A ship battery cluster circulating current suppression and intelligent pre-charge control system and method

By using real-time monitoring and dynamic load allocation, the problem of differentiated management of battery cluster health status in marine energy storage systems has been solved, thereby improving the safety, reliability, and economy of battery clusters, reducing failure rates and maintenance costs, and extending battery cluster lifespan.

CN123074070APending Publication Date: 2026-11-10SICHUAN CAMY NEW ENERGY CO LTD +1
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202611473306.8
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-09-21
Publication Date
2026-11-10

AI Technical Summary

Technical Problem

Existing ship energy storage systems lack real-time monitoring and differentiated management mechanisms for the health status of each battery cluster when operating in parallel. This can lead to battery clusters with poor health being assigned excessive loads, causing safety hazards such as overheating, overcurrent, or accelerated aging.

Method used

By acquiring data from multiple sources, synchronizing a unified clock, using extended Kalman filtering and deep temporal neural networks, the health status of battery clusters is monitored in real time, load is dynamically allocated and circulating current is suppressed, health assessment reports and maintenance strategies are generated, and intelligent decision-making is supported.

Benefits of technology

This has improved the safety, reliability, and economy of battery clusters, reduced failure rates and maintenance costs, extended battery cluster lifespan, and improved the overall operating efficiency of ship energy systems.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123074070A_ABST
    Figure CN123074070A_ABST
Patent Text Reader

Abstract

This invention relates to the field of marine power system technology, and particularly to a marine battery cluster circulating current suppression and intelligent pre-charge control system and method. The system includes: a multi-dimensional state perception module, configured with a multi-channel synchronous sampling unit, for collecting data from each battery cluster, and timestamping and aligning the data of each battery cluster using a unified clock reference, outputting time-series state data, wherein the data of each battery cluster includes voltage, current, and temperature; and a predictive health management module, which, based on the time-series state data, analyzes the time-series state data using a recurrent neural network model to predict the remaining usable life and the probability of failure. In this invention, by monitoring the health status of each battery cluster in real time and dynamically allocating load power accordingly, it is possible to effectively avoid safety risks such as overheating, overcurrent, or accelerated aging caused by excessive load on battery clusters with poor health status.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention relates to the field of marine power system technology, and in particular to a marine battery cluster circulating current suppression and intelligent pre-charge control system and method. Background Technology

[0002] With the increasing demands for ship electrification and energy conservation and emission reduction, energy storage systems are being used more and more widely in ship power systems. Parallel operation of battery clusters can improve system capacity and redundancy to meet the high power and high reliability power supply needs of ships. Ships face complex and ever-changing sea conditions and load fluctuations during navigation, which places higher demands on the safety, stability and energy management capabilities of energy storage systems. At the same time, battery clusters will have different health conditions during long-term use. How to reasonably allocate power and control operation based on the actual condition of each battery cluster, and how to effectively suppress circulating current and ensure system stability during parallel operation, have become key technical problems that urgently need to be solved in the design and operation management of ship energy storage systems.

[0003] Existing ship energy storage systems lack real-time monitoring and differentiated management mechanisms for the health status of each battery cluster when operating in parallel. This can lead to battery clusters with poor health being assigned excessive loads, causing safety hazards such as overheating, overcurrent, or accelerated aging. Summary of the Invention

[0004] To overcome the above deficiencies, this invention provides a ship battery cluster circulating current suppression and intelligent pre-charge control system and method, aiming to improve the problem in the prior art that ship energy storage systems lack a real-time monitoring and differentiated management mechanism for the health status of each battery cluster when the battery clusters are running in parallel.

[0005] The invention employs a multi-source acquisition module to obtain operational data such as voltage, current, and temperature from each battery cluster, and utilizes a unified clock synchronization mechanism for time alignment to address the issue of asynchronous data sampling across multiple clusters. To address data loss during sampling, an interpolation compensation algorithm is used to dynamically recover missing data, ensuring continuous and complete time-series data. An extended Kalman filter is employed to estimate key battery health parameters in real time, dynamically correcting model prediction errors and improving the accuracy and stability of health status assessment. Simultaneously, a deep temporal neural network is used to extract features and recognize patterns from long-term operational data, predicting the remaining usable life and potential failure trends of battery clusters. Finally, by fusing multi-layered analysis results, the system generates a health assessment report and maintenance strategy recommendations, providing real-time monitoring, risk warnings, and intelligent decision support for ship energy management, thereby improving the safety, reliability, and economy of the entire ship's electrical system.

[0006] In a first aspect, the present invention provides the following technical solution: a ship battery cluster circulating current suppression and intelligent pre-charge control system, comprising:

[0007] The multi-dimensional state perception module is equipped with a multi-channel synchronous sampling unit for collecting data from each battery cluster. It timestamps and aligns the data of each battery cluster with time sequence using a unified clock reference, and outputs time sequence state data. The data of each battery cluster includes voltage, current and temperature.

[0008] The predictive health management module, based on the time-series state data, analyzes the time-series state data through a recurrent neural network model to predict the remaining usable life and the probability of failure, and uses an extended Kalman filter algorithm to fuse the time-series state data to estimate the health status, and outputs a dynamic health weight as a comprehensive result.

[0009] The dynamic load allocation module, based on the dynamic health weight and the total power demand from the ship's energy management system, uses a multi-objective optimization algorithm, with the dynamic health weight as the allocation factor, to decouple the total power demand and generate allocation instructions and start / stop instructions containing the target power values ​​of each battery cluster.

[0010] The intelligent safety execution module, based on the allocation instructions and start / stop instructions, and by acquiring the voltage data of the battery clusters from the timing status data, generates control signals for controlling the power units of each battery cluster, provided that preset safety conditions are met.

[0011] By adopting the above technical solutions and employing a unified clock synchronization mechanism, high-precision alignment of operating data such as voltage, current, and temperature of all battery clusters on the timeline is ensured, avoiding data deviations and analysis errors caused by time differences in traditional solutions. Addressing common issues of missing or dropped frames during data acquisition, a dynamic interpolation compensation algorithm can recover missing values, ensuring the integrity and smoothness of time-series data and providing reliable input data for subsequent health assessments and predictive models.

[0012] In terms of health status assessment, the system, based on the extended Kalman filter algorithm, can estimate key health parameters of battery clusters in real time, such as internal resistance and state of health. It also dynamically corrects prediction errors based on actual battery operating data, thereby improving the accuracy and stability of health status assessment. Furthermore, the system incorporates a deep temporal neural network to extract features from long-term operating data of battery clusters, identifying degradation patterns in different clusters and predicting their remaining usable life and the probability of potential failures, thus providing early decision support for system maintenance and optimization.

