The data acquisition module of each cluster battery collects individual cell voltage and cluster temperature data in real time. The independent temperature monitoring module collects independent temperature data. The safety detection module collects data on smoke, combustible gas, liquid leakage status and electromagnetic interference intensity. The multi-physics auxiliary acquisition module collects vibration data. Then, the cluster controller summarizes the above data and calculates the cluster-level SOC, SOH and insulation resistance. When the total battery voltage is ≥1000V, the relative error is ≤20%. The data is uploaded to the domain control unit via CAN communication. The domain control unit synchronously collects ship navigation condition signals such as propulsion power and navigation status, and calls historical operating data, namely the operating conditions, battery status and environmental data of the past 3 months. S2: AI prediction and multiphysics coupling analysis. The AI ​​prediction module inputs real-time data to predict the development trend of differences in each cluster of cells. When the predicted difference exceeds the threshold, the equalization preparation is triggered in advance. The multiphysics coupling calculation module establishes a coupling model to analyze the mutual influence between multiple physical fields and outputs the coupling correction coefficient. S3: Balance Trigger Judgment. The domain control unit combines AI prediction results and multi-physics field coupling correction coefficients to determine whether to trigger balance. When the real-time inter-cluster SOC difference is ≥5%, SOH difference is ≥10%, temperature difference is ≥5℃, or the AI ​​predicts a future difference of ≥4%, or when the ship's operating conditions change, such as from light load to heavy load or when it stops for charging, the dynamic balance strategy is activated. If the safety detection module detects a safety anomaly, such as excessive combustible gas concentration or seawater leakage, the corresponding cluster is disconnected from the grid first, and the balance operation is suspended. S4: Dynamic balancing strategy execution. The domain control unit executes different dynamic balancing strategies according to the specific sailing mode of the ship. S5: Equalization State and AI Model Iteration. During the equalization process, the data acquisition module feeds back the individual voltage changes to the cluster controller every 100ms. The cluster controller calculates the equalization effect in real time and uploads it to the domain control unit. The domain control unit dynamically adjusts the equalization current and grid connection status based on the feedback data. When the SOC difference between clusters is ≤2.5% and the temperature difference is ≤2℃, the equalization operation stops. At the same time, the equalization data, multi-physics field data and operating condition data are stored in the database for online iterative training of the AI ​​prediction model to improve prediction accuracy.

[0011] Step S2 includes the following steps: S21: AI Predictive Equilibrium. The LSTM model of the AI ​​prediction module is input with real-time collected data on inter-cluster SOC, SOH, and temperature differences. Combined with historical operating data and multi-physics field data, such as vibration and electromagnetic interference, it predicts the development trend of differences in each cluster of cells in the next 3 to 5 minutes. When the predicted difference is ≥4%, the equilibrium preparation is triggered in advance. S22: Multiphysics coupling analysis. The multiphysics coupling calculation module establishes a temperature-vibration-electromagnetic coupling model and analyzes the mutual influence of the three. For example, increased vibration leads to a decrease in heat dissipation efficiency and electromagnetic interference affects communication stability. It outputs coupling correction coefficients. For example, when the vibration acceleration is ≥3g, the equalization current correction coefficient is 0.8. When the electromagnetic interference intensity is ≥80V / m, the communication verification frequency is increased by 2 times. In step S3, when the safety detection module detects smoke, excessive concentration of combustible gas, and seawater leakage, it combines multiphysics data and sends an alarm signal to the cluster control unit via CAN communication. The cluster control unit immediately controls the DO output to cut off the charging and discharging contactor of the corresponding battery cluster and triggers the ship's emergency stop signal through the expansion unit.

[0012] Compared with the prior art, the beneficial effects of the present invention are as follows: 1. This invention achieves global collaborative balancing of multiple battery clusters through a domain management architecture. By combining AI predictive balancing with dynamic adjustment strategies, it can predict the development trend of inter-cluster differences in advance, reduce the SOC difference between clusters, extend the ship's range, and rely on the individual voltage sampling accuracy and passive balancing function of the BCU unit to ensure the accuracy of intra-cluster balancing, extend the overall lifespan of the battery pack, thereby improving the balancing accuracy and range of multiple battery clusters.

