Battery power system control method and system based on digital twin system driving

Abstract

The invention discloses a battery power system control method and system based on digital twinborn system driving. The method comprises the following steps: acquiring operation state data of a battery power system in real time based on a digital twinborn model; inputting the acquired operation state data into a double-layer guide mechanism formed by twin model proactive decision control and real-time data dynamic collaborative correction, and generating a preliminary control instruction; optimizing the preliminary control instruction through a multi-agent collaborative decision-making mechanism to obtain an optimized control instruction; and based on the sequential relation dynamic graph structure and the cross-unit linkage constraint, verifying the optimized control instruction by adopting a model prediction control algorithm, and outputting a final control instruction. According to the invention, the requirements of efficient, safe and reliable operation of the heavy truck battery power system under complex working conditions can be met.

Description

Technical Field

[0001] The present invention relates to the field of power battery technology, and in particular to a battery power system control method and system driven by a digital twin system. Background Art

[0002] In the field of heavy-duty truck battery power system control, existing technologies mostly use control strategies based on empirical rules and independent subsystem control methods. A common practice is to first formulate a set of fixed charging, discharging and power output rules based on battery characteristics and motor operating parameters. For example, during the charging process, a constant charging current is set based on the current battery power and the preset power range, without considering the impact of the complex electrochemical processes and temperature changes inside the battery on charging efficiency and safety. In terms of power output control, it is mainly based on the driver's operating instructions, and the torque and speed of the drive motor are adjusted through the vehicle controller. There is a lack of comprehensive consideration of the battery's real-time health status, vehicle driving conditions and brake energy recovery potential. In addition, the battery management system, motor control system and vehicle driving control system usually operate independently, and the information exchange between the subsystems is limited, making it difficult to achieve collaborative optimization.

[0003] Therefore, the existing battery power system control method cannot meet the requirements of heavy truck battery power system for efficient, safe and reliable operation under complex working conditions. Summary of the Invention

[0004] The embodiments of the present invention provide a battery power system control method and system driven by a digital twin system, which can meet the requirements of efficient, safe and reliable operation of heavy-duty truck battery power systems under complex working conditions.

[0005] An embodiment of the present invention provides a battery power system control method based on digital twin system drive, including:

[0006] Obtain real-time operating status data of the battery power system based on the digital twin model;

[0007] The acquired operating status data is input into a two-layer guidance mechanism consisting of proactive decision-making and control by the twin model and dynamic collaborative correction of real-time data to generate preliminary control instructions;

[0008] The initial control instructions are optimized through a multi-agent collaborative decision-making mechanism to obtain optimized control instructions. The multi-agent collaborative decision-making mechanism includes: the energy management agent optimizes power distribution, the charging scheduling agent optimizes braking energy recovery, and the vehicle control agent optimizes driving strategy.

[0009] Based on the dynamic graph structure of timing relationships and cross-unit linkage constraints, the model predictive control algorithm is used to verify the optimized control instructions and output the final control instructions.

[0010] As an improvement to the above solution, the real-time acquisition of operating status data of the battery power system based on the digital twin model includes the following sub-steps:

[0011] Utilize sensors deployed at key locations within the battery power system to collect basic data on voltage, current, temperature, and battery state of charge, generating a raw data set containing a variety of basic information.

[0012] The original data set is denoised and then normalized to a specific numerical range to obtain preprocessed data;

[0013] The preprocessed data is input into the constructed digital twin model. The digital twin model calculates and matches the input data with its own model parameters, and outputs operating status data corresponding to the actual operating status of the battery power system.

[0014] As an improvement to the above solution, the acquired operating status data is input into a two-layer guidance mechanism consisting of proactive decision-making and control of the twin model and dynamic collaborative correction of real-time data to generate preliminary control instructions, including the following sub-steps:

[0015] Input the operating status data into the twin model. The twin model predicts the operating status of the battery power system in the future based on the preset algorithm and historical data, and obtains the predicted operating status results.

[0016] Based on the predicted operating state results, combined with the performance indicators and safety constraints of the battery power system, and using pre-set decision rules, multiple candidate proactive decision plans are generated to form a proactive decision plan set;

[0017] Compare and analyze the operating status data with the results predicted by the twin model, calculate the deviation between the two, and adjust some parameters of the twin model based on the deviation to obtain the corrected twin model;

[0018] The corrected twin model and the set of proactive decision-making schemes are comprehensively evaluated, and the most appropriate proactive decision-making scheme is selected based on the evaluation results and output as the preliminary control instruction.

[0019] As an improvement to the above solution, the preliminary control instructions are optimized through a multi-agent collaborative decision-making mechanism to obtain optimized control instructions. The multi-agent collaborative decision-making mechanism includes: the energy management agent optimizes power distribution, the charging scheduling agent optimizes braking energy recovery, and the vehicle control agent optimizes driving strategy, including the following sub-steps:

[0020] The preliminary control instructions are passed to the energy management agent, the charging scheduling agent, and the vehicle control agent. Each agent parses the preliminary control instructions according to its own task and obtains the instruction content related to its own task;

[0021] The energy management agent adjusts the power allocation instructions based on the battery's current state of charge, remaining power, battery health status, and vehicle power demand information, calculates a more reasonable power allocation plan, and obtains the optimized power allocation instructions;

[0022] The charging scheduling agent optimizes the braking energy recovery instructions based on the vehicle's driving status, battery charging capacity, and historical braking energy recovery data, determines the best braking energy recovery strategy, and obtains the optimized braking energy recovery instructions;

[0023] The vehicle control agent adjusts the driving strategy instructions based on road conditions, traffic rules, and vehicle dynamic characteristics, plans a more reasonable driving strategy, and obtains the optimized driving strategy instructions;

[0024] The optimized power distribution instructions, brake energy recovery instructions and driving strategy instructions are integrated to form optimized control instructions.

[0025] As an improvement to the above scheme, the method based on the temporal relationship dynamic graph structure and cross-unit linkage constraints uses a model predictive control algorithm to verify the optimized control instructions and output the final control instructions, including the following sub-steps:

[0026] Based on the optimized control instructions, combined with the connection relationships, information transmission relationships, and time series information between the various components of the battery power system, a time series dynamic graph structure is constructed. The meaning and attributes of each node and edge in the graph structure are clarified to obtain the initial time series dynamic graph.

[0027] Analyze the mutual influence and constraints between different units in the battery power system, convert these relationships into specific constraints, and add them to the initial timing relationship dynamic graph structure to form a timing relationship dynamic graph with cross-unit linkage constraints;

[0028] The dynamic graph of the time series relationship with cross-unit linkage constraints is used as the input of the model predictive control algorithm. The model predictive control algorithm simulates and evaluates the optimized control instructions based on the current system state and the predicted future state, and calculates various indicators during the execution of the instructions.

[0029] Based on the evaluation results of the model predictive control algorithm, it is judged whether the optimized control instructions meet the system requirements. If they do, they are output as the final control instructions. If they do not meet the requirements, the optimized control instructions are adjusted and simulated and evaluated again until the final control instructions that meet the requirements are obtained.