[0013] By integrating multiple prediction and assessment results, the system can generate detailed health assessment reports, propose intelligent maintenance strategies and energy scheduling schemes, and support predictive maintenance of ship energy systems. This proactive management approach can not only significantly reduce the rate of sudden battery cluster failures and unplanned downtime, but also lower maintenance costs, extend battery cluster lifespan, and improve the overall reliability, safety, and economic efficiency of ship energy systems.

[0014] Ultimately, the deployment of the system enables ships to monitor the health status of battery clusters in real time without human intervention, intelligently predict battery life, and make timely maintenance and replacement decisions, thereby achieving optimized management of the entire life cycle of battery clusters and improving the operating efficiency and overall economy of the ship's power system.

[0015] Secondly, the present invention provides the following technical solution: a method for suppressing circulating current and intelligent pre-charge control of ship battery clusters, the method comprising the following steps:

[0016] S01. Real-time acquisition of the operating data of the ship's multi-battery cluster to obtain time-series status data including voltage, current, and temperature;

[0017] S02. The time-series state data is fused, and the health status of each battery cluster is estimated using the extended Kalman filter algorithm, and dynamic health weights are generated.

[0018] S03. Based on the dynamic health weight and the total power demand from the ship's energy management system, a multi-objective optimization algorithm is used to decouple the total power demand to generate allocation instructions and start / stop instructions that include the target power value and start / stop status of each battery cluster.

[0019] S04. Based on the allocation instruction and start / stop instruction, and obtaining the timing status data, and under the premise of satisfying the preset safety conditions including intelligent pre-charge logic and circulating current suppression logic, generate control signals for controlling the power units of each battery cluster.

[0020] The present invention has the following beneficial effects:

[0021] 1. In this invention, by monitoring the health status of each battery cluster in real time and dynamically allocating the load power accordingly, it is possible to effectively avoid safety risks such as overheating, overcurrent, or accelerated aging caused by battery clusters with poor health bearing excessive loads. At the same time, by reducing the voltage difference between parallel battery clusters through the pre-charge control strategy, circulating current is suppressed, thereby reducing the risk of battery cluster damage and system efficiency reduction caused by circulating current. This fundamentally ensures the stable operation of the ship's power system under complex sea conditions and variable load conditions.

[0022] 2. In this invention, differentiated load allocation is performed based on the real-time health status of each battery cluster, so that battery clusters with good health status bear more power and battery clusters with poor health status bear a lighter load. This can prevent some battery clusters from rapidly degrading due to overuse, thereby extending the overall lifespan of each battery cluster. At the same time, by timely identifying and isolating faulty battery clusters, the spread of faults is prevented, thereby reducing the frequency of battery replacement and maintenance costs.

[0023] 3. In this invention, by assessing the health status and controlling the dynamic power distribution, each battery cluster can operate at its optimal operating point, thereby improving the overall energy conversion efficiency. At the same time, the pre-charge control and circulating current suppression strategy reduce the reactive circulating current loss during parallel operation. Furthermore, the real-time control mechanism based on the health status enables the system to respond quickly to the dynamic changes in the ship's load, ensuring stable power output under various operating conditions. Attached Figure Description

[0024] Figure 1 This invention presents an architecture for a ship battery cluster circulating current suppression and intelligent pre-charge control system. Detailed Implementation

[0025] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0026] Example 1:

[0027] In a first embodiment of the present invention, the present invention provides a ship battery cluster circulating current suppression and intelligent pre-charge control system, such as... Figure 1 As shown, it includes:

[0028] The multi-dimensional state perception module is equipped with a multi-channel synchronous sampling unit to collect data from each battery cluster. It timestamps and aligns the data of each battery cluster with timing using a unified clock reference, and outputs timing state data, including voltage, current and temperature for each battery cluster.

[0029] Furthermore, the steps of collecting data from each battery cluster, timestamping and aligning the data of each battery cluster using a unified clock reference, and outputting timing status data specifically include the following steps:

[0030] A multi-channel synchronous sampling unit is triggered with a preset sampling period to synchronously collect voltage, current and temperature data of each battery cluster.

[0031] The current time value of the unified clock reference is appended as a timestamp to the collected data corresponding to the sampling period to form a timestamped data frame;

[0032] The timestamped data frames are stored in a data buffer queue, and sorted, verified, and interpolated according to the timestamps to achieve time sequence alignment.

[0033] Aligned data is output from the data buffer queue using a preset data structure to form time-series state data.

[0034] Specifically, the multi-dimensional state perception module integrates real-time data acquisition hardware, such as multi-channel synchronous sampling units designed based on high-speed analog-to-digital converters (ADCs) and programmable gate arrays (FPGAs) or high-performance microcontrollers (MCUs). These units are directly connected to each battery cluster in the ship's electric propulsion system through sensor interfaces to monitor and acquire its voltage at both ends, charging and discharging current, and temperature data of key internal points in real time. Through a unified clock reference provided by an internal high-precision crystal oscillator or an external GNSS timing module, all acquired data are timestamped at the microsecond level and precise timing alignment is performed to ultimately form a continuous, complete, and time-synchronized multi-dimensional time-series state data stream.

[0035] First, the system's internal control logic, such as the clock management unit embedded in the FPGA, sends a synchronization trigger signal to the multi-channel synchronous sampling unit at a preset, fixed period, such as every 50 milliseconds. Upon receiving the trigger signal, the sampling unit simultaneously begins acquiring specified operating parameters for all on-grid battery clusters: including charging and discharging current and terminal voltage obtained through high-precision shunts and voltage detection circuits, and temperature data obtained through thermistors or thermocouple sensors. For example, for the first... Each battery cluster, at the sampling time Voltage values ​​will be acquired simultaneously. Current value and temperature value .

[0036] Next, the multi-dimensional state awareness module appends the precise time value of the current unified clock reference as a unique timestamp to all battery cluster data captured during that sampling period, forming a complete data frame. This unified clock reference is typically provided by a system-level hardware clock or a precise time source synchronized via the Network Time Protocol (NTP), ensuring time consistency of all data within the entire ship's electrical system. For example, if the current precise time is... The generated data frame can then be represented as a structure, containing... And all N battery clusters in The data vector consists of voltage, current, and temperature measured at each moment. Specifically, the data frame structure can be:

[0037] ;

[0038] in: Indicates a unified timestamp A complete frame of data is collected. Each frame of data corresponds to one synchronous sampling period.

[0039] A unified timestamp, which is the standard time obtained by the system through a unified clock reference (such as a hardware clock or NTP protocol), is used to identify the sampling time of the frame data.

