A battery stack pressure adaptive control system and method

Abstract

The application discloses a battery stack pressure self-adaptive control system and method, and relates to the technical field of battery systems.The system comprises a scheduling layer, a decision layer and an execution perception layer.The scheduling layer is used for determining a control target according to business and safety requirements, generating a reference band, a safety boundary and a pressure-current-temperature linkage control strategy, and compiling the pressure-current-temperature linkage control strategy into instruction sets corresponding to different devices.The decision layer is used for completing state estimation, critical stack pressure identification, pressure-current-temperature collaborative optimization and scheduling management by combining an expert system and a model predictive control technology, and obtaining final pressure, current and temperature control signals under the constraints of the reference band and the safety boundary, and the final pressure, current and temperature control signals are sent to the execution perception layer.The application can link pressure, current and temperature, and realize integrated management and control of battery safety and performance.

Description

Technical Field

[0001] This application relates to the field of battery system technology, and in particular to a battery stacking pressure adaptive control system and method. Background Technology

[0002] With the rapid development of new power systems, energy storage power stations, and distributed energy consumption, battery systems that balance high safety, long lifespan, and low cost have become critical infrastructure. Stack pressure (SP) permeates the entire lifecycle of battery cell manufacturing and operation: during assembly and formation, pressure determines the actual contact area and pore structure between electrodes / separators / electrolytes or solid-state electrolytes (SSE); during use, pressure profoundly alters the probability of void formation, dendrite deflection, and side reactions by affecting mechatronic coupling processes such as interface contact, ion transport, and metal flow (creep). Therefore, the setting of stack pressure is directly related to the alarm and protection thresholds of the Battery Management System (BMS), the coordination of fast charging strategies and thermal management, module / cluster consistency, and assembly efficiency, thereby affecting asset life assessment and total lifecycle maintenance costs. For energy storage scenarios, the ability to dynamically and traceably maintain an "appropriate pressure state" at different rates, temperatures, and aging stages is a key means to ensure long-term stable operation of the system, reduce the risk of thermal runaway and early failure, and improve the availability and reliability of unit energy output.

[0003] Existing methods for setting and controlling stacking pressure are mostly based on empirical solidification of existing material systems and fixed process windows. For example, assembly lines for traditional graphite / ternary systems often provide a relatively constant pre-compression range, supplemented by quality monitoring driven by the process capability index (Cpk). Other solutions treat pressure as an auxiliary independent variable, along with roll thickness, solvent content, and drying curve, and establish lookup and regression relationships through offline DOE (Design of Experiments) for subsequent batch reuse.

[0004] However, with the rapid iteration of new material systems (such as high-silicon anodes, lithium metal anodes, solid electrolytes, and nickel-rich / high-voltage cathodes) and more demanding operating boundaries (low temperature, fast charging, deep cycling), the "optimal window" for stacking pressure exhibits strong non-stationarity and cross-stage drift characteristics: the preference for pressure may be opposite during the lithium stripping and lithium plating stages, and the critical stacking pressure (CSP) and overpotential-pressure curves at different temperatures / rates also show systematic shifts; at the same time, even slight inhomogeneities in the in-plane pressure distribution are enough to induce local contact degradation and evolve into voids and dendrites. Relying solely on fixed pressure or static table lookup is insufficient to cope with this coupled complexity that evolves over time and under different operating conditions. Summary of the Invention

[0005] The purpose of this application is to provide a battery stack pressure adaptive control system and method that can link pressure, current and temperature to achieve integrated control of battery safety and performance.

[0006] To achieve the above objectives, this application provides the following solution.

[0007] In a first aspect, this application provides a battery stacking pressure adaptive control system, including a orchestration layer, a decision layer, and an execution perception layer; The orchestration layer is used to: determine control objectives based on business and security requirements; generate reference bands, security boundaries, and pressure-current-temperature linkage control strategies based on the control objectives; compile the pressure-current-temperature linkage control strategies into instruction sets corresponding to different devices; and distribute the reference bands, security boundaries, and instruction sets to the decision layer. The decision layer is used to: receive multi-source real-time signals uploaded by the execution perception layer; under the constraints of the reference band and the safety boundary, use the multi-source real-time signals to complete state estimation, critical stacking pressure identification, and pressure-current-temperature collaborative optimization and scheduling management combining expert system and model predictive control technology to achieve adaptive coordinated control under different rates, temperatures and aging stages, so as to obtain the final pressure, current and temperature control signals; and send the final pressure, current and temperature control signals to the execution perception layer. The execution sensing layer is used to: execute the final pressure, current and temperature control signals, and collect multi-source real-time signals.

[0008] Secondly, this application provides a battery stacking pressure adaptive control method, applied to a battery stacking pressure adaptive control system, the method comprising: The orchestration layer determines control objectives based on business and security requirements; based on the control objectives, a reference band, a safety boundary, and a pressure-current-temperature linkage control strategy are generated; the pressure-current-temperature linkage control strategy is compiled into instruction sets corresponding to different devices; and the reference band, the safety boundary, and the instruction sets are distributed to the decision layer. The system collects real-time signals from multiple sources through the perception layer and uploads them to the decision layer. Under the constraints of the reference band and the safety boundary, the decision layer utilizes multi-source real-time signals to complete state estimation, critical stacking pressure identification, and pressure-current-temperature collaborative optimization and scheduling management by combining expert systems and model predictive control technology. This enables adaptive coordinated control under different rates, temperatures, and aging stages to obtain the final pressure, current, and temperature control signals. The final pressure, current, and temperature control signals are then sent down to the execution perception layer. The final pressure, current, and temperature control signals are executed through the execution sensing layer, and multi-source real-time signals are continuously acquired.