[0030] Another embodiment of the present invention provides a battery power system control system driven by a digital twin system, including:

[0031] An acquisition module is used to obtain the operating status data of the battery power system in real time based on the digital twin model;

[0032] A generation module is used to input the acquired operating status data into a two-layer guidance mechanism consisting of proactive decision-making and control based on the twin model and dynamic collaborative correction based on real-time data to generate preliminary control instructions;

[0033] An optimization module is used to optimize the initial control instructions through a multi-agent collaborative decision-making mechanism to obtain optimized control instructions. The multi-agent collaborative decision-making mechanism includes: the energy management agent optimizes power distribution, the charging scheduling agent optimizes braking energy recovery, and the vehicle control agent optimizes driving strategy;

[0034] The verification module is used to verify the optimized control instructions based on the dynamic graph structure of the timing relationship and the cross-unit linkage constraints, and output the final control instructions using the model predictive control algorithm.

[0035] As an improvement to the above solution, the acquisition module is specifically configured to:

[0036] Utilize sensors deployed at key locations within the battery power system to collect basic data on voltage, current, temperature, and battery state of charge, generating a raw data set containing a variety of basic information.

[0037] The original data set is denoised and then normalized to a specific numerical range to obtain preprocessed data;

[0038] The preprocessed data is input into the constructed digital twin model. The digital twin model calculates and matches the input data with its own model parameters, and outputs operating status data corresponding to the actual operating status of the battery power system.

[0039] As an improvement to the above solution, the generation module is specifically used to:

[0040] Input the operating status data into the twin model. The twin model predicts the operating status of the battery power system in the future based on the preset algorithm and historical data, and obtains the predicted operating status results.

[0041] Based on the predicted operating state results, combined with the performance indicators and safety constraints of the battery power system, and using pre-set decision rules, multiple candidate proactive decision plans are generated to form a proactive decision plan set;

[0042] Compare and analyze the operating status data with the results predicted by the twin model, calculate the deviation between the two, and adjust some parameters of the twin model based on the deviation to obtain the corrected twin model;

[0043] The corrected twin model and the set of proactive decision-making schemes are comprehensively evaluated, and the most appropriate proactive decision-making scheme is selected based on the evaluation results and output as the preliminary control instruction.

[0044] As an improvement to the above solution, the optimization module is specifically used to:

[0045] The preliminary control instructions are passed to the energy management agent, the charging scheduling agent, and the vehicle control agent. Each agent parses the preliminary control instructions according to its own task and obtains the instruction content related to its own task;

[0046] The energy management agent adjusts the power allocation instructions based on the battery's current state of charge, remaining power, battery health status, and vehicle power demand information, calculates a more reasonable power allocation plan, and obtains the optimized power allocation instructions;

[0047] The charging scheduling agent optimizes the braking energy recovery instructions based on the vehicle's driving status, battery charging capacity, and historical braking energy recovery data, determines the best braking energy recovery strategy, and obtains the optimized braking energy recovery instructions;

[0048] The vehicle control agent adjusts the driving strategy instructions based on road conditions, traffic rules, and vehicle dynamic characteristics, plans a more reasonable driving strategy, and obtains the optimized driving strategy instructions;

[0049] The optimized power distribution instructions, brake energy recovery instructions and driving strategy instructions are integrated to form optimized control instructions.

[0050] As an improvement to the above solution, the verification module is specifically used to:

[0051] Based on the optimized control instructions, combined with the connection relationships, information transmission relationships, and time series information between the various components of the battery power system, a time series dynamic graph structure is constructed. The meaning and attributes of each node and edge in the graph structure are clarified to obtain the initial time series dynamic graph.

[0052] Analyze the mutual influence and constraints between different units in the battery power system, convert these relationships into specific constraints, and add them to the initial timing relationship dynamic graph structure to form a timing relationship dynamic graph with cross-unit linkage constraints;

[0053] The dynamic graph of the time series relationship with cross-unit linkage constraints is used as the input of the model predictive control algorithm. The model predictive control algorithm simulates and evaluates the optimized control instructions based on the current system state and the predicted future state, and calculates various indicators during the execution of the instructions.

[0054] Based on the evaluation results of the model predictive control algorithm, it is judged whether the optimized control instructions meet the system requirements. If they do, they are output as the final control instructions. If they do not meet the requirements, the optimized control instructions are adjusted and simulated and evaluated again until the final control instructions that meet the requirements are obtained.

[0055] Compared with the prior art, the embodiments of the present invention have the following beneficial effects:

[0056] First, a digital twin model collects real-time operating status data from all levels of the battery powertrain (from single cells to the entire pack, controller, and drive motor). Real-time mapping between the virtual model and the physical system ensures the comprehensiveness and accuracy of the acquired data. This data is then fed into a two-tiered guidance mechanism consisting of proactive decision-making and control based on the twin model and dynamic collaborative correction based on real-time data. Proactive decision-making predicts system status based on historical data and simulations, while real-time data correction adjusts model parameters based on actual operational deviations. These two mechanisms combine to generate preliminary control instructions, balancing the foresight and real-time nature of the predictions. Subsequently, a multi-agent collaborative decision-making mechanism is employed to decompose these preliminary instructions into subtasks, such as power allocation, regenerative braking, and driving strategy. Each agent independently calculates and collaboratively adjusts based on its own optimization objective (e.g., the energy management agent optimizes power allocation, and the charging scheduling agent optimizes regenerative braking), achieving a balance between local optimization and global optimization. Finally, a dynamic graph structure based on temporal relationships intuitively presents the spatiotemporal relationships between system components. Physical constraints are set by combining cross-unit linkage constraints. The optimized control instructions are then subjected to rolling optimization and feasibility verification using a model predictive control algorithm, ultimately outputting final control instructions that align with the system's actual operation. Compared with traditional control methods that have problems such as insufficient multi-physical field coupling, real-time decision-making lag, and poor subsystem coordination, the embodiments of the present invention use digital twin models to achieve accurate perception of system status, a two-layer guidance mechanism to enhance the dynamic adaptability of decision-making, multi-agent collaborative optimization of system resource allocation, and timing diagrams and model predictive control to ensure the feasibility of instructions, thereby achieving full-process and multi-dimensional precise control of the battery power system, significantly improving system operating efficiency, energy utilization and adaptability to working conditions. BRIEF DESCRIPTION OF THE DRAWINGS

[0057] Figure 1 This is a flow chart of a method for controlling a battery power system driven by a digital twin system provided by one embodiment of the present invention;

[0058] Figure 2 This is a structural diagram of a battery power system control system driven by a digital twin system provided by one embodiment of the present invention. DETAILED DESCRIPTION

[0059] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0060] See also Figure 1 , is a flow chart of a method for controlling a battery power system driven by a digital twin system according to an embodiment of the present invention. The method for controlling a battery power system driven by a digital twin system comprises the following steps:

[0061] S10, real-time acquisition of battery power system operating status data based on the digital twin model;

[0062] S11, input the acquired operating status data into a two-layer guidance mechanism consisting of proactive decision-making and control of the twin model and dynamic collaborative correction of real-time data to generate preliminary control instructions;

[0063] S12, optimizing the preliminary control instructions through a multi-agent collaborative decision-making mechanism to obtain optimized control instructions; wherein the multi-agent collaborative decision-making mechanism includes: the energy management agent optimizing power distribution, the charging scheduling agent optimizing braking energy recovery, and the vehicle control agent optimizing driving strategy;

[0064] S13, based on the timing relationship dynamic graph structure and cross-unit linkage constraints, the model predictive control algorithm is used to verify the optimized control instructions and output the final control instructions.