[0040] This represents the number of individual battery cells in the battery cluster.

[0041] Indicates the first Individual battery cells in time The voltage value at that time.

[0042] Indicates the first Individual battery cells in time The current value at that time.

[0043] Indicates the first Individual battery cells in time The temperature value at that time.

[0044] Indicates a unified timestamp The collection of voltage, current, and temperature data for all individual battery cells.

[0045] Subsequently, these timestamped data frames are continuously fed into a dedicated data buffer queue. To address potential transmission delays, out-of-order data packets, and even data loss caused by the complex ship network environment, this buffer queue performs strict sorting checks and data integrity processing. The system continuously monitors the timestamps of the data frames in the buffer queue. If it detects that the timestamps of adjacent data frames are reversed, it immediately reorders them according to the timestamps. Simultaneously, the system calculates the interval between the timestamps of adjacent data frames and compares it with a preset sampling period. If it detects that a data frame corresponding to a certain timestamp is missing, for example, two consecutive valid data frames... and If the time interval between data points exceeds one sampling period, it indicates data loss. In this case, the system will initiate an interpolation-filling algorithm to recover the data. For example, for lost voltage data... Linear interpolation can be used, based on the known previous valid data point. and the next valid data point The formula can be expressed as follows: ;

[0046] in: This represents the interpolation result for missing data points, i.e., the voltage (or other parameter) value obtained through algorithm compensation.

[0047] This represents the voltage value of the most recent valid sampled data before the missing point.

[0048] This represents the voltage value of the most recent valid sampled data after the missing point.

[0049] This represents the timestamp corresponding to the previous valid data point.

[0050] This represents the timestamp corresponding to the next valid data point.

[0051] This represents the timestamp corresponding to the missing data point.

[0052] For more complex or continuously lost data, advanced methods such as cubic spline interpolation can be used to ensure the smoothness and accuracy of the recovered data. In addition, when data frames enter the queue, CRC checks can be performed to verify the integrity of data transmission, and damaged data frames can be detected and discarded in a timely manner. A series of processes ensure the continuity, integrity and strict time alignment of the output data.

[0053] Finally, the multi-dimensional state awareness module outputs sorted, validated, and interpolated data from the data buffer queue using a pre-defined, standardized data structure, such as a table or array of structures containing timestamps and multi-dimensional feature vectors. This data is called time-series state data. Each output time point corresponds to the voltage, current, and temperature values ​​of all battery clusters at the moment of strict synchronization, forming a complete multi-dimensional vector. For example, the output time-series state data can be represented as a series of: Data records,

[0054] in, Indicates a unified timestamp The system time-series state vector (time-series state data record) is the set of key operating parameters of all battery clusters collected at the same time.

[0055] A unified timestamp, after calibration and alignment, ensures that data from different battery clusters are strictly synchronized in time.

[0056] This indicates the number of battery clusters, i.e., the total number of battery clusters in the system. Each independent battery cluster.

[0057] Indicates the first Individual battery clusters in time The voltage value (unit: V).

[0058] Indicates the first Individual battery clusters in time The current value (unit: A).

[0059] Indicates the first Individual battery clusters in time Temperature value (unit: °C).

[0060] These high-precision, time-synchronized, and complete time-series status data are directly provided as input to the subsequent health status estimation module and dynamic load allocation module, improving the quality of input data and thus ensuring the accuracy of battery cluster health status assessment and the reliability of dynamic load allocation decisions, significantly enhancing the operational safety and efficiency of the ship's battery cluster system.

[0061] The predictive health management module, based on the time-series state data, analyzes the time-series state data through a recurrent neural network model to predict the remaining usable life and the probability of failure, and uses an extended Kalman filter algorithm to fuse the time-series state data to estimate the health status, and outputs a dynamic health weight as a comprehensive result.

[0062] Furthermore, analyzing time-series state data using a recurrent neural network model to predict remaining usable lifetime and failure probability specifically includes the following steps:

[0063] The time-series state data is constructed into a feature sequence with a preset time step, which is then used as the input to the recurrent neural network model.

[0064] By using the recurrent layer units of the recurrent neural network model, temporally dependent features are extracted from the feature sequence, and hidden state vectors encoding historical information are generated.

[0065] The hidden state vector is input into the regression prediction head and the probability output head, respectively;

[0066] The regression prediction head outputs the predicted value of the remaining usable life, and the probability output head outputs the predicted value of the failure probability.

[0067] Furthermore, a preset battery state space model is established, which is used to characterize the nonlinear relationship between the time-series state data and the internal state vector containing the health state.

[0068] Based on the preset battery state space model and the health state estimate of the previous moment, a time update is performed to predict the prior value of the health state at the current moment.

[0069] Using the current measured data in the time series state data, the error between the prior value of the health status at the current moment and the current measured data is calculated, and the measurement is updated in combination with Kalman gain to correct the prior value of the health status at the current moment and obtain the posterior estimate of the health status.

[0070] The posterior estimate of the health status of each battery cluster is transformed using a preset mapping function to generate the dynamic health weight.

[0071] Specifically, the predictive health management module receives high-precision, time-synchronized time-series state data from the multi-dimensional state perception module. Based on this data, it uses a specially trained recurrent neural network model to proactively predict the remaining usable life (RUL) and potential failure probability of each battery cluster. Simultaneously, the module runs an extended Kalman filter algorithm in parallel to perform high-precision real-time estimation of the battery's state of health (SOH). Finally, these predictions and estimations are fused and transformed to generate a dynamic health weight reflecting the current health level, performance degradation, and future operational risks of each battery cluster. This weight serves as a key input for subsequent intelligent load allocation strategies, guiding the system to manage battery resources more effectively.

[0072] First, the time-series state data is constructed into a feature sequence with a preset time step, which is then used as input to a recurrent neural network model for each battery cluster on the ship. The system will extract and construct a dataset containing a fixed number of past data points from its continuous time-series state data stream. A dense and synchronized sequence of voltage, current, and temperature measurements at each time step is used as input features for the model, for example, at the current time step. The feature sequence input into the recurrent neural network It can be represented as a A matrix, where each row represents a time step. Data, i.e. ,in, , , Representing battery clusters At any moment The voltage, current and temperature measurements, Typically set to tens to hundreds of time steps to capture a sufficiently long historical operating pattern, this sequential input method gives recurrent neural networks the ability to capture long-term and complex time-dependent characteristics of batteries during charge-discharge cycles, such as small drifts in the voltage curve, slow increases in internal resistance, and continuous capacity decay.