[0009] According to the specific embodiments provided in this application, the following technical effects are disclosed: This application completes the conversion of business and security requirements into structured control objectives, the generation and maintenance of reference bands and security boundaries, and performs compilation and conversion through the orchestration layer. Under the constraints of reference bands and security boundaries issued by the orchestration layer, the decision layer uses multi-source real-time signals to complete state estimation, critical stacking pressure identification, and pressure-current-temperature collaborative optimization. Through state estimation and CSP identification, hybrid intelligent control (ES+MPC), and collaborative scheduling and fault diagnosis, the decision layer performs rolling optimization and graded rollback of pressure, charging and discharging current, and temperature while ensuring the rigid constraints of the security boundaries, realizing adaptive coordinated control under different rates, temperatures, and aging stages. Finally, the execution perception layer transforms the upper-layer decisions into precise physical execution and sends back multi-source real-time signals, providing a real and traceable data foundation for upper-layer strategy optimization and large language model / physical model updates. Through data interaction between the above three layers, this application realizes an integrated closed loop from knowledge to control, and has adaptive capabilities across material systems and multiple operating conditions. Attached Figure Description

[0010] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0011] Figure 1 This is a schematic diagram of the battery stacking pressure adaptive control system in one embodiment of this application.

[0012] Figure 2 This is a partial judgment diagram of the battery stacking pressure adaptive control method in one embodiment of this application. Detailed Implementation

[0013] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0014] This application provides a battery stack pressure adaptive control system and method. Under the expected functional safety framework of BMS, it relies on multi-source online signals to lock the CSP in real time and link pressure, current and temperature to achieve integrated control of battery safety and performance.

[0015] To make the above-mentioned objectives, features and advantages of this application more apparent and understandable, the application will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0016] In one exemplary embodiment, such as Figure 1 As shown, a battery stacking pressure adaptive control system is provided. It is a multi-level intelligent system architecture based on the co-strategy of a large language model and a physical model, including an orchestration layer, a decision layer and an execution perception layer.

[0017] Each layer interacts with data via industrial Ethernet, CAN bus, or other wired / wireless communication links, forming a closed-loop measurement and control system from data acquisition, state estimation, policy calculation to physical execution and policy feedback. The system as a whole uses a large language model as the core knowledge engine and a physical model as the constraint kernel to achieve coordinated optimization control of stacked pressure, current, and temperature under safety boundary constraints, and supports adaptive migration across material systems, rate ranges, and temperature conditions.

[0018] (a) Arrangement layer.

[0019] The orchestration layer, centered on a large language model, is responsible for transforming business and security requirements into structured objectives and executable control schemes. Specifically, the orchestration layer is used to: determine control objectives based on business and security requirements; generate reference bands, security boundaries, and pressure-current-temperature linkage control strategies based on the control objectives; compile the pressure-current-temperature linkage control strategies into instruction sets corresponding to different devices; and distribute the reference bands, security boundaries, and instruction sets to the decision layer.

[0020] The orchestration layer is deployed on servers or edge computing nodes with relatively high computing power. It is the knowledge and strategy center for the co-governance of large language models and physical models. The orchestration layer includes a target parsing and evidence retrieval module, a rule and security library module, and an instruction compilation and auditing module. Through these three modules, reference bands and security boundaries are extracted from literature, historical experiments and standard operating procedures to generate "pressure-current-temperature" linkage control strategies and test plans. These are then uniformly compiled into cross-device executable and auditable instruction scripts, providing a consistent knowledge and strategy entry point for lower-level decision-making and execution.

[0021] (1) The target parsing and evidence retrieval module is used to: call and use a preset large language model to extract control targets from business and security requirements, and perform structured parsing of the control targets to convert them into performance indicators and constraints of various devices in the battery system; thus, this application transforms the control targets described in natural language (i.e., business and security requirements) into specific, quantifiable physical parameters or operational constraints, i.e., performance indicators and constraints, such as target coulombic efficiency range, temperature upper limit, and allowable stacking pressure range, after semantic understanding and structured extraction by the large language model. Then, based on the performance indicators and constraints, evidence is retrieved from the preset knowledge base to generate a reference band RLLM (Reinforced Learning Logic Model) to provide evidence required for decision support of the system. The preset database includes relevant literature, historical experimental data, and standard operating procedures. The reference band is a reference value, which is a different reference value generated by the trained large language model according to different inputs (such as battery type, temperature, charge / discharge rate, etc.).

[0022] In short, the target parsing and evidence retrieval module integrates a large language model fine-tuned with knowledge of the field, used for large language model construction and training, target parsing, evidence retrieval, and reference strip generation. The specific internal settings of the target parsing and evidence retrieval module are as follows: (11) Construction and training of pre-set large language models.

[0023] 1) In the initial stage, domain preprocessing is performed on the general large language model, and papers, standards, patents, production process cards, experimental records, SOP documents and other materials related to battery stacking pressure, solid / semi-solid system, low temperature fast charging, electrode design, BMS protection, etc. are constructed into a domain corpus.

[0024] 2) By employing supervised fine-tuning and instruction fine-tuning, the large language model can understand concepts such as critical stacking pressure, reference band, safety boundary, and pressure-current-temperature linkage, and can output structured control constraints and candidate strategies.

[0025] 3) During online operation, the runtime data and anomaly handling results returned by the instruction compilation and auditing module are periodically used to incrementally update or retrain the large language model, thereby improving the accuracy and adaptability of reference bands and policy generation.

[0026] (12) Target analysis.

[0027] 1) The target parsing and evidence retrieval module receives business objectives and security requirements from the upper-level monitoring system or operation and maintenance personnel, such as maintaining the critical stacking pressure platform area at a certain rate / temperature / cycle stage, and achieving maximum available energy output within the safety boundary.