[0065] Compared with the prior art, the embodiments of the present invention have the following beneficial effects:

[0066] First, a digital twin model collects real-time operating status data from all levels of the battery powertrain (from single cells to the entire pack, controller, and drive motor). Real-time mapping between the virtual model and the physical system ensures the comprehensiveness and accuracy of the acquired data. This data is then fed into a two-tiered guidance mechanism consisting of proactive decision-making and control based on the twin model and dynamic collaborative correction based on real-time data. Proactive decision-making predicts system status based on historical data and simulations, while real-time data correction adjusts model parameters based on actual operational deviations. These two mechanisms combine to generate preliminary control instructions, balancing the foresight and real-time nature of the predictions. Subsequently, a multi-agent collaborative decision-making mechanism is employed to decompose these preliminary instructions into subtasks, such as power allocation, regenerative braking, and driving strategy. Each agent independently calculates and collaboratively adjusts based on its own optimization objective (e.g., the energy management agent optimizes power allocation, and the charging scheduling agent optimizes regenerative braking), achieving a balance between local optimization and global optimization. Finally, a dynamic graph structure based on temporal relationships intuitively presents the spatiotemporal relationships between system components. Physical constraints are set by combining cross-unit linkage constraints. The optimized control instructions are then subjected to rolling optimization and feasibility verification using a model predictive control algorithm, ultimately outputting final control instructions that align with the system's actual operation. Compared with traditional control methods that have problems such as insufficient multi-physical field coupling, real-time decision-making lag, and poor subsystem coordination, the embodiments of the present invention use digital twin models to achieve accurate perception of system status, a two-layer guidance mechanism to enhance the dynamic adaptability of decision-making, multi-agent collaborative optimization of system resource allocation, and timing diagrams and model predictive control to ensure the feasibility of instructions, thereby achieving full-process and multi-dimensional precise control of the battery power system, significantly improving system operating efficiency, energy utilization and adaptability to working conditions.

[0067] As one example, the real-time acquisition of operating status data of a battery power system based on a digital twin model includes the following sub-steps:

[0068] Utilize sensors deployed at key locations within the battery power system to collect basic data on voltage, current, temperature, and battery state of charge, generating a raw data set containing a variety of basic information.

[0069] The original data set is denoised and then normalized to a specific numerical range to obtain preprocessed data;

[0070] The preprocessed data is input into the constructed digital twin model. The digital twin model calculates and matches the input data with its own model parameters, and outputs operating status data corresponding to the actual operating status of the battery power system.

[0071] In this embodiment, a quantitative matching system between data features and control strategies is established to achieve precise decision-making for optimizing the operation of a new energy heavy-duty truck battery powertrain. First, the action sequence and expected system response of each candidate control strategy are abstracted into a feature vector representation, converting the abstract control strategy into quantifiable and comparable digital information, providing a unified data foundation for subsequent evaluation. Next, the cosine similarity metric is used to calculate the similarity between the unified feature representation of the system's operating state and the feature vectors of each candidate control strategy. Cosine similarity effectively measures the degree of directional similarity between vectors, thereby quantifying the degree of match between each strategy and the current system state. The candidate control strategies are then ranked based on the calculated similarity values, and the strategy with the highest similarity value is selected as the optimal control strategy, ensuring that the selected strategy best matches the actual system operating state. Finally, the optimal control strategy is converted into specific control instructions to drive the actual operation of the battery powertrain. In summary, by establishing a standardized feature representation and quantitative evaluation system, this embodiment avoids the subjectivity and blindness inherent in traditional control strategy selection and achieves a precise mapping from data features to control strategies. This scientific matching and evaluation mechanism can quickly and accurately select the solution that best suits the current system state from multiple candidate control strategies, significantly improving the efficiency and accuracy of operation optimization decisions for new energy heavy-duty truck battery power systems, ensuring that the system always operates in the optimal state under complex and changeable working conditions, and improving energy utilization efficiency and system stability.

[0072] For ease of understanding, the working process of this embodiment is as follows:

[0073] First, a sensor array is deployed at the core components of the battery power system. High-precision voltage sensors are connected to the positive and negative terminals of each battery module to measure cell voltage. Hall effect current sensors are connected in series in the main circuit to obtain real-time charge and discharge currents. Temperature sensors are installed at key locations within the battery pack and on the outer shell to monitor core and ambient temperatures. Furthermore, an existing algorithm combining ampere-hour integration with Kalman filtering is used to accurately estimate the battery's state of charge (SOC). These sensors continuously collect data at a fixed frequency, integrating the collected voltage, current, temperature, and SOC data into a raw data set.

[0074] Next, the raw data set is preprocessed. First, the data is denoised using a median filter. This effectively removes random noise by replacing the value at a specific position in the data sequence with the median value of the data within a certain range. After denoising, the data is mapped uniformly to the [0, 1] range using the min-max normalization method. This method uses a linear transformation based on the maximum and minimum values in the raw data to compress the data into a specified range, facilitating subsequent processing.

[0075] Finally, the preprocessed data is input into the constructed digital twin model. The digital twin model uses an improved recurrent neural network (RNN) structure combined with an adaptive weight adjustment algorithm. Its core calculation logic is:

[0076] h t =σ(W xh x t +W hh h t-1 +b h )

[0077] y t =W hy h t +b y

[0078] Among them, x t Represents the pre-processed data input at time t, including voltage, current and other information; h t is the hidden layer state at time h; σ is the activation function, which is used to introduce nonlinearity; W xh is the weight matrix from the input layer to the hidden layer, W hh is the weight matrix from hidden layer to hidden layer, b h is the bias vector of the hidden layer; W hy is the weight matrix from the hidden layer to the output layer, b y is the bias vector of the output layer. During model training, the existing backpropagation algorithm dynamically adjusts the weight matrix and bias vector based on the error between the actual operating data and the model output data, ensuring that the operating status data output by the model accurately reflects the actual operating conditions of the battery power system.

[0079] As one example, the acquired operating status data is input into a two-layer guidance mechanism consisting of proactive decision-making control of a twin model and dynamic collaborative correction of real-time data to generate preliminary control instructions, including the following sub-steps:

[0080] Input the operating status data into the twin model. The twin model predicts the operating status of the battery power system in the future based on the preset algorithm and historical data, and obtains the predicted operating status results.

[0081] Based on the predicted operating state results, combined with the performance indicators and safety constraints of the battery power system, and using pre-set decision rules, multiple candidate proactive decision plans are generated to form a proactive decision plan set;

[0082] Compare and analyze the operating status data with the results predicted by the twin model, calculate the deviation between the two, and adjust some parameters of the twin model based on the deviation to obtain the corrected twin model;

[0083] The corrected twin model and the set of proactive decision-making schemes are comprehensively evaluated, and the most appropriate proactive decision-making scheme is selected based on the evaluation results and output as the preliminary control instruction.