[0073] Next, through the recurrent layer units of the recurrent neural network model, time-dependent features are extracted from the feature sequence, and hidden state vectors encoding historical information are generated. These recurrent layer units, such as Long Short-Term Memory (LSTM) networks or Gated Recurrent Units (GRUs), are designed to process and remember key information in the time-series data for each time step of the input feature sequence. The loop unit will, based on its internal control mechanism and the hidden state of the previous moment, and current input It finely updates its internal cell states and outputs the hidden state at the current moment. Its core recursive calculation can be abstracted as = This hidden state vector It is a high-dimensional numerical representation that dynamically encodes and condenses all important historical operational information and timing patterns from the beginning of the sequence to the current moment. It is particularly good at identifying nonlinear and slowly changing timing features such as battery aging and performance degradation.

[0074] Then, the hidden state vector of the last time step of the recurrent neural network, or the hidden state vector after weighted aggregation by the attention mechanism, is input into two independent output modules: the regression prediction head and the probability output head. This modular design allows the system to optimize different network structures and activation functions for predicting the remaining lifetime (a continuous value) and predicting the failure probability (a value between 0 and 1), respectively, in order to improve their respective prediction accuracy and robustness.

[0075] Finally, the regression prediction head outputs the predicted remaining usable lifetime, and the probability output head outputs the predicted failure probability. The regression prediction head typically consists of one or more fully connected layers, with its last layer employing a linear activation function to directly output a continuous scalar value, namely the predicted remaining usable lifetime (RUL) of the battery cluster. For example, the prediction head can detect hidden state vectors. Mapped to ;

[0076] in, The predicted remaining usable life (RUL) of the battery cluster.

[0077] is the weight matrix of the RUL prediction layer, used to map hidden state features to the lifetime prediction output space.

[0078] The final hidden state vector of a recurrent neural network (such as LSTM or GRU) represents a comprehensive representation of the battery's historical operating characteristics.

[0079] This is a bias term for the RUL prediction layer, used to adjust the baseline level of the model output.

[0080] The probability output header also consists of fully connected layers, but its last layer typically uses the sigmoid activation function to output a probability value between 0 and 1, which is the predicted probability of the battery cluster failing within a specific time period in the future. ,For example, ;

[0081] in This is the predicted probability of battery cluster failure, and its value ranges from [0,1].

[0082] The sigmoid activation function is defined as follows: This is used to map the model output to probability values.

[0083] This is the weight matrix of the fault prediction layer, used to extract the relationship between hidden states and fault risks.

[0084] This is the final hidden state vector of the network, which contains information about the battery's operating state and degradation features.

[0085] This is the bias term for the fault prediction layer.

[0086] These real-time updated RUL and failure probability predictions provide ship managers with an important forward-looking view of the future health and potential risks of batteries, supporting predictive maintenance and risk-avoidance decisions.

[0087] Furthermore, the extended Kalman filter algorithm is used to fuse time-series state data to estimate the health status, and the comprehensive output is a dynamic health weight. The specific steps include:

[0088] First, a pre-defined battery state-space model is established. This model characterizes the nonlinear relationship between time-series state data and the internal state vector containing the healthy state. This model is the foundation of the Extended Kalman Filter (EKF) algorithm, and it rigorously describes the internal states of the battery that cannot be directly measured (such as the healthy state of state (SOH) and internal ohmic impedance). diffusion impedance The model consists of two core parts, which study how internal states (such as capacity decay factor, etc.) evolve with time and operating conditions, and how these internal states affect observable time-series state data (such as voltage, current, and temperature) through the battery's physical and electrochemical properties.

[0089] State transition equation: ;

[0090] Observation equation: ;

[0091] in, express The internal state vector at time t, its core component is SOH. express Input at any given time, including charging and discharging current and ambient temperature ; express The core component of the observation vector at time t is the battery terminal voltage. , It is a nonlinear function that describes the evolution of the internal state over time, such as the decay model of SOH or the dynamic change model of battery capacity. It is a nonlinear function that describes the relationship between the internal state and the observable terminal voltage, and is usually based on the equivalent circuit model of the battery, such as the Thevenin model or the Randles model. and These are process noise and measurement noise, which follow a Gaussian distribution, used to characterize model uncertainty and sensor error. The functional forms and parameters in these models are identified and calibrated offline using a large amount of battery cycle life experimental data to simulate actual battery behavior.

[0092] Next, based on the state-space model and the health state estimate from the previous time step, a time update is performed to predict the prior health state value at the current time step. The EKF algorithm first uses the posterior state estimate from the previous time step (i.e., the corrected health state estimate). and input during that time period Through the linearized state transition equation To predict the prior state estimate at the current moment. Simultaneously, the algorithm calculates the prior error covariance matrix. This matrix quantifies the uncertainty of the predicted state. The time update process can be represented as:

[0093] State prediction: ;

[0094] in, For a moment The predicted state vector represents the estimated state of the system at the current moment.

[0095] For the previous moment The state estimate.

[0096] This is the state transition function, which describes the evolution of the system state over time.

[0097] To control the input quantity, it represents the impact of external operations or environmental conditions on the system state.

[0098] Covariance prediction: ;

[0099] in, For a moment The prediction covariance matrix represents the uncertainty of the current state estimate.

[0100] Let be the posterior covariance matrix of the previous time step.

[0101] The state transition Jacobian matrix is ​​given by the state transition function. It is formed by the first-order partial derivatives with respect to the state variables.

[0102] This is the process noise covariance matrix, which reflects the influence of model prediction errors and system disturbances.

[0103] in, It is a state transition function At point The first-order Taylor expansion (Jacobi matrix) at a given point represents how the state error is propagated from the previous time step to the current time step. It is the covariance matrix of process noise, reflecting the impact of model inaccuracies or unmodeled dynamics on state estimation. This step uses the battery's own dynamic model to predict the long-term evolution trend of SOH.

[0104] Then, using the current measured data in the time-series state data, the error between the prior value and the measured value is calculated, and the Kalman gain is used to update the measurement to correct the prior value, thereby obtaining the posterior estimate of the health state, after acquiring the actual measurement data at the current time. After (e.g., measured voltage), EKF will compare it with the observed value predicted based on the prior state. (e.g., model predicted voltage) are compared to calculate the measurement residual. This residual directly reflects the difference between the model predictions and actual observations. Subsequently, the algorithm calculates the Kalman gain. The gain is a weighting factor that determines the extent to which current measurement data is trusted to correct previous predictions. The calculation of the Kalman gain takes into account the prior error covariance matrix. and measurement noise covariance matrix (Reflecting sensor accuracy), the measurement update process can be represented as:

[0105] Kalman gain: ;

[0106] in: Kalman Gain is the Kalman gain matrix, representing the weight balance factor between model predictions and measured data during measurement updates.