[0028] 2) Semantic understanding and constraint extraction are performed on natural language and structured input through a pre-set large language model, and the results are transformed into formal control objectives, performance indicators and constraints, including: target coulomb efficiency range, allowable stacking pressure range, upper temperature limit, current change rate constraint, allowable perturbation amplitude, etc.

[0029] (13) Evidence retrieval and reference band generation.

[0030] 1) Based on the analyzed control objective, that is, the performance indicators and constraints obtained in the previous objective analysis, retrieve literature, experimental data and historical operation data that match the current material system, rate, temperature and cycle stage from the local and / or cloud knowledge base.

[0031] 2) Based on the reasoning of the large language model, and by integrating the prior knowledge of the physical model, a reference band that matches the current working conditions is generated.

[0032] 3) Send the above reference tapes in a structured manner to the rules and security library module and the decision layer for subsequent control and judgment.

[0033] (2) The rules and security library module is used to store and maintain the security boundaries and pressure-current-temperature linkage control strategies of various devices, and integrate emergency response SOPs (Standard Operating Procedures); and to perform version management and permission management for each version of the security boundaries, pressure-current-temperature linkage control strategies and emergency response SOPs.

[0034] Specifically, the rules and security library module stores and maintains safety thresholds and operating condition rules (such as pressure, temperature, and current) for various devices, and provides real-time querying and decision support. It also integrates emergency response standard operating procedures (SOPs) to handle system anomalies or failures. Furthermore, version control ensures the traceability and version control of these rules and security standards. In short, the rules and security library module is used to uniformly store and maintain the safety boundaries and operating condition rules required for system operation, serving as a rigid constraint source for decision-making in control and fault handling. The specific internal settings of the rules and security library module are as follows: (21) Safety boundary management unit: Stores safety boundaries such as stacking pressure upper and lower limits, temperature upper and lower limits, current upper limit, overpotential threshold, acoustic emission and impedance abnormality threshold corresponding to different battery chemical systems, module structures, stacking methods and cooling structures; and supports indexing safety boundaries according to the dimension of "material system - operating condition - life stage" to provide the decision-making layer with safety boundary configuration matching the current operating condition.

[0035] (22) Rules and SOPs Management Unit: Stores the rules corresponding to the pressure-current-temperature linkage control strategy, including anomaly / fault identification rules, degraded operation rules, rollback strategy rules, etc.; stores emergency response SOPs, which are used to guide the decision-making level to perform operations such as disconnecting faulty branches, degrading operation, strengthening cooling, data preservation and alarm level classification according to the pre-approved steps when abnormal conditions are triggered.

[0036] (23) Version governance and traceability unit: Perform version management and permission management for each version of the safety boundary, pressure-current-temperature linkage control strategy and emergency response SOP, record the creation time, effective time, modifier and applicable scope; then provide the current effective version information to the instruction compilation and auditing module for subsequent responsibility traceability and strategy evolution analysis.

[0037] The security boundary and system operation strategy rules can be updated by the target parsing and evidence retrieval module based on the reasoning results of the large language model. After manual review, they are written into the rules and security library module to realize the closed-loop evolution of "data-model-rules".

[0038] (3) The instruction compilation and auditing module is used to: receive the reference band, the safety boundary and the pressure-current-temperature linkage control strategy; compile the pressure-current-temperature linkage control strategy into instruction sets corresponding to different devices; track and record the execution process of each instruction in the instruction set in real time; and send the reference band, the safety boundary and the instruction set to the decision layer.

[0039] Specifically, the instruction compilation and auditing module compiles the generated strategies and rules into instruction sets suitable for different devices (such as API calls, PLC instructions, etc.). Simultaneously, it is responsible for recording the instruction execution process in real time, including timestamps and hash chains of instructions, events, and data, ensuring the traceability and verifiability of the execution process. Based on this, instruction auditing and error backtracking are performed to guarantee operational transparency and accountability.

[0040] The instruction compilation and auditing module can be located between the orchestration layer and the decision / execution awareness layer. It is used to translate the control strategies generated by the upper layer into executable instructions for the lower-level devices and to audit and record the control process. The internal settings of the instruction compilation and auditing module are as follows: (31) Instruction compilation unit.

[0041] 1) Receive the reference band and security boundary jointly determined by the target parsing and evidence retrieval module and the rule and security library module, as well as the pressure-current-temperature linkage control strategy output by the decision layer.

[0042] 2) Based on the interface specifications of different devices in the execution perception layer, compile the pressure-current-temperature linkage control strategy into executable instruction sets such as API calls, PLC logic instructions, and motion control commands.

[0043] 3) Structure and encapsulate instructions from the same batch or the same type of operating conditions to form reusable executable process cards and control scripts.

[0044] (32) Process audit and traceability unit: records the generation time, issuance time, target object, key parameters and execution results of each instruction; timestamps and hash chains are used to record key events (including security boundary touch, rollback action trigger, cell fault determination, etc.) to achieve anti-tampering and provable security of the control process; periodically transmits audit data, observed stack pressure platform area location, abnormal recovery time and other indicators back to the target parsing and evidence retrieval module for online updates of the large language model and rule base.

[0045] (ii) Decision-making level.

[0046] The decision layer is used to: receive multi-source real-time signals uploaded by the execution perception layer; under the constraints of the reference band and the safety boundary, use the multi-source real-time signals to complete state estimation, critical stacking pressure identification, pressure-current-temperature collaborative optimization and scheduling management combining expert system and model predictive control technology, and realize adaptive coordinated control under different rates, temperatures and aging stages to obtain the final pressure, current and temperature control signals; and send the final pressure, current and temperature control signals to the execution perception layer.