[0013] 2. This invention optimizes the system by coupling multiple physics fields and integrating environmental data from multiple dimensions such as temperature, vibration, and electromagnetic interference. This enables the balancing strategy to dynamically adapt to the complex operating conditions of ships, improves the mean time between failures (MTBF) of ships in harsh navigation environments, and ensures long-term stable operation. This enhances the environmental adaptability of the dynamic grid-connected balancing system for multi-cluster batteries.

[0014] 3. This invention integrates the smoke, combustible gas, and liquid detection functions of the safety detection unit, and links them with the equalization control. When an abnormality is detected, the corresponding battery cluster is immediately disconnected from the grid and an emergency stop output is triggered. At the same time, through the insulation resistance detection function of the cluster control unit, combined with the dynamic adjustment of the safety threshold under multi-physical field coupling, safety hazards are eliminated.

[0015] 4. This invention is trained based on historical data of the ship under all operating conditions. It can predict the trend of SOC and temperature differences between clusters in advance, trigger the equalization operation in advance, reduce the equalization time, and avoid the impact of sudden differences on the ship's power output. Attached Figure Description

[0016] Figure 1 This is a schematic diagram illustrating the workflow of a multi-cluster battery dynamic grid-connected balancing system with a domain management architecture proposed in this invention. Figure 2 This is a schematic diagram of the data acquisition process of the security unit in a multi-cluster battery dynamic grid-connected balancing system with a domain management architecture proposed in this invention. Figure 3 This is a flowchart illustrating a dynamic grid-connected balancing method for multi-cluster batteries with a domain management architecture proposed in this invention. Detailed Implementation

[0017] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0018] It should be understood that, when used in this specification and the appended claims, the terms "comprising" and "including" indicate the presence of the described features, integrals, steps, operations, elements and / or components, but do not exclude the presence or addition of one or more other features, integrals, steps, operations, elements, components and / or collections thereof.

[0019] It should also be understood that the terminology used in this specification is for the purpose of describing particular embodiments only and is not intended to limit the invention. As used in this specification and the appended claims, the singular forms “a,” “an,” and “the” are intended to include the plural forms unless the context clearly indicates otherwise.

[0020] It should also be further understood that the term "and / or" as used in this specification and the appended claims refers to any combination of one or more of the associated listed items and all possible combinations, and includes such combinations.

[0021] Example, refer to Figure 1 and Figure 2 In the figure, BMAU is the domain control unit, BSMU is the cluster control unit, BCU is the data acquisition module, ITMU is the independent temperature monitoring module, and SDU is the safety detection module. A multi-cluster battery dynamic grid-connected equalization system oriented towards domain management architecture includes a domain control unit, a cluster control unit, a data acquisition and safety unit, and a human-machine interaction extension unit. The domain control unit adopts the domain management unit of model ARS-DCU-G1IS, which is installed inside the ship's domain management box. It includes an AI prediction processing module and a multiphysics coupling calculation module. As the control core of the system, the AI ​​prediction processing module adopts an embedded LSTM neural network model and is responsible for predicting the inter-cluster difference trend based on historical data. The multiphysics coupling calculation module is responsible for fusing temperature, vibration and electromagnetic data for coupling analysis. The LSTM neural network model embedded in the AI ​​prediction processing module has the following structure: The input layer receives time series data from the past 10 minutes, with a time step of 1 minute. The features include the maximum difference in SOC between clusters, the maximum difference in SOH, the maximum difference in temperature, ship propulsion power, the root mean square value of vibration acceleration, and the mean value of electromagnetic interference intensity. Then, two LSTM hidden layers are connected in sequence, each containing 64 hidden units and using the tanh activation function. The output layer is a fully connected layer that outputs the predicted value of the maximum difference in SOC between clusters in the next 3 to 5 minutes.

[0022] During model training, offline supervised learning was performed using over 100,000 samples obtained from real-ship testing. The loss function was mean squared error (MSE), the optimizer was Adam, the initial learning rate was 0.001, and the batch size was 128. Training continued until the validation loss converged. After training, the model parameters were quantized and deployed on the embedded processor of the domain control unit. During inference, real-time feature sequences were input, and predicted difference values ​​were output. When the maximum predicted inter-cluster SOC difference was ≥4%, an early equalization preparation flag was generated.