[0107] Let be the prior covariance matrix, representing the magnitude of the uncertainty in the system state before the measurement update.

[0108] The observation matrix (derived from the observation function) (Obtained by the first-order Taylor expansion at the current estimation point), used to map the state variables to the observation space.

[0109] The noise covariance matrix is ​​measured to reflect the magnitude and accuracy of the sensor's measurement noise.

[0110] The total covariance of the predicted observation error, taking into account both model and measurement noise.

[0111] Status Update: ;

[0112] in, This is the posterior state estimate (the final estimate after measurement update).

[0113] This is the prior state estimate (the state value obtained during the prediction phase).

[0114] This is the Kalman gain matrix.

[0115] : Measurement residuals or innovations represent the difference between actual and predicted measurements.

[0116] Covariance update: ;

[0117] in, Let be the posterior covariance matrix, representing the uncertainty of the system state after the measurement update.

[0118] It is an identity matrix.

[0119] This is the Kalman gain matrix.

[0120] This is the observation matrix.

[0121] Let be the prior covariance matrix.

[0122] in, It is the observation function At point The first-order Taylor expansion (Jacobi matrix) at the point represents how the state error affects the observation error. It is an identity matrix. Through measurement updates, EKF effectively integrates real-time measurement data to correct biases in model predictions, making the estimated SOH value closer to the true health condition of the battery, significantly improving the accuracy of the estimation and robustness to measurement noise.

[0123] Finally, the posterior health state estimates of each battery cluster are transformed using a preset mapping function to generate dynamic health weights, which are then derived from the posterior state estimates output by the EKF algorithm. Extract the current SOH estimate for each battery cluster. These SOH values ​​are input into a carefully designed preset mapping function. In this process, it is converted into a dynamic health weight between 0 and 1.

[0124] For example, a mapping function can be defined as:

[0125] ;

[0126] in This is a preset minimum acceptable health status threshold. Battery clusters with a health status below this threshold may have their weights significantly reduced or even reduced to zero. The closer a battery cluster's health status is to 100%, the closer its weight is to 1. This dynamic health weight is a comprehensive indicator that quantifies the relative health status, performance level, and load-bearing capacity of each battery cluster. It will serve as a key allocation factor in the multi-objective optimization algorithm, guiding the system to prioritize battery clusters with better health status to bear more load. This will maximize the lifespan of the entire battery cluster system and improve its operational economy while ensuring the overall safety of the system.

[0127] This predictive health management module integrates deep learning and Kalman filtering algorithms to accurately monitor the health status and degradation trend of ship battery clusters in real time, thereby improving the reliability and safety of system operation and effectively avoiding the risk of failure due to unclear battery health status.

[0128] The dynamic load allocation module, based on dynamic health weights and total power requirements from the ship's energy management system, employs a multi-objective optimization algorithm, using dynamic health weights as allocation factors to decouple the total power requirements and generate allocation instructions and start / stop instructions containing target power values ​​for each battery cluster.

[0129] Furthermore, based on dynamic health weights and the total power demand from the ship's energy management system, a multi-objective optimization algorithm is adopted, using dynamic health weights as the allocation factor, to decouple the total power demand and generate allocation instructions and start / stop instructions containing the target power values ​​of each battery cluster. Specifically, this includes the following steps:

[0130] Construct a multi-objective optimization model;

[0131] Solving multi-objective optimization problems: Using multi-objective optimization algorithms, solve the multi-objective optimization model to obtain a set of Pareto optimal solutions;

[0132] Select the optimal solution and generate instructions: From the Pareto optimal solution set, determine the final allocation scheme according to the preset decision strategy, and extract the target power value and start / stop status of each battery cluster according to the final allocation scheme, and generate allocation instructions and start / stop instructions respectively.

[0133] Determine the decision variables, including the target power value and start / stop status of each battery cluster;

[0134] Define an objective function that includes the following:

[0135] Objective function 1: Minimize the deviation between the sum of the target power values ​​of each battery cluster and the total power demand from the ship's energy management system;

[0136] Objective function two: Using the dynamic health weight as an allocation factor, differentially adjust the target power value of each battery cluster;

[0137] The definitions include the maximum and minimum power output limits for each battery cluster, the conditions for meeting the total power demand, and the constraints related to the start / stop status and power output logic of each battery cluster.

[0138] Specifically, the dynamic load allocation module is the decision-making center for the ship's battery cluster health status estimation and dynamic load allocation system. It receives real-time dynamic health weights from the predictive health management module and the global total power demand from the ship's energy management system (BMS). Using a multi-objective optimization algorithm, with the dynamic health weights as the key allocation factor, it intelligently decouples the total power demand, thereby calculating the target power value and corresponding start / stop status for each battery cluster. Finally, the module generates allocation and start / stop instructions containing these specific commands, guiding the intelligent safety execution module to precisely control the power units of each battery cluster, ensuring the entire battery system operates under efficient, safe, and long-life conditions.

[0139] Furthermore, based on dynamic health weights and the total power demand from the ship's energy management system, a multi-objective optimization algorithm is employed, using the dynamic health weights as the allocation factor, to decouple the total power demand. The generation of allocation instructions and start / stop instructions containing the target power values ​​for each battery cluster is specifically achieved through the following steps.

[0140] First, a multi-objective optimization model is constructed, which is the mathematical foundation of the entire dynamic load allocation process. It transforms the operating status, health status, and system requirements of the battery cluster into solvable mathematical expressions.

[0141] Next, the multi-objective optimization problem is solved using multi-objective optimization algorithms, such as Non-Dominated Sorting Genetic Algorithm (NSGA-II), Multi-Objective Particle Swarm Optimization (MOPSO), or Decomposition-Based Multi-Objective Evolutionary Algorithm (MOEA / D). The constructed multi-objective optimization model is solved through iterative search, generating a Pareto optimal solution set. Each solution in the Pareto optimal solution set represents a possible battery cluster power allocation scheme and start-stop state combination. The improvement of one objective function will not lead to the deterioration of at least one other objective function. For example, one solution may maximize the protection of battery clusters with poor health while meeting the total power demand, but not all battery clusters may be at their high-efficiency operating point. Another solution may pursue the highest overall efficiency, but may lead to uneven load distribution among some battery clusters. The algorithm continues to iterate until the preset convergence condition or the maximum number of iterations is reached.