[0047] The decision-making layer, deployed in the battery test bench, high-voltage control unit, or energy management controller, is the core of online control and fault diagnosis. It includes a state estimation and CSP identification module, a hybrid intelligent control module, and a collaborative scheduling and fault diagnosis module. Under the reference band and safety boundary constraints issued by the orchestration layer, it uses multi-source real-time signals to complete state estimation, critical stacking pressure identification, and pressure-current-temperature collaborative optimization. It also makes fault judgments and rollback decisions based on rules and SOPs.

[0048] (1) The state estimation and CSP identification module is used to: receive multi-source real-time signals uploaded by the execution perception layer, including at least stacking pressure, terminal voltage, operating current, temperature, displacement, acoustic emission signal, and impedance spectrum; and perform quality assessment and data repair, state estimation, and critical stacking pressure identification on the multi-source real-time signals.

[0049] The state estimation and CSP identification module is connected to the multimodal acquisition and synchronization module of the execution sensing layer. Specifically, the state estimation and CSP identification module estimates the current state of the battery system in real time by processing and analyzing sensor data. It uses various methods, such as signal filtering, feature extraction, and small disturbance analysis, to identify the system's operating state and stage (e.g., charging, discharging, dormancy). Based on this, the module can identify the "critical stacking pressure" plateau region, providing crucial information for subsequent decision-making. The internal settings of the state estimation and CSP identification module are as follows: (11) Data quality assessment and repair subunit.

[0050] 1) Receive real-time signals such as stacking pressure, terminal voltage, operating current, temperature, displacement, acoustic emission, and online impedance uploaded by the multi-modal acquisition and synchronization module.

[0051] 2) Within a preset time window (e.g., a 10-second sliding window), check the number of sampling points, sampling interval, and fluctuation level to determine whether the data is complete and whether there is obvious noise, dropped points, or impact.

[0052] 3) For data that does not meet the quality requirements, use methods such as filtering, interpolation, and trajectory completion to repair it; if the quality requirements are still not met after repair, mark the corresponding sensor channel as untrusted and submit the result to the collaborative scheduling and fault diagnosis module for expected safety processing.

[0053] (12) Operating status estimation unit: Under the premise of qualified data quality, the operating status of the cell and / or module is estimated, including SOC, SOH, current working stage (charging / discharging / resting), current rate, equivalent circuit model parameters, etc.; features such as pressure-capacity curve, pressure-electric derivative (dP / dQ), and coulombic efficiency are extracted to form status features for CSP identification and anomaly detection.

[0054] (13) Critical Stacking Pressure (CSP) Identification Unit.

[0055] 1) Within the safety boundary constraints, perform a small-range pressure scan or utilize natural perturbation data during operation to construct the relationship between "pressure-coulomb efficiency / overpotential / impedance".

[0056] 2) Through feature analysis and curve fitting, identify the plateau region that meets the criteria of "continuous adjustment of pressure yielding diminishing returns and no abnormal peaks", and obtain the initial estimate of CSP and its confidence level.

[0057] 3) The estimated location, width, and related features of the CSP platform area are transmitted to the hybrid intelligent control module.

[0058] (2) The hybrid intelligent control module is used to: employ a combination of an expert system (ES) and model predictive control (MPC) to perform rolling optimization control of stacking pressure, charge / discharge current, and temperature under safety boundary constraints, based on the operations corresponding to the instruction set. This allows for real-time updates of the pressure-current-temperature linkage control strategy, enabling precise control of the battery system's operating state while meeting requirements for safety, efficiency, and lifespan. The specific settings of the hybrid intelligent control module are as follows: (21) The expert system subunit calls the emergency response SOP in the orchestration layer through the expert system, and uses the safety boundary to make a logical judgment on the state estimation and the state characteristics provided by the CSP identification module to obtain a preliminary decision.

[0059] 1) Based on the rules and SOPs in the rules and security library module, perform logical judgment on the state characteristics provided by the state estimation and CSP identification module to identify whether the current state is normal operation, reference band deviation, abnormal warning or fault state.

[0060] 2) When signs such as dP / dQ sawtooth, sudden acoustic emission, or impedance abnormality are detected, preliminary degradation suggestions (i.e., preliminary decisions) are given, such as reducing the multiplier, limiting the stacking pressure adjustment speed, and requiring enhanced cooling.

[0061] (22) Model predictive control subunit: Under the reference band and the safety boundary constraints, a multivariate coupled physical model of stacked pressure-current-temperature is established. Based on the preset target, the preliminary decision and the multivariate coupled physical model, rolling optimization is performed to determine the optimal control sequence and the corresponding backoff sequence candidate scheme, and to determine a new pressure-current-temperature linkage control strategy.

[0062] 1) Under the reference band and safety boundary constraints issued by the orchestration layer, establish a multivariable coupled physical model or equivalent model of stacking pressure-current-temperature.

[0063] 2) With the goals of maintaining the CSP platform region, improving coulombic efficiency, slowing down the growth of interface impedance, and controlling the temperature rise, the pressure setting, charge / discharge current setting, and thermal management setting are continuously optimized over a period of time; the optimal control sequence and corresponding backoff sequence candidate schemes are output, and the pressure-current-temperature control strategy is updated in real time.

[0064] (23) Reference band consistency determination and perturbation optimization unit: compares the state features provided by the state estimation and CSP identification module with the reference band generated by the orchestration layer to determine whether the current running point is within the reference band; when the running point deviates from the reference band, performs a preset small perturbation on the stacking pressure within the safety boundary, and updates the CSP and forms a pressure fine-tuning suggestion in combination with the performance index changes fed back by the state estimation and CSP identification module; and transmits the pressure fine-tuning suggestion and the abnormal flag to the collaborative scheduling and fault diagnosis module; wherein, the abnormal flag refers to the current running point deviating from the reference band.