[0084] This embodiment leverages the advantages of digital twin models to achieve precise control of the battery power system by building a closed-loop "prediction-decision-correction-evaluation" mechanism. First, real-time operating status data is input into the twin model. Preset algorithms and historical data are used to predict the system's future operating status, enabling the system to make forward-looking decisions. Next, based on the prediction results, combined with the battery power system's performance indicators and safety constraints, multiple candidate proactive decision-making solutions are generated using pre-set decision rules, providing diverse options for system operation. Subsequently, by comparing actual operating status data with the model's prediction results, deviations are calculated and the twin model parameters are dynamically adjusted to ensure a high degree of match between the model and the physical system. Finally, a comprehensive evaluation is conducted on the corrected twin model and the set of proactive decision-making solutions, selecting the solution that best suits the current system state as the initial control instruction. In summary, this embodiment effectively overcomes the problems of decision-making lag and disconnection between model and actual operation in traditional control methods. It achieves early intervention in system operation trends through proactive decision-making, improves model accuracy through dynamic correction using real-time data, and ensures the rationality and effectiveness of control instructions through comprehensive evaluation. It significantly enhances the real-time, accuracy and adaptability of battery power system control, and improves system operation efficiency and safety.

[0085] In this embodiment, the acquired operating status data is input into a two-layer guidance mechanism consisting of proactive decision-making and control by the twin model and dynamic collaborative correction of real-time data. The specific workflow for generating preliminary control instructions is as follows:

[0086] Step 1: Input the real-time operating status data obtained based on the digital twin model into the twin model. The twin model uses an improved time series prediction algorithm and uses the formula Make predictions. represents the predicted operating status result vector at the future time τ, including the predicted values of battery voltage, current, temperature, and state of charge; X t-i is the actual operating status data at the current time t and the previous n times; W i is the weight matrix corresponding to the data at that moment, which is obtained by training and optimizing historical data; b is the bias vector; σ is the activation function, which performs a nonlinear transformation on the weighted summation result to make the prediction result fit the actual change trend of the battery power system, and then output the predicted operating status result.

[0087] Step 2: Based on the above predicted operating state results, combined with the performance indicators of the battery power system (such as energy conversion efficiency, power output stability) and safety constraints (such as battery overcharge and over-discharge thresholds, temperature safety range). Use the pre-set decision rules based on the multi-objective optimization framework to construct the objective function F(a) = ω1·E(a) + ω2·S(a) + ω3·L(a). Among them, a represents the decision plan; E(a) is the system energy efficiency index under plan a; S(a) is the safety risk index under plan a, the lower the value, the safer; L(a) is the battery life loss index under plan a; ω1, ω2, ω3 are the weight coefficients of the corresponding indicators, which are set according to actual needs. By enumerating different control parameter combinations, multiple candidate proactive decision plans are generated, involving power regulation strategies, temperature control strategies, etc., and finally forming a proactive decision plan set A = {a1, a2, ..., a m}.

[0088] Step 3: Compare and analyze the real-time operating status data with the results predicted by the twin model one by one according to the parameters, and calculate the deviation. Taking the battery voltage as an example, the deviation is calculated as Where V real is the actual measured battery voltage, is the battery voltage predicted by the twin model. Based on the deviation of all parameters, the gradient descent algorithm with adaptive learning rate is used to adjust some parameters of the twin model. Let the model parameter be θ, and the parameter update formula is Among them, θ new and θ old are the updated parameters and the parameters before the update respectively; η is the learning rate dynamically adjusted according to the deviation; is the gradient of the loss function with respect to the parameter θ. The loss function is constructed based on the mean square error between the predicted value and the actual value, thereby obtaining the calibrated twin model.

[0089] Step 4: Comprehensively evaluate the corrected twin model and proactive decision-making solution set. i ∈A, simulate the system operation status after the implementation of the solution with the help of the corrected twin model to obtain a series of evaluation index data. Construct the evaluation function Among them, Score(a i ) is option a i The comprehensive score of j Is the jth evaluation metric Metric j Weight; evaluation indicator Metric jThis includes indicators such as prediction accuracy improvement, system stability, and energy utilization improvement. Based on the evaluation results, the proactive decision-making solution with the highest score is selected and converted into corresponding control command parameters. This is output as the preliminary control command for subsequent control of the battery power system.

[0090] As one example, the multi-agent collaborative decision-making mechanism is used to optimize the preliminary control instructions to obtain optimized control instructions. The multi-agent collaborative decision-making mechanism includes: the energy management agent optimizes power distribution, the charging scheduling agent optimizes braking energy recovery, and the vehicle control agent optimizes driving strategy, including the following sub-steps:

[0091] The preliminary control instructions are passed to the energy management agent, the charging scheduling agent, and the vehicle control agent. Each agent parses the preliminary control instructions according to its own task and obtains the instruction content related to its own task;

[0092] The energy management agent adjusts the power allocation instructions based on the battery's current state of charge, remaining power, battery health status, and vehicle power demand information, calculates a more reasonable power allocation plan, and obtains the optimized power allocation instructions;

[0093] The charging scheduling agent optimizes the braking energy recovery instructions based on the vehicle's driving status, battery charging capacity, and historical braking energy recovery data, determines the best braking energy recovery strategy, and obtains the optimized braking energy recovery instructions;

[0094] The vehicle control agent adjusts the driving strategy instructions based on road conditions, traffic rules, and vehicle dynamic characteristics, plans a more reasonable driving strategy, and obtains the optimized driving strategy instructions;

[0095] The optimized power distribution instructions, brake energy recovery instructions and driving strategy instructions are integrated to form optimized control instructions.

[0096] This embodiment aims to build a multi-agent collaborative decision-making system, achieving refined optimization of battery power system control instructions through division of labor, collaboration, and information exchange. First, the preliminary control instructions are disassembled and transmitted to the three agents: energy management, charging scheduling, and vehicle control. Each agent interprets the instructions based on its own functional positioning and clarifies the division of tasks. Subsequently, the energy management agent dynamically adjusts the power allocation plan based on core parameters such as the battery state of charge and remaining power to ensure that energy supply and vehicle power requirements are accurately matched. The charging scheduling agent optimizes the braking energy recovery strategy based on the vehicle's driving status, battery charging capacity, and historical recovery data to maximize energy recovery efficiency. The vehicle control agent plans driving strategies based on real-time road conditions, traffic regulations, and vehicle dynamics, balancing energy consumption and driving efficiency. Finally, the optimized instructions output by each agent are integrated to form optimized control instructions that take into account energy management, recovery efficiency, and driving performance. In summary, this embodiment effectively solves the problem of independent operation and lack of coordination of each subsystem in traditional control methods. Through the division of labor and cooperation of multiple intelligent agents, targeted optimization of each link of the battery power system operation is achieved. At the same time, information sharing and interaction are used to achieve global optimization, which significantly improves the system's energy utilization efficiency, endurance and operational stability, ensuring that all parts of the system operate in a coordinated and efficient manner under complex working conditions.

[0097] In this embodiment, the preliminary control instructions are optimized through a multi-agent collaborative decision-making mechanism. The specific workflow is as follows:

[0098] The first step is to transmit preliminary control instructions to the energy management agent, charging scheduling agent, and vehicle control agent in a unified communication protocol format. Each agent deconstructs the preliminary control instructions according to a pre-set instruction parsing protocol. For example, the energy management agent separates the instruction parameters related to power distribution, including the initial power distribution ratio and power adjustment range; the charging scheduling agent extracts basic control instructions for brake energy recovery, such as the recovery power lower limit and recovery time threshold; and the vehicle control agent reads the initial instructions for the driving strategy, including the initial vehicle speed setting and the driving path guidance direction. This allows each agent to obtain the instruction information corresponding to its own task, laying the foundation for subsequent optimization operations.