[0023] Furthermore, the domain control unit is equipped with 4 isolated CAN communication channels and 4 isolated 485 communication channels, supporting real-time interaction of multi-cluster battery status data and multi-physical field data. The domain control unit also has 4 DI inputs and 4 DO outputs, which can realize emergency stop signal detection and status indication. The cluster control unit adopts the ARS-Cluster Control Unit-G1IS cluster management unit, which is installed inside the ship's high-voltage box. It has an isolated CAN communication interface and a vibration sensor signal acquisition interface, supports total voltage measurement from 0 to 1500V (accuracy ≤ ±0.5%FS), current detection from -500 to +500A (accuracy ≤ ±0.5%FS), and SOC estimation accuracy ≤ 5%. It also has 3 isolated CAN communication interfaces to ensure efficient communication with the domain control unit and the acquisition unit. A new vibration sensor signal acquisition interface is added to receive vibration data from the ship's battery compartment (acceleration range 0 to 5g). The cluster control unit is responsible for executing commands from the domain control unit, monitoring the battery status within the cluster, and equalization control. The vibration sensor signal acquisition interface is directly electrically connected to the vibration sensor in the multi-physics auxiliary acquisition module to receive the analog or digital vibration signals output by the sensor. After being filtered by the cluster control unit, the signals are uploaded to the domain control unit via the CAN bus. Furthermore, each battery cluster is equipped with a cluster controller, which has 3-channel isolated CAN communication to establish communication with the domain control unit and data acquisition module, as well as a vibration sensor signal acquisition interface to access vibration data from the ship's battery compartment. The data acquisition safety unit includes a data acquisition module, an independent temperature monitoring module, a safety detection module, and a multi-physics field auxiliary acquisition module. The data acquisition module adopts the model ARS-BCU-G1CA2D1, which supports the acquisition of voltages of 4-36 series batteries and 1-36 channels of temperature monitoring (10kNTC). The single cell voltage sampling period is ≤100ms, and it has a 70mA passive equalization function. It uploads data to the cluster control unit through 2-channel isolated CAN communication. The independent temperature monitoring module adopts the model ARS-ITMU-G1D1, which supports 1-48 channels of temperature monitoring and has a temperature detection accuracy of -20 to 65℃≤±1℃, providing accurate temperature data for the balancing strategy. The safety detection module uses model ARS-SDU-G1, which integrates 2 liquid sensors (seawater leakage detection), 1 smoke sensor, and 1 combustible gas sensor. It also adds an electromagnetic interference intensity detection submodule (detection range 0 to 100V / m) and links with the cluster control unit through 1 isolated CAN communication. The multi-physics field-assisted acquisition module includes an integrated vibration sensor (range 0 to 5g, accuracy ±0.01g) and an electromagnetic interference sensor, which respectively collect battery compartment vibration data and electromagnetic interference data and upload them synchronously to the cluster control unit; Furthermore, the data acquisition module supports voltage acquisition of 4 to 36 battery cells and temperature monitoring of 1 to 36 channels, the independent temperature monitoring module supports temperature monitoring of 1 to 48 channels, and the safety detection module integrates 2 liquid sensors, 1 smoke sensor, 1 combustible gas sensor and electromagnetic interference intensity detection module. The human-machine interaction expansion unit is equipped with a BeijerX2pro 10-inch display (model X2pro-10), which supports Chinese and English bilingual switching, balanced status display, AI prediction result visualization, parameter configuration, historical data export and alarm prompts; the expansion unit (EXIU, model ARS-EXIU-G1) provides 4 emergency stop inputs and 1 emergency stop output to meet the emergency control needs of ships. Electromagnetic interference (EMI) intensity data is uniformly collected by the EMI intensity detection submodule within the safety detection module. This submodule uses the data for both safety alarm judgment and uploads it to the domain control unit via CAN communication for use by the multiphysics coupling calculation module. The multiphysics auxiliary acquisition module only retains the vibration sensor and does not repeatedly collect EMI data to simplify the system architecture and ensure data consistency.