[0065] (3) The collaborative scheduling and fault diagnosis module is used to: combine the new pressure-current-temperature linkage control strategy and the state characteristics provided by the state estimation and CSP identification module, and comprehensively allocate the thermal management resources, stack pressure regulation resources and safety switching and monitoring resources of the battery system through a multi-objective coordination algorithm, and jointly optimize the thermal management strategy, stack pressure regulation strategy, cell balancing strategy and graded rollback strategy to obtain the final pressure, current and temperature control signals, thereby achieving the comprehensive optimization of efficiency and lifespan under the premise of satisfying the safety boundary constraints; perform anomaly / fault diagnosis based on the state characteristics provided by the state estimation and CSP identification module, and determine the corresponding execution measurement and rollback decision.

[0066] In short, the collaborative scheduling and fault diagnosis module is the comprehensive coordination and safety adjudication unit at the decision-making level. It is used to achieve the collaborative allocation of resources such as thermal management, stack pressure regulation, and cell balancing, and to perform fault determination and rollback control according to the Standard Operating Procedure (SOP). The internal configuration of the collaborative scheduling and fault diagnosis module is as follows: (31) Multi-objective collaborative scheduling unit: Combining the pressure-current-temperature control strategy output by the hybrid intelligent control module and the current operating status of each module, it performs comprehensive scheduling of resources such as cooling / heating unit, pressurization execution unit, cell balancing unit and safety switching unit; Under the premise of meeting the rigid constraints of safety boundary, it realizes the joint optimization of thermal management strategy, stacking pressure regulation strategy, cell balancing strategy and graded backoff strategy.

[0067] (32) Anomaly / Fault Diagnosis Unit: 1) Based on the changes in stacking pressure difference between each cell and adjacent cells, acoustic emission characteristics, and trends in overpotential and impedance changes, combined with the judgment rules given by the rules and safety library module, cell faults and sensor faults are distinguished.

[0068] 2) When a cell fault is determined, the SOP shall be issued with instructions such as disconnecting the branch and / or cluster, setting the maximum power to zero, strengthening cooling, and issuing multi-level alarms.

[0069] 3) When a sensor failure is determined, the expected safety procedures are executed, including disabling the faulty sensor, enabling the alternative channel, initiating degradation monitoring, and escalating to cell failure handling if an anomaly occurs in the alternative channel.

[0070] (33) Strategy execution and rollback decision-making unit: 1) When it is detected that the short-term forecast may touch the safety boundary, execute the rollback sequence of "flow reduction - short-term pressurization - temperature rise assessment - reassessment".

[0071] 2) Send the final determined stacking pressure setting, charging / discharging current setting, and thermal management setting to the corresponding module in the execution perception layer for execution.

[0072] 3) After summarizing the abnormal events, handling processes and results, feed them back to the instruction compilation and auditing module for audit recording and subsequent policy updates.

[0073] (iii) Execution perception layer.

[0074] The execution sensing layer is used to: execute the final pressure, current and temperature control signals, and collect multi-source real-time signals.

[0075] The execution perception layer, deployed within the battery test bench, module / cluster-level control unit, and sensor network, serves as the interface layer for interaction between the system and physical objects. Under the control commands of the decision layer, the execution perception layer performs physical execution and provides high-quality, time-aligned multimodal measurement data. The execution perception layer includes a pressure execution module (corresponding to the attached module). Figure 1 The specific settings are as follows: pressurization actuator, charge / discharge and thermal management module, and multi-modal acquisition and synchronization module.

[0076] (1) The pressurization execution module is used to: receive pressure and displacement signals from the multimodal acquisition and synchronization module; and control the internal pressure of the battery system in a closed loop based on the final pressure control signal and the pressure and displacement signals.

[0077] The pressurization execution module precisely controls the pressure actuation unit to regulate the internal pressure of the system, ensuring accurate pressure application and meeting the system's operational requirements. Through a high-precision execution mechanism, it achieves precise control over the battery system's operating state.

[0078] (11) Actuator: The actuator adopts a motor-screw mechanism, hydraulic / pneumatic actuator or other form of controllable pressure application device to apply controllable stacking pressure to the battery cells and / or modules; at the same time, it supports force control mode and displacement control mode, and can accurately adjust the actual stacking pressure according to the feedback signal.

[0079] (12) Closed-loop control unit: Receives stacked pressure setpoints from the collaborative scheduling and fault diagnosis module and / or hybrid intelligent control module; performs closed-loop control based on pressure and displacement signals transmitted back by the multimodal acquisition and synchronization module to ensure that the actual pressure tracks the setpoints and limits the rate of pressure change and maximum pressure to meet safety boundary requirements; and executes rapid depressurization or pressure limiting strategies in fault or rollback scenarios.

[0080] (2) The charge / discharge and thermal management module is used to: coordinately regulate the charging and discharging process and temperature of the battery system based on the final current control signal and the final temperature control signal. This includes the implementation of constant current control, constant voltage control, and pulse charging sequences. Simultaneously, it manages the battery temperature through a thermal control unit to ensure the system operates within a safe temperature range. Through precise current and temperature control, the battery's lifespan and performance are optimized.

[0081] (21) Charge and discharge control unit: electrically connected to power converter, power supply or load equipment, supports various strategies such as constant current, constant voltage and pulse charging; receives charge and discharge current setting value and current change rate limit instruction issued by the collaborative scheduling and fault diagnosis module; when protection is triggered, execute current reduction, stop charging and discharging or switch to safe power mode.

[0082] (22) Thermal management unit: It is connected to liquid cooling plate, air cooling channel or phase change heat dissipation structure and can perform cooling and heating; it receives temperature set value and thermal power limit value issued by the decision layer and adjusts cooling / heating intensity according to the feedback information of temperature sensor; it cooperates to perform enhanced cooling, preheating and other operations in abnormal or retreating state to reduce the risk of thermal runaway.