[0099] Step 2: The energy management agent obtains the current state of charge (SOC) of the battery (value range 0-1, estimated by ampere-hour integration combined with Kalman filter algorithm), the remaining power Q remain (unit: Ah), calculated by multiplying the battery capacity and the state of charge), battery health status indicator SOH (reflects the degree of battery performance degradation, 1 represents a brand new state, and is comprehensively evaluated through parameters such as battery internal resistance and capacity attenuation), and vehicle power demand information, including current vehicle speed v (unit: km / h), collected in real time by the vehicle speed sensor), acceleration demand areq (Unit: m / s 2 , calculated based on the driver’s accelerator pedal depth and the vehicle dynamics model).

[0100] An improved adaptive power allocation algorithm is used, and the algorithm formula is:

[0101]

[0102] Among them, P opt is the optimized power distribution value (unit W); P total is the total power that the battery can currently output (unit: W, related to the battery voltage and internal resistance); P rated is the rated power of the battery (unit: W); P dyn It is the dynamic power (unit W) calculated based on the vehicle power demand, through the vehicle dynamics equation P dyn =F·v, where F is the driving force required for the vehicle to move (unit: N), which is determined by factors such as vehicle mass, driving resistance, and acceleration resistance; ω soc 、ω soh 、ω demand They are the weight coefficients of state of charge, battery health status, and power demand. The initial values are set based on experience and dynamically adjusted through an online reinforcement learning algorithm with the goal of minimizing battery life loss, maximizing energy utilization efficiency, and ensuring that power response meets demand, thereby obtaining the optimized power allocation instructions.

[0103] Step 3: The charging scheduling agent is based on the vehicle's driving status, including the current speed v (unit km / h), acceleration a (unit m / s 2 ), brake pedal travel p brake (unit: mm, reflects the braking intensity); the battery charging capacity is determined by the current battery temperature T (unit: °C, measured by a temperature sensor), state of charge SOC and other parameters, according to the charging characteristic curve provided by the battery manufacturer to determine the current maximum allowable charging power P max-charge (unit: W); and historical data on brake energy recovery, such as the average recovery efficiency η under different driving conditions over a period of time avg .

[0104] Using the improved braking energy recovery optimization algorithm, the theoretical energy E that can be recovered during braking is first calculated according to the kinetic energy theorem. theo :

[0105]

[0106] Where m is the total mass of the vehicle (in kg), v start is the vehicle speed at the start of braking (unit: m / s), v endis the vehicle speed at the end of braking (in m / s).

[0107] Then, the actual recovered energy E is determined by combining the battery charging capacity and historical recovery efficiency. real :

[0108]

[0109] Where η is the real-time recovery efficiency (ranging from 0 to 1) obtained through adaptive fuzzy control algorithm based on historical data and current working conditions; P recover-theo is the theoretical recovery power (unit W), given by Calculation, Δt is the braking duration (unit: s). Finally, according to the actual recovered energy E real , determine the optimal braking energy recovery strategy, including parameters such as recovery power and recovery time, and then obtain the optimized braking energy recovery instruction.

[0110] Step 4: The vehicle control agent obtains the road slope α, curvature k and traffic rules information such as speed limit v according to the road condition information through on-board sensors (such as lidar, camera) and map data. limit , traffic light status; and the vehicle's dynamic characteristics, including vehicle mass m, tire friction coefficient μ, maximum driving force F max .

[0111] The improved multi-objective driving strategy optimization algorithm is used to construct the objective function:

[0112] J=λ time ·t est +λ energy ·E est +λ safety ·S eval

[0113] Among them, t est is the expected travel time, estimated by road length, speed limit, and vehicle dynamics model; E est To estimate energy consumption, calculate based on the driving path and vehicle power system characteristics; S eval is the driving risk assessment value (dimensionless), which is determined by comprehensively considering factors such as road conditions and traffic rule violation risks; time ,λ energy ,λ safety It is the weight coefficient of the corresponding indicator, which is dynamically adjusted according to travel demand and vehicle status.

[0114] Through genetic algorithms, under the conditions of satisfying traffic rules and vehicle dynamics constraints (such as maximum acceleration, maximum braking force, and tire grip limit), parameters such as driving speed v, acceleration a, and steering angle θ are iteratively optimized to solve the minimum value of the objective function, thereby planning a more reasonable driving strategy, including the optimal vehicle speed sequence, acceleration / deceleration timing, steering operation plan, etc., and obtaining the optimized driving strategy instructions.

[0115] Step 5: Integrate the optimized power distribution instructions, brake energy recovery instructions, and driving strategy instructions according to the pre-defined control instruction data structure. The power distribution parameters in the power distribution instructions, the recovery strategy parameters in the brake energy recovery instructions, and the driving parameters in the driving strategy instructions are uniformly encoded into a complete instruction data packet. The data packet contains fields such as instruction identification, parameter type, parameter value, and checksum, ultimately forming the optimized control instructions for precise control of the battery power system.

[0116] As one example, the method of verifying the optimized control instructions using a model predictive control algorithm based on a dynamic graph structure of a timing relationship and cross-unit linkage constraints and outputting a final control instruction includes the following sub-steps:

[0117] Based on the optimized control instructions, combined with the connection relationships, information transmission relationships, and time series information between the various components of the battery power system, a time series dynamic graph structure is constructed. The meaning and attributes of each node and edge in the graph structure are clarified to obtain the initial time series dynamic graph.

[0118] Analyze the mutual influence and constraints between different units in the battery power system, convert these relationships into specific constraints, and add them to the initial timing relationship dynamic graph structure to form a timing relationship dynamic graph with cross-unit linkage constraints;

[0119] The dynamic graph of the time series relationship with cross-unit linkage constraints is used as the input of the model predictive control algorithm. The model predictive control algorithm simulates and evaluates the optimized control instructions based on the current system state and the predicted future state, and calculates various indicators during the execution of the instructions.

[0120] Based on the evaluation results of the model predictive control algorithm, it is judged whether the optimized control instructions meet the system requirements. If they do, they are output as the final control instructions. If they do not meet the requirements, the optimized control instructions are adjusted and simulated and evaluated again until the final control instructions that meet the requirements are obtained.

[0121] In this embodiment, based on the optimized control instructions, a dynamic graph structure of time-series relationships is constructed to transform the physical connections, information transmission, and time series characteristics between the various components of the battery power system into a graph theory model, intuitively presenting the dynamic evolution of the system. Next, the linkage constraints between different units are analyzed (such as the mutual influence between battery charging and discharging and temperature, power distribution, and driving strategy), converted into mathematical constraints, and integrated into the dynamic graph structure to form a constraint model that is closer to physical reality. This constraint model is then input into the model predictive control algorithm. Through rolling optimization and multi-step prediction, the execution effect of the control instructions in the future time domain is simulated, and the impact of the instructions on system performance indicators is quantitatively evaluated. Finally, the control instructions are iteratively adjusted based on the evaluation results until the multi-dimensional requirements of system safety and efficiency are met, and the final control instructions are output. In summary, this embodiment captures the spatiotemporal characteristics of the system through a time-series dynamic graph, utilizes cross-unit linkage constraints to ensure the physical feasibility of the instructions, and combines the rolling optimization capabilities of model predictive control to effectively solve the problem that traditional verification methods are difficult to handle complex system coupling relationships and dynamic changes, significantly improving the safety, stability, and robustness of the control instructions, and achieving global optimal control of the battery power system under multiple constraints.