[0024] Furthermore, the interactive display supports bilingual (Chinese and English) switching, balanced status display, visualization of AI prediction results, parameter configuration, historical data export, and alarm prompts.

[0025] A method for dynamic grid-connected balancing of multi-cluster batteries oriented towards a domain management architecture includes the following steps: S1: Acquires status data and multi-physics data. The data acquisition module of each cluster battery collects individual cell voltage and cluster temperature data in real time. The independent temperature monitoring module collects independent temperature data. The safety detection module collects data on smoke, combustible gas, liquid leakage status, and electromagnetic interference intensity. The multi-physics auxiliary acquisition module collects vibration data. Then, the cluster controller summarizes the above data and calculates the cluster-level SOC, SOH, and insulation resistance. When the total battery voltage is ≥1000V, the relative error of the cluster-level SOC estimation is ≤20%; when the total battery voltage is <1000V, the relative error is ≤10%, to ensure the estimation accuracy across the entire voltage range. The data is then uploaded to the domain control unit via CAN communication. The domain control unit synchronously collects ship navigation condition signals such as propulsion power and navigation status, and retrieves historical operating data (including operating conditions, battery status, and environmental data) from the past 3 months. S2: AI prediction and multiphysics coupling analysis. The AI ​​prediction module inputs real-time data to predict the development trend of differences in each cluster of cells. When the predicted difference exceeds the threshold, the equalization preparation is triggered in advance. The multiphysics coupling calculation module establishes a coupling model to analyze the mutual influence between multiple physical fields and outputs the coupling correction coefficient. Step S2 includes the following steps: S21: AI Predictive Equilibrium. The LSTM model of the AI ​​prediction module is input with real-time collected data on inter-cluster SOC, SOH, and temperature differences. Combined with historical operating data and multi-physics field data, such as vibration and electromagnetic interference, it predicts the development trend of differences in each cluster of cells in the next 3 to 5 minutes. When the predicted difference is ≥4%, the equilibrium preparation is triggered in advance. S22: Multiphysics coupling analysis. The multiphysics coupling calculation module establishes a temperature-vibration-electromagnetic coupling model and analyzes the mutual influence of the three. For example, increased vibration leads to a decrease in heat dissipation efficiency and electromagnetic interference affects communication stability. It outputs coupling correction coefficients. For example, when the vibration acceleration is ≥3g, the equalization current correction coefficient is 0.8. When the electromagnetic interference intensity is ≥80V / m, the communication verification frequency is increased by 2 times. Based on the above, the training process of AI prediction equilibrium is based on real ship test data, covering 10 typical working conditions such as light load, heavy load, and berthing, and more than 100,000 samples for offline training. After training, the embedded chip deployed in the domain control unit supports online incremental training. In multiphysics coupling optimization, the core correction formula is:

[0026] in, To balance the current and limit it to the range of 35mA~70mA, The rated balancing current of the data acquisition unit is 70mA in this invention. The comprehensive equalization current correction coefficient, with a value ranging from 0.5 to 1.0, is obtained by coupling vibration, temperature, and electromagnetic single-field coefficients. For the vibration field equilibrium current correction coefficient, in At version 1.0, The value was 0.9. The value was 0.8, provided by the multiphysics-assisted acquisition module for vibration environment of the ship's battery compartment; This is the temperature field equalization current correction factor, which is 1.0 when T < 60℃. The value was 0.9. The value is 0.7, collected by the acquisition unit; This is the correction factor for the electromagnetic field equilibrium current. The value is 1.0; The value was 0.9. The value was 0.8.