[0083] (23) Safety interruption and alarm execution unit: Through connection with high voltage relays, fuses and switch matrix, it performs branch disconnection, cluster disconnection or system power-off when the collaborative scheduling and fault diagnosis module issues a fault judgment; drives multi-level audible and visual alarm devices and remote alarm interface to report the fault level and handling information to the upper monitoring system.

[0084] (3) The multimodal acquisition and synchronization module is used to: configure at least a stacked pressure sensor, displacement sensor, voltage sensor, current sensor, temperature sensor, acoustic sensor, and online impedance testing unit, and synchronously acquire stacked pressure, displacement, terminal voltage, operating current, temperature, acoustic emission signal, and impedance spectrum at a preset sampling frequency to ensure that the data are aligned in time, facilitating subsequent analysis and processing. In addition, it also has data quality control capabilities, which can repair dropped points, remove noise, and ensure the accuracy of the data.

[0085] (31) Multimodal signal acquisition unit: equipped with stacked pressure sensor, displacement sensor, voltage sensor, current sensor, temperature sensor, acoustic sensor and online impedance test unit; synchronously acquire stacked pressure, terminal voltage, working current, temperature, displacement, acoustic emission signal and impedance spectrum according to preset sampling frequency.

[0086] (32) Clock synchronization and data alignment unit: Using a unified clock source or timestamp synchronization mechanism, the data from different sensors and different nodes are time aligned; the aligned data is packaged and uploaded to the state estimation and CSP identification module as the basis for state estimation and CSP identification, data quality assessment and anomaly detection.

[0087] (33) Data quality control unit: performs preliminary quality checks on the collected data, including sampling loss detection and obvious abnormal mutation detection; uploads the data quality flag along with the original data to provide a basis for the data quality assessment and repair subunit of the state estimation and CSP identification module.

[0088] In summary, the orchestration layer in this application primarily completes the transformation of business / security requirements into formalized control objectives, the generation and maintenance of reference bands and security boundaries, and the scripting and auditable logging of control strategies. Under the constraints of reference bands, rules, and security boundaries issued by the orchestration layer, the decision layer utilizes multi-source real-time signals to complete state estimation, critical stacking pressure identification, and pressure-current-temperature collaborative optimization. Through modules such as state estimation and CSP identification, hybrid intelligent control (ES+MPC), and collaborative scheduling and fault diagnosis, the decision layer, while ensuring rigid constraints on security boundaries, performs rolling optimization and graded rollback of pressure, charging / discharging current, and temperature to achieve adaptive coordinated control under different rates, temperatures, and aging stages. The execution perception layer translates upper-level decisions into precise physical execution and provides time-aligned, quality-controllable multimodal data feedback. Through modules such as pressurization execution, charging / discharging and thermal management, and multimodal acquisition and synchronization, this layer uniformly manages execution channels such as pressure, current, and temperature, as well as acquisition channels such as pressure, displacement, voltage, current, temperature, and acoustic signals. This forms a vendor-independent, standardized, and auditable measurement and control link, providing a real and traceable data foundation for upper-level strategy optimization and large language / physical model updates. This application achieves an integrated closed loop from knowledge to control, possessing adaptive capabilities across material systems and multiple operating conditions.

[0089] Based on the same inventive concept, this application also provides a method. The solution provided by this method is similar to the solution described in the above system. Therefore, the specific limitations of one or more method embodiments provided below can be found in the system limitations above, and will not be repeated here.

[0090] In an exemplary embodiment, a battery stack pressure adaptive control method is provided, applied to the above-described battery stack pressure adaptive control system, the method comprising the following steps 100-400.

[0091] Step 100: The orchestration layer determines the control objectives based on business and security requirements; based on the control objectives, a reference band, a safety boundary, and a pressure-current-temperature linkage control strategy are generated; the pressure-current-temperature linkage control strategy is compiled into instruction sets corresponding to different devices; and the reference band, the safety boundary, and the instruction sets are sent to the decision layer.

[0092] Step 200: Collect multi-source real-time signals through the perception layer and upload them to the decision layer.

[0093] Step 300: Under the constraints of the reference band and the safety boundary, the decision layer uses the multi-source real-time signals to complete state estimation, critical stacking pressure identification, and pressure-current-temperature collaborative optimization and scheduling management by combining expert system and model predictive control technology to achieve adaptive coordinated control under different rates, temperatures and aging stages, so as to obtain the final pressure, current and temperature control signals; and sends the final pressure, current and temperature control signals to the execution perception layer.

[0094] Step 400: The final pressure, current and temperature control signals are executed through the execution sensing layer, and multi-source real-time signals are continued to be collected.

[0095] In one specific application, the decision layer includes a state estimation and CSP identification module, a hybrid intelligent control module, and a cooperative scheduling and fault diagnosis module. The method further includes: (1) The state estimation and CSP identification module is used to perform quality assessment and data repair, state estimation and critical stacking pressure identification on the multi-source real-time signal.

[0096] (2) After the quality assessment is qualified, the hybrid intelligent control module adopts a combination of expert system and model predictive control technology. Under the safety boundary constraints, according to the operation corresponding to the instruction set, the stacking pressure, charging and discharging current and temperature are rolled optimization control to update the pressure-current-temperature linkage control strategy in real time.