[0122] In this embodiment, based on the temporal relationship dynamic graph structure and cross-unit linkage constraints, the model predictive control algorithm is used to verify the optimized control instructions and output the final control instructions. The specific workflow is as follows:

[0123] Step 1: Based on the optimized control instructions, we start to build a dynamic graph structure of the timing relationship of the battery power system. The core components such as battery modules, charge and discharge controllers, drive motors, and thermal management systems are set as nodes in the graph structure. Each node is given rich attributes to represent its operating status. For example, the battery module node attribute can be represented as Node bat =[V bat ,I bat ,T bat ,SOC bat ], where V bat Represents the battery voltage, which is acquired in real time by the voltage sensor; I bat Represents the battery charge and discharge current, measured by the current sensor; T bat is the battery temperature, monitored by the temperature sensor; SOC bat is the battery state of charge, estimated using the existing algorithm of ampere-hour integration combined with Kalman filtering.

[0124] The physical connection lines and information transmission channels between components are defined as edges in the graph. Taking the battery module and the charge and discharge controller as an example, the energy transmission line between them constitutes an edge, and the attributes of the edge include the transmission power P ij(represents the power transmitted from node i to node j, obtained by combining the voltage and current data with the power calculation formula) and the transmission delay τ ij (The time required for information to travel from node i to node j is determined by factors such as line length and transmission rate).

[0125] According to the time series information, the state changes of each node and the attribute updates of the edges at different times are connected in series with a time interval of \(0.5\) seconds, so as to construct the initial time series relationship dynamic graph G init =(N, E, T). N is the node set, which includes all key component nodes of the system; E is the edge set, which covers the connection and information transmission relationship between components; T is the time series set, which records the status data at each moment.

[0126] Step 2: In-depth analysis of the interaction between each unit in the battery power system, converting it into specific constraints and integrating it into the initial dynamic diagram.

[0127] In terms of energy, following the law of conservation of energy, an energy balance constraint is established: the battery output power P bat-out Must be equal to the power consumed by the drive motor P motor , thermal management system operating power P thermal and line power loss P loss The sum of P bat-out =P motor +P thermal +P loss Among them, P motor Calculated based on the motor's operating characteristic curve and input voltage and current; P thermal Determined by the working mode and operating parameters of the thermal management system; P loss Estimated by the square relationship between line resistance and current.

[0128] At the temperature level, considering the principles of heat conduction and heat exchange, the thermal balance constraint is determined. bat Exceeds the preset threshold T thresh When the thermal management system needs to start the cooling mode, and use the corresponding power P cool The temperature is lowered and a certain heat exchange rate relationship is met to ensure that the system temperature is within a safe range.

[0129] These constraints are added to the initial temporal relationship dynamic graph structure to restrict and adjust the properties of nodes and edges in the graph. If the state data at a certain moment does not meet the constraints, such as the calculated P bat-out If it is not equal to the right side of the equation, the state is considered invalid. In this way, a dynamic graph G with cross-unit linkage constraints is formed. constraint=(N, E, T, C), where C is the set of constraints.

[0130] Step 3: The dynamic graph of the time series relationship with cross-unit linkage constraints is used as the input of the improved model predictive control (MPC) algorithm. This improved MPC algorithm introduces a hierarchical optimization strategy based on the traditional framework.

[0131] The algorithm firstly calculates the current system state S cur , that is, the real-time operating parameters of each node, and the prediction model of the next M time steps, simulate the execution of the optimized control instructions. The prediction model is constructed using an improved linear regression algorithm to predict the battery state of charge change ΔSOC pred For example, the formula is: ΔSOC pred =α1×I pred +α2×T pred +α3×P out-pred +β. Among them, I pred is the predicted charge and discharge current, estimated by analyzing historical current data and current control instruction trends; T pred is the predicted battery temperature, combined with the ambient temperature, heat dissipation and the working status of the thermal management system; P out-pred is the predicted battery output power, calculated based on the control instructions and system operation model; α1, α2, α3 and β are coefficients obtained through training with a large amount of historical data. The training process is optimized using the least squares method to reduce the prediction error.

[0132] During the simulation, the algorithm synchronously calculates various key indicators during the instruction execution process. Energy efficiency η energy The calculation formula is Among them, P useful is the effective output power of the system, such as the power of the drive motor used for vehicle driving; P total is the total input power of the system, that is, the power provided by the battery. System operation stability σ stable It is measured by calculating the standard deviation of key parameters (such as voltage, current, temperature, etc.) within a certain period of time. The smaller the standard deviation, the higher the stability. bat Based on factors such as the battery's charge and discharge depth, temperature, and number of charge and discharge cycles, a comprehensive assessment is conducted using empirical formulas combined with real-time data.

[0133] Step 4: Compare and judge the various indicators calculated by the model predictive control algorithm with the pre-set system requirement thresholds. Set the energy utilization efficiency η energy Need to be no less than 88%, system operation stability σ stable The fluctuation range does not exceed ±3%, and the battery health loss is L bat No more than (0.08% and other specific threshold conditions in a single run.

[0134] If all indicators meet the requirements, it means that the optimized control instruction meets the system operation requirements and is directly output as the final control instruction. If any indicator does not meet the requirements, the instruction adjustment mechanism is activated. The improved adaptive step gradient descent method is used to adjust the key parameters in the control instruction. For parameters that affect key indicators, such as the power allocation ratio r power , temperature adjustment threshold T set etc., according to The formula is adjusted. Among them, Parameter new and Parameter old are the parameters after adjustment and before adjustment respectively; γ is the adaptive learning rate, which is dynamically adjusted according to the historical effect of parameter adjustment and the current gradient size; is the gradient of the objective function F (constructed by integrating various indicators) with respect to this parameter. The adjusted control instructions are then fed back into the timing dynamic graph with cross-unit linkage constraints and the model predictive control algorithm for simulation and evaluation. This process is repeated until all indicators meet the system requirements, ultimately outputting the final control instructions that meet the requirements.

[0135] It should be noted that the details of the relevant algorithm technologies that are not described in detail above can be referred to the existing relevant algorithm technologies and will not be repeated here.

[0136] See also Figure 2 , is a schematic diagram of a battery power system control system based on a digital twin system drive according to an embodiment of the present invention. The battery power system control system based on a digital twin system drive includes:

[0137] An acquisition module 10 is used to obtain operating status data of the battery power system in real time based on the digital twin model;

[0138] The generation module 11 is used to input the acquired operating status data into a two-layer guidance mechanism consisting of proactive decision-making control of the twin model and dynamic collaborative correction of real-time data to generate preliminary control instructions;

[0139] An optimization module 12 is configured to optimize the preliminary control instructions through a multi-agent collaborative decision-making mechanism to obtain optimized control instructions. The multi-agent collaborative decision-making mechanism includes: an energy management agent optimizing power distribution, a charging scheduling agent optimizing braking energy recovery, and a vehicle control agent optimizing driving strategies.

[0140] The verification module 13 is used to verify the optimized control instructions based on the time sequence relationship dynamic graph structure and cross-unit linkage constraints using the model predictive control algorithm and output the final control instructions.