[0142] Then, the optimal solution is selected and instructions are generated. From the generated Pareto optimal solution set, the final allocation scheme is determined according to a preset decision-making strategy. The decision-making strategy can be based on expert experience rules, fuzzy logic, or further evaluation of solutions on the Pareto front using decision analysis methods such as entropy weighting and analytic hierarchy process, selecting the best compromise scheme that balances all optimization objectives. For example, in a ship emergency, the scheme that meets the total power requirement and has the fastest response speed may be prioritized; during normal cruising, the scheme that balances efficiency and battery life may be prioritized. Once the final allocation scheme is selected, the module extracts the target power value and start / stop status of each battery cluster from the scheme. These extracted specific values ​​represent the expected power output of each battery cluster. and its corresponding start / stop status (e.g., 1 indicates start-up, 0 indicates stop-up). Based on these values, the module generates allocation instructions containing the target power values ​​of each battery cluster, as well as start-stop instructions containing the start-stop status of each battery cluster, and sends them to the intelligent safety execution module.

[0143] Furthermore, constructing a multi-objective optimization model specifically includes the following steps.

[0144] First, for each battery cluster on the ship ( ,in (Total number of battery clusters), define its target power value. As a continuous decision variable, its value ranges from the minimum available power of the battery cluster. To maximum available power Between. At the same time, define a binary decision variable. To represent battery clusters The start / stop status, among which This indicates that the battery cluster is in operation. This indicates that the battery cluster is in a stopped state, and these decision variables together determine the final load distribution scheme.

[0145] Next, define the objective function, which includes the following:

[0146] Objective function 1: Minimize the deviation between the sum of the target power values ​​of each battery cluster and the total power requirement, ensuring that the total power allocated to all battery clusters accurately matches the total power requirement from the ship's energy management system. Its mathematical expression is:

[0147] ;

[0148] in, This represents the total output power of all battery clusters in operation, achieved by minimizing its... The absolute deviation ensures that the system can accurately respond to the ship's instantaneous power demand.

[0149] Objective Function 2: Using dynamic health weights as allocation factors, differentially adjust the target power values ​​of each battery cluster. This objective aims to closely link load allocation with the health status of the battery clusters, ensuring that battery clusters with better health bear relatively higher loads, while those with poorer health bear relatively lower loads, thereby extending the overall system lifespan. Its mathematical expression can be designed as minimizing the sum of squared power deviations after health weighting:

[0150] ;

[0151] in, It is a battery cluster from the predictive health management module. The dynamic health weight (a value between 0 and 1, with a higher weight for better health). Represents battery clusters The "unit health load" after considering health weights, and This represents the average distribution of the total system load across all operating battery clusters based on their health status. By minimizing this weighted bias, battery clusters with higher health status can be encouraged to bear more actual power, while those with lower health status can bear less actual power, thus achieving health-sensitive differential load distribution.

[0152] Finally, constraints are defined, including the maximum and minimum power output limits of each battery cluster, the conditions for meeting the total power demand, and the logical association between the start / stop state of each battery cluster and its power output. These constraints are the boundaries and rules that must be strictly followed in the optimization problem.

[0153] The power output of each battery cluster must be within its own physical limitations:

[0154] ;

[0155] The total power requirement must be within the range of the total system capacity:

[0156] ;

[0157] There is a logical relationship between start / stop status and power output:

[0158] like (Stop), then ;

[0159] like (Run) ;

[0160] These constraints ensure that the optimization results are feasible and safe at the practical physical and operational levels.

[0161] This dynamic load allocation module integrates advanced multi-objective optimization algorithms, enabling it to respond in real time to changes in the ship's power demand and perform intelligent and precise load allocation based on the dynamic health status of the battery clusters.

[0162] The intelligent safety execution module, based on the allocation instructions and start / stop instructions, and by acquiring the voltage data of the battery clusters from the timing status data, generates control signals for controlling the power units of each battery cluster, provided that preset safety conditions are met.

[0163] Furthermore, based on the allocation instructions and start / stop instructions, and by acquiring voltage data from the timing status data, the generation of control signals for controlling each battery cluster power unit, under the premise of meeting preset safety conditions, specifically includes the following steps:

[0164] Perform intelligent pre-charging:

[0165] Receive start / stop commands;

[0166] When the start / stop command instructs the battery cluster to switch from a non-operating state to an operating state, the real-time voltage of the battery cluster and the system bus voltage are obtained from the timing status data.

[0167] Determine whether the voltage difference between the real-time voltage of the battery cluster and the bus voltage exceeds the preset pre-charge safety threshold;

[0168] If the voltage difference exceeds the preset precharge safety threshold, a precharge control signal is generated to control the power unit of the battery cluster to perform precharge operation until the voltage difference drops below the preset precharge safety threshold.

[0169] If the voltage difference is within the preset pre-charge safety threshold or pre-charge is completed, a start-up control signal is generated to control the power units of the battery cluster to be connected to the bus.

[0170] Perform circulating current suppression and safe output:

[0171] Receive allocation instructions;

[0172] During the parallel operation of multiple battery clusters, real-time current and voltage data of each parallel battery cluster are acquired from the timing status data;

[0173] Analyze real-time current data to determine whether there is an imbalance in the current distribution between each parallel battery cluster that exceeds the preset circulating current suppression threshold, or whether there is a preset overload threshold.

[0174] If an imbalance or overload exists, a circulating current suppression control signal is generated to fine-tune the output parameters of the relevant battery cluster power units in order to eliminate the imbalance or alleviate the overload.

[0175] If no imbalance or overload is detected, a power output control signal is generated to control the power units of each battery cluster to output power according to the allocation instructions.

[0176] Specifically, the intelligent safety execution module is the final execution and safety assurance unit of the ship's battery cluster health status estimation and dynamic load allocation system. It receives allocation instructions and start / stop instructions generated by the dynamic load allocation module, which include the target power value and start / stop status of each battery cluster. It also acquires voltage and current data from the time-series status data provided by the multi-dimensional status perception module in real time. Under the premise of strictly meeting the preset safety conditions, it generates control signals to directly drive and manage the power units of each battery cluster (such as DC / DC converters or bidirectional DC / AC inverters), ensuring the safe start-up, stable operation and efficient power output of the battery cluster system, while actively suppressing potential operational risks.

[0177] Furthermore, based on the allocation instructions and start / stop instructions, and by acquiring voltage data from the timing status data, the generation of control signals for controlling each battery cluster power unit, under the premise of meeting preset safety conditions, specifically includes the following steps:

[0178] First, intelligent pre-charging is performed to avoid high current surges caused by voltage mismatch.

[0179] The system receives start / stop commands from the dynamic load distribution module. When the start / stop command specifies a battery cluster, for example, a battery cluster... When it is necessary to switch from a non-operating state to an operating state and be connected to the system bus, the intelligent safety execution module immediately starts the pre-charge process.