[0097] (3) By combining the new pressure-current-temperature linkage control strategy and the state characteristics provided by the state estimation and CSP identification module, the thermal management resources, stacking pressure regulation resources and safety switching and monitoring resources of the battery system are comprehensively allocated through the multi-objective coordination algorithm. The thermal management strategy, stacking pressure regulation strategy, cell balancing strategy and graded rollback strategy are jointly optimized to obtain the final pressure, current and temperature control signals, thereby achieving the comprehensive optimization of efficiency and lifespan under the premise of satisfying the safety boundary constraints. Based on the state characteristics provided by the state estimation and CSP identification module, anomaly / fault diagnosis is performed, and the corresponding execution measurement and rollback decision is determined.

[0098] like Figure 2 As shown, the specific algorithm steps executed by the decision layer can identify and lock the stable range of critical stacking pressure online for different material systems under different magnification, temperature and cycling stages, and achieve linkage optimization with current and temperature within the safety boundary to form an auditable closed-loop control.

[0099] Step 1: Initialization and Basic Security Verification (Security Boundaries + Data Quality). Raw data is provided by the multimodal acquisition and synchronization module of the execution perception layer. Data quality assessment and state estimation are completed by the state estimation and CSP identification module of the decision layer. Basic security verification is then performed using the rules of the orchestration layer and the security boundaries provided by the security library module.

[0100] Specifically, the multimodal acquisition and synchronization module acquires and aligns real-time signals such as stacking pressure, terminal voltage, operating current, temperature, acoustic emission signal, and online impedance of a battery cell within a 10-second sliding time window, and sends the aligned multimodal data to the state estimation and CSP recognition module. Simultaneously, the target parsing and evidence retrieval module and the rule and security library module in the orchestration layer, based on a pre-trained large language model, pre-load and distribute corresponding reference bands, safety boundaries, and handling SOPs according to the current battery system and operating conditions to constrain subsequent state assessment and control decisions.

[0101] Secondly, the hybrid intelligent control module of the decision-making layer executes... Figure 2 The judgment in the process is as follows: ① Within 10 seconds, whether the data received by the multimodal acquisition and synchronization module is complete and has not crossed the safety boundary; if yes, then proceed to normal monitoring; if no, then proceed to anomaly / fault candidate identification.

[0102] During routine monitoring: The multimodal acquisition and synchronization module captures 10 seconds of data every 60 seconds to check data integrity (number of points acquired / number of points to be acquired) and stability (whether there are obvious jitters or sudden changes). At the same time, the state estimation and CSP identification module observes signs of instability (abnormal peaks in the pressure-capacity derivative, sudden acoustic emission), and finally determines whether the data quality is qualified. Figure 2 The result of judgment ② in the middle.

[0103] If the data quality is satisfactory, proceed to the reference band consistency determination step ②.

[0104] If the data quality is unsatisfactory, perform data repair (denoising + trajectory completion) and then proceed. Figure 2 In the data quality judgment ③, if the data quality is not good, the sensor channel is considered unreliable, the channel is disabled from participating in control, and the expected safety process of the sensor is switched to. If the data quality is good, the process returns to judgment ① and monitoring continues.

[0105] During anomaly / fault candidate identification: the decision-making layer's collaborative scheduling and fault diagnosis module performs fault diagnosis. The diagnostic steps include: comparing with adjacent cells, calculating pressure difference changes (abnormal channel to adjacent channel, adjacent to adjacent), and simultaneously observing whether acoustic emission / overpotential is intensifying. This is then executed... Figure 2 Judgment ④: The pressure difference between the abnormal channel and the adjacent channel continues to increase (and is greater than the change between adjacent and next adjacent channels), and acoustic emission or overpotential is significantly aggravated; if so, the cell is identified as faulty and protection is immediately initiated (disconnect the branch / cluster, disconnect the high-voltage relay, set the maximum power to zero, enhance cooling, activate the three-level alarm, and preserve data); if not, the sensor is identified as faulty and the expected safety mode is activated (disable the sensor, activate the first-level alarm, and activate the alternative channel; if acoustic emission / abnormal peak / overpotential surge occurs in the alternative channel within 5 minutes, the protection is upgraded to cell fault protection; otherwise, degraded monitoring is maintained).

[0106] Step Two: Reference Band Consistency Determination ("Fitness Check" Involving a Large Language Model): The hybrid intelligent control module at the decision layer compares real-time data with the reference band item by item: whether the pressure is within the given range, whether the overpotential is within the expected range, and whether the key impedance indicators are normal. This is then executed. Figure 2 Judgment ⑤: Whether the real-time data falls within the reference band.

[0107] If so, the platform maintenance process begins: the current stacking pressure-current-temperature settings are maintained within the reference band constraints, and the operating data is reviewed and fine-tuned according to a preset cycle.

[0108] If not, proceed to step three: pre-calibration and perturbation optimization (de-fitting).

[0109] Step 3: Pre-calibration and perturbation optimization (bringing pressure back to the CSP platform area): The decision-making layer's state estimation and CSP identification module first performs a small-scale, low-overhead pressure scan to obtain the "pressure-Coulomb efficiency / overpotential" curve, providing an initial CSP guess; subsequently, the hybrid intelligent control module applies a small perturbation to calculate "which direction to adjust better," forming a pressure fine-tuning suggestion. Execution follows the fine-tuning. Figure 2 Judgment ⑥: Whether the platform's criteria have been met (e.g., "Continuing to adjust will not significantly improve the situation", and there are no abnormal peaks).

[0110] If yes, the platform maintains (keeping the current small perturbation and review frequency); if no, the projection is updated (updating the pressure within the safe upper and lower limits), and then it returns to the micro-perturbation to continue searching until judgment ⑥ is met or a safe rollback is triggered.

[0111] At this point, the critical stacking pressure can be found to optimize battery performance. Then, step four is performed to ensure battery safety under optimal performance: under the critical stacking pressure, and taking into account other factors such as temperature, actively predict whether the battery will touch the safety boundary in the short term. If it will, adjust it in advance.