[0027] S3: Balance Trigger Judgment. The domain control unit combines AI prediction results and multi-physics field coupling correction coefficients to determine whether to trigger balance. When the real-time inter-cluster SOC difference is ≥5%, SOH difference is ≥10%, temperature difference is ≥5℃, or the AI ​​predicts a future difference of ≥4%, or when the ship's operating conditions change, such as from light load to heavy load or when it stops for charging, the dynamic balance strategy is activated. If the safety detection module detects a safety anomaly, such as excessive combustible gas concentration or seawater leakage, the corresponding cluster is disconnected from the grid first, and the balance operation is suspended. In step S3, when the safety detection module detects smoke, excessive concentration of combustible gas, and seawater leakage, it combines multiphysics data and sends an alarm signal to the cluster control unit via CAN communication. The cluster control unit immediately controls the DO output to cut off the charging and discharging contactor of the corresponding battery cluster and triggers the ship's emergency stop signal through the expansion unit. Based on the above, the dynamic correction formula for the communication verification frequency is:

[0028] in, The corrected CAN communication verification frequency, in this invention, is 100 to 300 Hz, decoupled from the equalization current correction. It is the basic CAN communication verification frequency, a standard communication parameter for ship BMS, and is 100Hz in this invention; and As the weighting coefficient for the influence of electromagnetic / vibration, in this invention, , ; This is the normalized value of electromagnetic interference intensity, ranging from 0 to 1. The maximum electromagnetic interference in the ship's battery compartment is set to a value of [value missing]. Electromagnetic interference intensity The electromagnetic interference detection submodule in the safety detection module collects data, which is installed between the battery compartment and the high-voltage box. The collected data is directly uploaded to the domain control unit via CAN communication. This is the normalized value of vibration acceleration, ranging from 0 to 1. The maximum vibration acceleration of the ship during navigation is taken as 5g. Vibration data is collected by a multi-physics-assisted acquisition module with a sensor (range 0 to 5g, accuracy ±0.01g) installed inside the ship's battery compartment. The collected data is filtered by the cluster control unit to remove instantaneous vibration interference from the ship's navigation and then uploaded to the domain control unit via CAN communication.

[0029] S4: Dynamic balancing strategy execution. The domain control unit executes different dynamic balancing strategies according to the specific operating mode of the ship. In this invention, the ship's electrical network adopts a DC bus architecture. Each battery cluster is connected to the DC bus via a controllable DC contactor and a bidirectional DC / DC converter. The domain control unit sends grid-connected / off-grid commands to each cluster control unit via the CAN bus, controlling the on / off state of the corresponding contactor and the operating mode (charging or discharging) of the DC / DC converter. During light-load navigation, if the ship's electrical network has surplus generating power, the domain control unit selects one or more battery clusters with the lowest SOC, closes their contactors, and sets the DC / DC converter to constant current charging mode. The charging current is dynamically adjusted according to the multiphysics correction coefficient, and the individual cell balance within the cluster is achieved in combination with the passive balancing of the BCU; high SOC battery clusters remain off-grid or in standby mode. During berthing charging, all battery clusters are connected to the shore power rectifier for charging. The domain control unit coordinates the charging power distribution of each cluster, prioritizing equalization charging of battery clusters with good insulation and minimal environmental disturbance. When the ship is sailing lightly loaded, the domain control unit commands the battery clusters with lower SOC to prioritize connecting to the grid for charging and equalization. Combined with the multi-physics coupling correction coefficient, if the vibration acceleration is ≤2g and the electromagnetic interference intensity is ≤50V / m, the cluster control unit controls the BCU to perform passive equalization with a rated current of 70mA; if the vibration acceleration is >2g, the equalization current is reduced to 50mA to avoid loose wiring. When the ship is under heavy load, the domain control unit shuts down the grid connection of battery clusters with a temperature ≥65℃, instructs the independent temperature monitoring module to start cooling control, and adjusts the cooling wind speed based on the coupled model, that is, appropriately reducing the wind speed when the vibration is large to reduce structural resonance; after the temperature ≤55℃, the grid connection is restored, and combined with the AI ​​prediction results, the equalization power is allocated in advance to clusters with large future differences. When the ship is docked for charging, based on the insulation resistance test results and multi-physics field data, priority is given to equalizing battery clusters that are undamaged, have insulation resistance ≥1MΩ, vibration ≤1g, and electromagnetic interference ≤30V / m. The domain control unit coordinates the controllers of each cluster to perform equalization operations synchronously, and the AI ​​model corrects the equalization time in real time to ensure that the battery clusters are fully charged in a consistent manner. S5: Equalization State and AI Model Iteration. During the equalization process, the data acquisition module feeds back the individual voltage changes to the cluster controller every 100ms. The cluster controller calculates the equalization effect in real time and uploads it to the domain control unit. The domain control unit dynamically adjusts the equalization current and grid connection status based on the feedback data. When the SOC difference between clusters is ≤2.5% and the temperature difference is ≤2℃, the equalization operation is stopped. At the same time, the equalization data, multi-physics field data and operating condition data are stored in the database for online iterative training of the AI ​​prediction model. The specific method is as follows: The domain control unit continuously records complete data for each balancing event. When the number of new records reaches 500 or 72 hours have passed since the last model update, online incremental training is automatically initiated when the ship is docked and the CPU utilization of the domain control unit is below 30%. During training, the parameters of the currently deployed LSTM model are used as initial values, and 10 rounds of fine-tuning are performed using a small batch of samples mixed with the new data and the original training set. The learning rate is reduced to 0.0001, and the training is terminated if the early stopping strategy fails to reduce the loss. If the updated model is evaluated and the prediction error decreases by more than 5% compared to the old model, the original model is replaced; otherwise, the original model is retained and logged, thus ensuring that the prediction capability continuously improves over time.