[0141] Compared with the prior art, the embodiments of the present invention have the following beneficial effects:

[0142] First, a digital twin model collects real-time operating status data from all levels of the battery powertrain (from single cells to the entire pack, controller, and drive motor). Real-time mapping between the virtual model and the physical system ensures the comprehensiveness and accuracy of the acquired data. This data is then fed into a two-tiered guidance mechanism consisting of proactive decision-making and control based on the twin model and dynamic collaborative correction based on real-time data. Proactive decision-making predicts system status based on historical data and simulations, while real-time data correction adjusts model parameters based on actual operational deviations. These two mechanisms combine to generate preliminary control instructions, balancing the foresight and real-time nature of the predictions. Subsequently, a multi-agent collaborative decision-making mechanism is employed to decompose these preliminary instructions into subtasks, such as power allocation, regenerative braking, and driving strategy. Each agent independently calculates and collaboratively adjusts based on its own optimization objective (e.g., the energy management agent optimizes power allocation, and the charging scheduling agent optimizes regenerative braking), achieving a balance between local optimization and global optimization. Finally, a dynamic graph structure based on temporal relationships intuitively presents the spatiotemporal relationships between system components. Physical constraints are set by combining cross-unit linkage constraints. The optimized control instructions are then subjected to rolling optimization and feasibility verification using a model predictive control algorithm, ultimately outputting final control instructions that align with the system's actual operation. Compared with traditional control methods that have problems such as insufficient multi-physical field coupling, real-time decision-making lag, and poor subsystem coordination, the embodiments of the present invention use digital twin models to achieve accurate perception of system status, a two-layer guidance mechanism to enhance the dynamic adaptability of decision-making, multi-agent collaborative optimization of system resource allocation, and timing diagrams and model predictive control to ensure the feasibility of instructions, thereby achieving full-process and multi-dimensional precise control of the battery power system, significantly improving system operating efficiency, energy utilization and adaptability to working conditions.

[0143] As one example, the acquisition module is specifically configured to:

[0144] Utilize sensors deployed at key locations within the battery power system to collect basic data on voltage, current, temperature, and battery state of charge, generating a raw data set containing a variety of basic information.

[0145] The original data set is denoised and then normalized to a specific numerical range to obtain preprocessed data;

[0146] The preprocessed data is input into the constructed digital twin model. The digital twin model calculates and matches the input data with its own model parameters, and outputs operating status data corresponding to the actual operating status of the battery power system.

[0147] As one example, the generating module is specifically configured to:

[0148] Input the operating status data into the twin model. The twin model predicts the operating status of the battery power system in the future based on the preset algorithm and historical data, and obtains the predicted operating status results.

[0149] Based on the predicted operating state results, combined with the performance indicators and safety constraints of the battery power system, and using pre-set decision rules, multiple candidate proactive decision plans are generated to form a proactive decision plan set;

[0150] Compare and analyze the operating status data with the results predicted by the twin model, calculate the deviation between the two, and adjust some parameters of the twin model based on the deviation to obtain the corrected twin model;

[0151] The corrected twin model and the set of proactive decision-making schemes are comprehensively evaluated, and the most appropriate proactive decision-making scheme is selected based on the evaluation results and output as the preliminary control instruction.

[0152] As one example, the optimization module is specifically configured to:

[0153] The preliminary control instructions are passed to the energy management agent, the charging scheduling agent, and the vehicle control agent. Each agent parses the preliminary control instructions according to its own task and obtains the instruction content related to its own task;

[0154] The energy management agent adjusts the power allocation instructions based on the battery's current state of charge, remaining power, battery health status, and vehicle power demand information, calculates a more reasonable power allocation plan, and obtains the optimized power allocation instructions;

[0155] The charging scheduling agent optimizes the braking energy recovery instructions based on the vehicle's driving status, battery charging capacity, and historical braking energy recovery data, determines the best braking energy recovery strategy, and obtains the optimized braking energy recovery instructions;

[0156] The vehicle control agent adjusts the driving strategy instructions based on road conditions, traffic rules, and vehicle dynamic characteristics, plans a more reasonable driving strategy, and obtains the optimized driving strategy instructions;

[0157] The optimized power distribution instructions, brake energy recovery instructions and driving strategy instructions are integrated to form optimized control instructions.

[0158] As one example, the verification module is specifically used to:

[0159] Based on the optimized control instructions, combined with the connection relationships, information transmission relationships, and time series information between the various components of the battery power system, a time series dynamic graph structure is constructed. The meaning and attributes of each node and edge in the graph structure are clarified to obtain the initial time series dynamic graph.

[0160] Analyze the mutual influence and constraints between different units in the battery power system, convert these relationships into specific constraints, and add them to the initial timing relationship dynamic graph structure to form a timing relationship dynamic graph with cross-unit linkage constraints;

[0161] The dynamic graph of the time series relationship with cross-unit linkage constraints is used as the input of the model predictive control algorithm. The model predictive control algorithm simulates and evaluates the optimized control instructions based on the current system state and the predicted future state, and calculates various indicators during the execution of the instructions.

[0162] Based on the evaluation results of the model predictive control algorithm, it is judged whether the optimized control instructions meet the system requirements. If they do, they are output as the final control instructions. If they do not meet the requirements, the optimized control instructions are adjusted and simulated and evaluated again until the final control instructions that meet the requirements are obtained.

[0163] It should be noted that the system embodiment described above is merely illustrative, wherein the units described as separate components may or may not be physically separate, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed across multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the present embodiment. In addition, in the drawings of the system embodiment provided by the present invention, the connection relationship between the modules indicates that there is a communication connection between them, which can be specifically implemented as one or more communication buses or signal lines. A person of ordinary skill in the art can understand and implement the present invention without inventive work.

[0164] The above is a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present invention. These improvements and modifications are also considered to be within the scope of protection of the present invention.

Claims

1. A battery power system control method based on digital twin system drive, characterized in that: include: Obtain real-time operating status data of the battery power system based on the digital twin model; The acquired operating status data is input into a two-layer guidance mechanism consisting of proactive decision-making and control by the twin model and dynamic collaborative correction of real-time data to generate preliminary control instructions; The initial control instructions are optimized through a multi-agent collaborative decision-making mechanism to obtain optimized control instructions. The multi-agent collaborative decision-making mechanism includes: the energy management agent optimizes power distribution, the charging scheduling agent optimizes braking energy recovery, and the vehicle control agent optimizes driving strategy. Based on the dynamic graph structure of timing relationships and cross-unit linkage constraints, the model predictive control algorithm is used to verify the optimized control instructions and output the final control instructions.

2. The battery power system control method based on digital twin system drive according to claim 1, characterized in that: The real-time acquisition of the operating status data of the battery power system based on the digital twin model includes the following sub-steps: Utilize sensors deployed at key locations within the battery power system to collect basic data on voltage, current, temperature, and battery state of charge, generating a raw data set containing a variety of basic information. The original data set is denoised and then normalized to a specific numerical range to obtain preprocessed data; The preprocessed data is input into the constructed digital twin model. The digital twin model calculates and matches the input data with its own model parameters, and outputs operating status data corresponding to the actual operating status of the battery power system.