[0180] At this time, the module acquires battery clusters from the timing status data. Real-time voltage and the real-time voltage of the system bus These voltage data are high-precision measurements from the multi-dimensional state sensing module that have been time-synchronized and calibrated.

[0181] Subsequently, the module determines the battery cluster. Real-time voltage With system bus voltage The absolute value of the voltage difference between Has the preset precharge safety threshold been exceeded? This threshold is typically set based on the electrical characteristics of the power unit and battery cluster, such as 5V or 10V, to limit the inrush current that may be generated at the moment of closing the circuit.

[0182] If the absolute value of the voltage difference exceeds the preset pre-charge safety threshold, that is The module then generates a precharge control signal, which is sent to the battery cluster. The power unit is instructed to start its internal pre-charge circuit. The power unit usually charges or discharges the battery cluster slowly through a series current-limiting resistor or a controlled soft-start circuit, so that its voltage gradually approaches the bus voltage. This process continues, and the module continuously monitors the voltage difference until the absolute value of the voltage difference drops below the preset pre-charge safety threshold.

[0183] If battery cluster The voltage difference is within the preset pre-charge safety threshold, that is... If the pre-charging operation has been completed and the voltage difference meets the requirements, the module generates a start-up control signal, which instructs the battery cluster to start operation. The power units are fully turned on, for example, by closing the main contactor or the main power circuit of the DC / DC converter, so that the battery pack is smoothly and without impact connected to the system bus and begins normal operation.

[0184] Next, circulating current suppression and safe output are performed. This process is designed to ensure current balance and safe load during operation of multiple parallel battery clusters.

[0185] The system receives allocation instructions from the dynamic load allocation module. During the parallel operation of multiple battery clusters, the intelligent safety execution module continuously acquires the real-time current data of each parallel battery cluster from the timing status data. and voltage data (in (Represents the battery cluster number operating in parallel).

[0186] The module analyzes real-time current data to accurately determine whether the current distribution between parallel battery clusters meets the preset circulating current suppression threshold. The above imbalance may indicate the presence of an actual output current in a particular battery cluster. Exceeding its preset overload threshold The determination of current imbalance can be achieved by calculating the deviation between the actual current of each battery cluster and the target current in its allocation command, or by calculating the maximum relative deviation between the currents of parallel battery clusters. For example, if there exists a certain Make or any If so, it is considered that there is an imbalance or overload.

[0187] If a current imbalance or overload is detected, the module immediately generates a circulating current suppression control signal. This signal is sent to the power unit of the relevant battery cluster, instructing it to fine-tune its output parameters, such as adjusting the output voltage setpoint or power command of the DC / DC converter, thereby actively correcting current deviations or mitigating overloads. For example, for a battery cluster with excessive output current, its power command can be slightly reduced; for a battery cluster with insufficient output current, its power command can be slightly increased. However, all adjustments are strictly limited to the safe operating range of the battery cluster. This active fine-tuning aims to eliminate current imbalances between parallel battery clusters, suppress potential circulating currents, mitigate overload risks, and prevent excessive wear or damage to local battery clusters.

[0188] If no current imbalance or overload is detected, it indicates that each battery cluster is operating well and as expected. At this time, the module generates a power output control signal, instructing the power unit of each battery cluster to strictly follow the target power value set in the allocation instruction. Power output is achieved. For example, by adjusting the PWM duty cycle or other control parameters of the power unit, the actual output power of each battery cluster is precisely matched with the allocation command.

[0189] By implementing intelligent pre-charging and circulating current suppression strategies, this module ensures the safe and stable operation of the ship's battery cluster system under various operating conditions, effectively avoiding problems such as start-up shock, parallel circulating current, and local overload, thereby improving system reliability and service life.

[0190] Example 2:

[0191] The lack of real-time monitoring and differentiated management mechanisms for the health status of individual battery clusters in parallel operation of marine energy storage systems can lead to battery clusters with poor health being assigned excessive loads, potentially causing safety hazards such as overheating, overcurrent, or accelerated aging. To address these issues, this invention provides a method for suppressing circulating current and intelligent pre-charging control of marine battery clusters. The method includes the following steps:

[0192] S01. Real-time acquisition of operational data from multiple battery clusters on the ship to obtain time-series status data including voltage, current, and temperature;

[0193] S02. By fusing time-series state data, the health status of each battery cluster is estimated using the extended Kalman filter algorithm, and dynamic health weights are generated.

[0194] S03. Based on dynamic health weights and the total power demand from the ship's energy management system, a multi-objective optimization algorithm is used to decouple the total power demand in order to generate allocation instructions and start / stop instructions that include the target power values ​​and start / stop states of each battery cluster.

[0195] S04. Based on the allocation instructions and start / stop instructions, and after obtaining timing status data, and under the premise of meeting the preset safety conditions including intelligent pre-charge logic and circulating current suppression logic, generate control signals for controlling the power units of each battery cluster.

[0196] Specifically, the system collects real-time operational data such as voltage, current, and temperature of multiple battery clusters on the ship using sensors to form time-series status data. Then, an extended Kalman filter algorithm is used to fuse this time-series data to dynamically estimate the health status of each battery cluster and generate dynamic health weights that reflect the health level. Next, based on the total power demand issued by the ship's energy management system and combined with the dynamic health weights of each battery cluster, a multi-objective optimization algorithm decouples the total power demand into specific power allocation commands and start / stop commands for each battery cluster. Finally, when executing these commands, intelligent pre-charging logic and circulating current suppression logic are introduced as preset safety conditions. When a battery cluster is connected, its voltage is first adjusted to be close to the bus voltage through the pre-charging circuit before being connected to the grid to avoid inrush current. At the same time, during parallel operation, the output of each battery cluster is monitored and adjusted in real time to suppress circulating current, thereby generating safe and reliable control signals to drive the power units of each battery cluster to operate according to the optimized allocation scheme.

[0197] Finally, it should be noted that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A ship battery cluster circulating current suppression and intelligent pre-charge control system, characterized in that, include: The multi-dimensional state perception module is equipped with a multi-channel synchronous sampling unit for collecting data from each battery cluster. It timestamps and aligns the data of each battery cluster with time sequence using a unified clock reference, and outputs time sequence state data. The data of each battery cluster includes voltage, current and temperature. The predictive health management module, based on the time-series state data, analyzes the time-series state data through a recurrent neural network model to predict the remaining usable life and the probability of failure, and uses an extended Kalman filter algorithm to fuse the time-series state data to estimate the health status, and outputs a dynamic health weight as a comprehensive result. The dynamic load allocation module, based on the dynamic health weight and the total power demand from the ship's energy management system, uses a multi-objective optimization algorithm, with the dynamic health weight as the allocation factor, to decouple the total power demand and generate allocation instructions and start / stop instructions containing the target power values ​​of each battery cluster. The intelligent safety execution module, based on the allocation instructions and start / stop instructions, and by acquiring the voltage data of the battery clusters from the timing status data, generates control signals for controlling the power units of each battery cluster, provided that preset safety conditions are met.