[0030] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A multi-cluster battery dynamic grid-connected balancing system oriented towards a domain management architecture, characterized in that, It includes a domain control unit, a cluster control unit, a data acquisition security unit, and a human-machine interaction extension unit; The domain control unit includes an AI prediction processing module and a multiphysics coupling calculation module, serving as the control core of the system. The AI ​​prediction processing module employs an embedded LSTM neural network model to predict inter-cluster difference trends based on historical data. The multiphysics coupling calculation module integrates temperature, vibration, and electromagnetic data for coupling analysis and outputs coupling correction coefficients. The domain control unit is also used to generate dynamic equilibrium commands based on the inter-cluster difference trends and the coupling correction coefficients. The cluster control unit has an isolated CAN communication interface and a vibration sensor signal acquisition interface, which are used to execute the dynamic equalization command, monitor the battery status in the cluster, and perform equalization control. The data acquisition safety unit includes a data acquisition module, an independent temperature monitoring module, a safety detection module, and a multi-physics field auxiliary acquisition module. The data acquisition module has a passive equalization function and uploads data to the cluster control unit through isolated CAN communication. The independent temperature monitoring module provides temperature data for the equalization strategy. The safety detection module integrates an electromagnetic interference intensity detection submodule and is linked with the cluster control unit through CAN communication. The data acquisition module supports battery voltage acquisition and temperature monitoring. The independent temperature monitoring module is used for temperature monitoring. The safety detection module integrates a liquid sensor, a smoke sensor, a combustible gas sensor, and an electromagnetic interference intensity detection submodule. The electromagnetic interference intensity detection submodule provides electromagnetic interference data, which is also used by the multiphysics coupling calculation module. The multiphysics-assisted acquisition module includes a vibration sensor for acquiring battery compartment vibration data and uploading it to the cluster control unit; The human-machine interaction extension unit includes a human-machine interaction module and an extension module. The human-machine interaction module is equipped with an interactive display, and the extension module consists of an emergency stop input and an emergency stop output for responding to ship emergency control.

2. The multi-cluster battery dynamic grid-connected balancing system with a domain management architecture as described in claim 1, characterized in that, The domain control unit is equipped with 4 isolated CAN communication channels and 4 isolated 485 communication channels, supporting real-time interaction of multi-cluster battery status data and multi-physical field data. The domain control unit also has 4 DI inputs and 4 DO outputs to realize emergency stop signal detection and status indication.

3. The multi-cluster battery dynamic grid-connected balancing system with a domain management architecture as described in claim 1, characterized in that, Each battery cluster is equipped with a cluster controller, which has 3-channel isolated CAN communication to establish communication with the domain control unit and the data acquisition module. The vibration sensor signal acquisition interface is connected to the vibration data of the ship's battery compartment.