3. The battery power system control method based on digital twin system drive according to claim 1, characterized in that: The acquired operating status data is input into a two-layer guidance mechanism consisting of proactive decision-making and control of the twin model and dynamic collaborative correction of real-time data to generate preliminary control instructions, including the following sub-steps: Input the operating status data into the twin model. The twin model predicts the operating status of the battery power system in the future based on the preset algorithm and historical data, and obtains the predicted operating status results. Based on the predicted operating state results, combined with the performance indicators and safety constraints of the battery power system, and using pre-set decision rules, multiple candidate proactive decision plans are generated to form a proactive decision plan set; Compare and analyze the operating status data with the results predicted by the twin model, calculate the deviation between the two, and adjust some parameters of the twin model based on the deviation to obtain the corrected twin model; The corrected twin model and the set of proactive decision-making schemes are comprehensively evaluated, and the most appropriate proactive decision-making scheme is selected based on the evaluation results and output as the preliminary control instruction.

4. The battery power system control method based on digital twin system drive according to claim 1, characterized in that: The multi-agent collaborative decision-making mechanism optimizes the preliminary control instructions to obtain optimized control instructions. The multi-agent collaborative decision-making mechanism includes: the energy management agent optimizes power distribution, the charging scheduling agent optimizes braking energy recovery, and the vehicle control agent optimizes driving strategy, including the following sub-steps: The preliminary control instructions are passed to the energy management agent, the charging scheduling agent, and the vehicle control agent. Each agent parses the preliminary control instructions according to its own task and obtains the instruction content related to its own task; The energy management agent adjusts the power allocation instructions based on the battery's current state of charge, remaining power, battery health status, and vehicle power demand information, calculates a more reasonable power allocation plan, and obtains the optimized power allocation instructions; The charging scheduling agent optimizes the braking energy recovery instructions based on the vehicle's driving status, battery charging capacity, and historical braking energy recovery data, determines the best braking energy recovery strategy, and obtains the optimized braking energy recovery instructions; The vehicle control agent adjusts the driving strategy instructions based on road conditions, traffic rules, and vehicle dynamic characteristics, plans a more reasonable driving strategy, and obtains the optimized driving strategy instructions; The optimized power distribution instructions, brake energy recovery instructions and driving strategy instructions are integrated to form optimized control instructions.

5. The battery power system control method based on digital twin system drive according to claim 1, characterized in that: The method of verifying the optimized control instructions using a model predictive control algorithm based on the time sequence relationship dynamic graph structure and cross-unit linkage constraints and outputting the final control instructions includes the following sub-steps: Based on the optimized control instructions, combined with the connection relationships, information transmission relationships, and time series information between the various components of the battery power system, a time series dynamic graph structure is constructed. The meaning and attributes of each node and edge in the graph structure are clarified to obtain the initial time series dynamic graph. Analyze the mutual influence and constraints between different units in the battery power system, convert these relationships into specific constraints, and add them to the initial timing relationship dynamic graph structure to form a timing relationship dynamic graph with cross-unit linkage constraints; The dynamic graph of the time series relationship with cross-unit linkage constraints is used as the input of the model predictive control algorithm. The model predictive control algorithm simulates and evaluates the optimized control instructions based on the current system state and the predicted future state, and calculates various indicators during the execution of the instructions. Based on the evaluation results of the model predictive control algorithm, it is determined whether the optimized control instructions meet the system requirements. If they do, they are output as the final control instructions. If the requirements are not met, the optimized control instructions are adjusted and simulated and evaluated again until the final control instructions that meet the requirements are obtained.

6. A battery power system control system based on digital twin system drive, characterized in that: include: An acquisition module is used to obtain the operating status data of the battery power system in real time based on the digital twin model; A generation module is used to input the acquired operating status data into a two-layer guidance mechanism consisting of proactive decision-making and control based on the twin model and dynamic collaborative correction based on real-time data to generate preliminary control instructions; An optimization module is used to optimize the initial control instructions through a multi-agent collaborative decision-making mechanism to obtain optimized control instructions. The multi-agent collaborative decision-making mechanism includes: the energy management agent optimizes power distribution, the charging scheduling agent optimizes braking energy recovery, and the vehicle control agent optimizes driving strategy; The verification module is used to verify the optimized control instructions based on the dynamic graph structure of the timing relationship and the cross-unit linkage constraints, and output the final control instructions using the model predictive control algorithm.

7. The battery power system control system based on digital twin system drive according to claim 6, characterized in that: The acquisition module is specifically used for: Utilize sensors deployed at key locations within the battery power system to collect basic data on voltage, current, temperature, and battery state of charge, generating a raw data set containing a variety of basic information. The original data set is denoised and then normalized to a specific numerical range to obtain preprocessed data; The preprocessed data is input into the constructed digital twin model. The digital twin model calculates and matches the input data with its own model parameters, and outputs operating status data corresponding to the actual operating status of the battery power system.

8. The battery power system control system based on digital twin system drive according to claim 6, characterized in that: The generation module is specifically used for: Input the operating status data into the twin model. The twin model predicts the operating status of the battery power system in the future based on the preset algorithm and historical data, and obtains the predicted operating status results. Based on the predicted operating state results, combined with the performance indicators and safety constraints of the battery power system, and using pre-set decision rules, multiple candidate proactive decision plans are generated to form a proactive decision plan set; Compare and analyze the operating status data with the results predicted by the twin model, calculate the deviation between the two, and adjust some parameters of the twin model based on the deviation to obtain the corrected twin model; The corrected twin model and the set of proactive decision-making schemes are comprehensively evaluated, and the most appropriate proactive decision-making scheme is selected based on the evaluation results and output as the preliminary control instruction.

9. The battery power system control system based on digital twin system drive according to claim 6, characterized in that: The optimization module is specifically used for: The preliminary control instructions are passed to the energy management agent, the charging scheduling agent, and the vehicle control agent. Each agent parses the preliminary control instructions according to its own task and obtains the instruction content related to its own task; The energy management agent adjusts the power allocation instructions based on the battery's current state of charge, remaining power, battery health status, and vehicle power demand information, calculates a more reasonable power allocation plan, and obtains the optimized power allocation instructions; The charging scheduling agent optimizes the braking energy recovery instructions based on the vehicle's driving status, battery charging capacity, and historical braking energy recovery data, determines the best braking energy recovery strategy, and obtains the optimized braking energy recovery instructions; The vehicle control agent adjusts the driving strategy instructions based on road conditions, traffic rules, and vehicle dynamic characteristics, plans a more reasonable driving strategy, and obtains the optimized driving strategy instructions; The optimized power distribution instructions, brake energy recovery instructions and driving strategy instructions are integrated to form optimized control instructions.

10. The battery power system control system based on digital twin system drive according to claim 6, characterized in that: The verification module is specifically used for: Based on the optimized control instructions, combined with the connection relationships, information transmission relationships, and time series information between the various components of the battery power system, a time series dynamic graph structure is constructed. The meaning and attributes of each node and edge in the graph structure are clarified to obtain the initial time series dynamic graph. Analyze the mutual influence and constraints between different units in the battery power system, convert these relationships into specific constraints, and add them to the initial timing relationship dynamic graph structure to form a timing relationship dynamic graph with cross-unit linkage constraints; The dynamic graph of the time series relationship with cross-unit linkage constraints is used as the input of the model predictive control algorithm. The model predictive control algorithm simulates and evaluates the optimized control instructions based on the current system state and the predicted future state, and calculates various indicators during the execution of the instructions. Based on the evaluation results of the model predictive control algorithm, it is determined whether the optimized control instructions meet the system requirements. If they do, they are output as the final control instructions. If the requirements are not met, the optimized control instructions are adjusted and simulated and evaluated again until the final control instructions that meet the requirements are obtained.

Images