2. The ship battery cluster circulating current suppression and intelligent pre-charge control system according to claim 1, characterized in that, The steps of collecting data from each battery cluster, timestamping and aligning the data of each battery cluster using a unified clock reference, and outputting timing status data specifically include the following steps: The multi-channel synchronous sampling unit is triggered with a preset sampling period to synchronously collect data from each battery cluster; The current time value of the unified clock reference is appended as a timestamp to the collected data corresponding to the sampling period to form a timestamped data frame; The timestamped data frames are stored in a data buffer queue, and sorting, verification, and interpolation are performed based on the timestamps to achieve time alignment. Aligned data is output from the data buffer queue using a preset data structure to form the timing state data.

3. The ship battery cluster circulating current suppression and intelligent pre-charge control system according to claim 1, characterized in that, The step of analyzing the time-series state data using a recurrent neural network model to predict remaining usable lifetime and failure probability specifically includes the following steps: The time-series state data is constructed into a feature sequence with a preset time step, which is then used as input to a recurrent neural network model. The recurrent layer units of the recurrent neural network model extract temporal-dependent features from the feature sequence and generate a hidden state vector that encodes historical information. The hidden state vector is input into the regression prediction head and the probability output head, respectively; The regression prediction head outputs the predicted value of the remaining usable life, and the probability output head outputs the predicted value of the failure probability.

4. The ship battery cluster circulating current suppression and intelligent pre-charge control system according to claim 1, characterized in that, The step of fusing the time-series state data using the extended Kalman filter algorithm to estimate the health status and outputting a dynamic health weight specifically includes the following steps: A preset battery state space model is established, which is used to characterize the nonlinear relationship between the time-series state data and the internal state vector containing the health state. Based on the preset battery state space model and the health state estimate of the previous moment, a time update is performed to predict the prior value of the health state at the current moment. Using the current measured data in the time series state data, the error between the prior value of the health status at the current moment and the current measured data is calculated, and the measurement is updated in combination with Kalman gain to correct the prior value of the health status at the current moment and obtain the posterior estimate of the health status. The posterior estimate of the health status of each battery cluster is transformed using a preset mapping function to generate the dynamic health weight.

5. The ship battery cluster circulating current suppression and intelligent pre-charge control system according to claim 1, characterized in that, The step of generating allocation instructions and start / stop instructions containing target power values ​​for each battery cluster based on the dynamic health weight and the total power demand from the ship's energy management system, using a multi-objective optimization algorithm with the dynamic health weight as the allocation factor, and decoupling the total power demand, specifically includes the following steps: Construct a multi-objective optimization model; Solving multi-objective optimization problems: Using multi-objective optimization algorithms, the multi-objective optimization model is solved to obtain a set of Pareto optimal solutions; Select the optimal solution and generate instructions: From the Pareto optimal solution set, determine the final allocation scheme according to the preset decision strategy, and extract the target power value and start / stop status of each battery cluster according to the final allocation scheme, and generate the allocation instruction and start / stop instruction respectively.

6. The ship battery cluster circulating current suppression and intelligent pre-charge control system according to claim 5, characterized in that, The construction of the multi-objective optimization model specifically includes the following steps: Determine the decision variables, including the target power value and start / stop status of each battery cluster; Define an objective function that includes the following: Objective function 1: Minimize the deviation between the sum of the target power values ​​of each battery cluster and the total power demand from the ship's energy management system; Objective function two: Using the dynamic health weight as an allocation factor, differentially adjust the target power value of each battery cluster; The definitions include the maximum and minimum power output limits for each battery cluster, the conditions for meeting the total power demand, and the constraints related to the start / stop status and power output logic of each battery cluster.

7. The ship battery cluster circulating current suppression and intelligent pre-charge control system according to claim 1, characterized in that, The step of generating control signals for controlling each battery cluster power unit based on the allocation command and start / stop command, and obtaining voltage data from the timing state data, under the premise of meeting preset safety conditions, specifically includes the following steps: Perform intelligent pre-charging: Receive the start / stop command; When the start / stop command instructs the battery cluster to switch from a non-operating state to an operating state, the real-time voltage of the battery cluster and the system bus voltage are obtained from the timing status data; Determine whether the voltage difference between the real-time voltage of the battery cluster and the bus voltage exceeds a preset pre-charge safety threshold. If the voltage difference exceeds the preset precharge safety threshold, a precharge control signal is generated to control the power unit of the battery cluster to perform a precharge operation until the voltage difference drops below the preset precharge safety threshold. If the voltage difference is within the preset precharge safety threshold or precharge is completed, a start-up control signal is generated to control the power unit of the battery cluster to be connected to the bus. Perform circulating current suppression and safe output: Receive the allocation instruction; During the parallel operation of multiple battery clusters, the real-time current and voltage data of each parallel battery cluster in the time-series state data are acquired; Analyze the real-time current data to determine whether there is an imbalance in the current distribution among the parallel battery clusters that exceeds a preset circulating current suppression threshold, or whether there is a preset overload threshold. If an imbalance or overload exists, a circulating current suppression control signal is generated to fine-tune the output parameters of the relevant battery cluster power units in order to eliminate the imbalance or alleviate the overload. If no imbalance or overload is detected, a power output control signal is generated to control each battery cluster power unit to output power according to the allocation instruction.

8. A method for suppressing circulating current and intelligent pre-charge control of ship battery clusters, characterized in that, The method for a ship battery cluster circulating current suppression and intelligent pre-charge control system according to any one of claims 1-7 includes the following steps: S01. Real-time acquisition of the operating data of the ship's multi-battery cluster to obtain time-series status data including voltage, current, and temperature; S02. The time-series state data is fused, and the health status of each battery cluster is estimated using the extended Kalman filter algorithm, and dynamic health weights are generated. S03. Based on the dynamic health weight and the total power demand from the ship's energy management system, a multi-objective optimization algorithm is used to decouple the total power demand to generate allocation instructions and start / stop instructions that include the target power value and start / stop status of each battery cluster. S04. Based on the allocation instruction and start / stop instruction, and obtaining the timing status data, and under the premise of satisfying the preset safety conditions including intelligent pre-charge logic and circulating current suppression logic, generate control signals for controlling the power units of each battery cluster.