4. The multi-cluster battery dynamic grid-connected balancing system with a domain management architecture as described in claim 1, characterized in that, The interactive display supports switching between Chinese and English, balanced status display, visualization of AI prediction results, parameter configuration, historical data export, and alarm prompts.

5. A method for dynamic grid-connected balancing of multi-cluster batteries oriented to a domain management architecture, applied to a dynamic grid-connected balancing system for multi-cluster batteries oriented to a domain management architecture as described in any one of claims 1-4, characterized in that, Includes the following steps: S1: Acquire status data and multi-physics field data. The data acquisition module of each cluster cell collects individual cell voltage and cluster temperature data in real time. The independent temperature monitoring module collects independent temperature data. The safety detection module collects data on smoke, combustible gas, liquid leakage status and electromagnetic interference intensity. The multi-physics field auxiliary acquisition module collects vibration data. Then, the cluster controller summarizes the above data and calculates the cluster-level SOC, SOH and insulation resistance. The data is then uploaded to the domain control unit via CAN communication. The domain control unit synchronously collects ship navigation condition signals and retrieves historical operating data. S2: AI prediction and multiphysics coupling analysis. The AI ​​prediction module inputs real-time data to predict the development trend of differences in each cluster of cells. When the predicted difference exceeds the threshold, the equalization preparation is triggered in advance. The multiphysics coupling calculation module establishes a coupling model to analyze the mutual influence between multiple physical fields and outputs the coupling correction coefficient. Step S2 includes the following steps: S21: AI Predictive Equilibrium. The LSTM model of the AI ​​prediction module is input with real-time collected data on inter-cluster SOC, SOH, and temperature differences. Combined with historical operating data and multi-physics data, it predicts the future development trend of differences among cells in each cluster. When the predicted difference exceeds the threshold, it triggers the equilibrium preparation in advance. S22: Multiphysics coupling analysis. The multiphysics coupling calculation module establishes a temperature-vibration-electromagnetic coupling model, analyzes the mutual influence of the three, and outputs coupling correction coefficients. S3: Balance trigger judgment. The domain control unit combines the AI ​​prediction results and the multi-physics coupling correction coefficient to determine whether to trigger balance. When the real-time inter-cluster difference exceeds the threshold, or when the ship's operating conditions change, the dynamic balance strategy is activated. If the security detection module detects a security anomaly, it will prioritize disconnecting the corresponding cluster from the grid and suspend the balancing operation. S4: Dynamic balancing strategy execution. The domain control unit executes different dynamic balancing strategies based on the ship's light-load navigation, heavy-load navigation, or berthing charging mode: During light-load navigation, the domain control unit commands battery clusters with lower State of Charge (SOC) to prioritize charging equalization and adjusts the equalization current based on the multi-physics coupling correction coefficient. During heavy-load navigation, the domain control unit shuts down the grid connection of battery clusters whose temperature exceeds a preset threshold, initiates cooling control, and adjusts the cooling parameters based on the coupling model; During charging while parked, the domain control unit selects battery clusters that meet safety conditions for equalization based on insulation resistance detection results and multi-physics data, and coordinates the controllers of each cluster to perform equalization operations synchronously. S5: Equalization State and AI Model Iteration. During the equalization process, the data acquisition module feeds back the individual voltage changes to the cluster controller every 100ms. The cluster controller calculates the equalization effect in real time and uploads it to the domain control unit. The domain control unit dynamically adjusts the equalization current and grid connection status based on the feedback data. When the SOC difference and temperature difference between clusters are less than the threshold, the equalization operation stops. At the same time, the equalization data, multi-physics field data and operating condition data are stored in the database for online iterative training of the AI ​​prediction model to improve prediction accuracy.

6. The method for dynamic grid-connected balancing of multi-cluster batteries oriented towards a domain management architecture according to claim 5, characterized in that, In step S3, when the safety detection module detects smoke, excessive concentration of combustible gas, and seawater leakage, it combines multiphysics data and sends an alarm signal to the cluster control unit via CAN communication. The cluster control unit immediately controls the DO output to cut off the charging and discharging contactor of the corresponding battery cluster and triggers the ship's emergency stop signal through the expansion unit.

Images