[0112] Step Four: Collaborative and Safety Projection (Pressure-Current-Temperature, these three variables are controlled together): Within the safety boundary, the pressurization execution module and the charging / discharging and thermal management module simultaneously consider the combination of pressure, charging / discharging current, and temperature to predict in advance whether the boundary will be approached within a short period of time. This is then executed... Figure 2 Judgment ⑦: Will short-term predictions touch the safety boundary?

[0113] If yes, execute the rollback sequence (first reduce current, then briefly increase pressure, then increase temperature, then re-evaluate; if still abnormal, protect / shut down); if no, maintain the current settings and return to step three / platform to maintain the loop.

[0114] Step 5: Policy Update and Feedback (Making the Large Language Model More Accurate with Use): Upload the data, anomalies, handling, and effects obtained in the above steps to the orchestration layer; based on this, the preset large language model refreshes the reference band, security boundaries, SOPs, and sampling / control plans, while exporting executable scripts and tooling parameters for reuse in the next batch or across platforms. This process allows for closer targeting in the next application, resulting in fewer scans and faster platform region locking.

[0115] Compared with the prior art, this application has the following advantages: 1. The battery stacking pressure adaptive control method, which adopts a large language model-physical model co-strategy, enables various materials to automatically optimize and maintain the critical stacking pressure plateau region within the safety boundary under various conditions of rate / temperature / cycle stages, based on multiple indicators such as ES+MPC, dP / dQ, CE, and η. This significantly improves coulombic efficiency, reduces interface impedance growth, and maintains robustness under harsh conditions such as low temperature and fast charging.

[0116] 2. It adopts a pressure-current-temperature triple-coordinated and provably safe closed-loop control mechanism, which can generate strategies and SOPs for handling anomalies from a large language model and implement projection control by MPC constrained by physical inequalities. When dP / dQ sawtooth, acoustic emission burst or EIS anomaly occurs, the system self-heals and relocks the critical stacking pressure, actively suppressing short-circuit risks and significantly shortening the anomaly recovery time.

[0117] 3. Adopting an integrated executable process card automatic generation and end-side migration method for manufacturing and operation and maintenance, it can translate the knowledge organization and physical verification of large language models into equipment API / PLC scripts and distill them into lightweight agents, realizing a direct connection from "text knowledge to executable control". While maintaining safety boundaries and traceability, it reduces DOE costs, reduces batch variance, and accelerates rapid migration across chemical systems / bench.

[0118] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0119] This document uses specific examples to illustrate the principles and implementation methods of this application. The descriptions of the above embodiments are only for the purpose of helping to understand the methods and core ideas of this application. Furthermore, those skilled in the art will recognize that, based on the ideas of this application, there will be changes in the specific implementation methods and application scope. Therefore, the content of this specification should not be construed as a limitation of this application.

Claims

1. A battery stacking pressure adaptive control system, characterized in that, The system includes an orchestration layer, a decision-making layer, and an execution awareness layer; The orchestration layer is used for: determining control objectives based on business and security requirements; generating reference bands, safety boundaries, and pressure-current-temperature linkage control strategies based on the control objectives; compiling the pressure-current-temperature linkage control strategies into instruction sets corresponding to different devices; and distributing the reference bands, safety boundaries, and instruction sets to the decision layer. The orchestration layer calls a preset large language model to perform semantic understanding and constraint extraction on business objectives and security requirements provided by the upper-level monitoring system or maintenance personnel in natural language or structured form, converting them into target coulomb efficiency ranges, allowable stacking pressure ranges, temperature upper limits, current change rate constraints, and allowable perturbation amplitudes. Based on the performance indicators and constraints obtained from the conversion, literature, experimental data and historical operating data that match the current material system, magnification, temperature and cycling stage are retrieved from the preset knowledge base to generate the reference band that matches the current operating conditions. The decision layer is used to: receive multi-source real-time signals uploaded by the execution perception layer; under the constraints of the reference band and the safety boundary, utilize the multi-source real-time signals to complete state estimation, critical stacking pressure identification, and pressure-current-temperature collaborative optimization and scheduling management combining expert system and model predictive control technology to achieve adaptive coordinated control under different rates, temperatures and aging stages, so as to obtain the final pressure, current and temperature control signals; and send the final pressure, current and temperature control signals to the execution perception layer. The decision-making layer includes a state estimation and CSP identification module, a hybrid intelligent control module, and a collaborative scheduling and fault diagnosis module. The state estimation and CSP identification module is used to: receive multi-source real-time signals uploaded by the execution sensing layer, including at least stacking pressure, terminal voltage, operating current, temperature, displacement, acoustic emission signal, and impedance spectrum; and based on the multi-source real-time signals, perform a small-range pressure scan within the reference band and the safety boundary constraints, or utilize natural perturbation data during operation to construct the relationship between stacking pressure and coulombic efficiency, overpotential, and impedance. By using feature analysis and curve fitting, a CSP plateau region is identified that satisfies the decreasing performance gains characterized by the coulombic efficiency, overpotential, and impedance when the stacking pressure is further adjusted, and that the overpotential and impedance have no abnormal peak values. The location, width, and confidence level of the CSP plateau region are obtained, and the location, width, and confidence level of the CSP plateau region are transmitted to the hybrid intelligent control module. The hybrid intelligent control module is used to: employ a combination of expert system and model predictive control technology to perform rolling optimization control on stacking pressure, charging / discharging current, and temperature under the safety boundary constraints, based on the operation corresponding to the instruction set, so as to update the pressure-current-temperature linkage control strategy in real time; wherein, the expert system calls a preset emergency response SOP and uses the safety boundary to perform logical judgment on the state characteristics provided by the state estimation and CSP identification module, including the position, width, and confidence level of the CSP platform area, to obtain a preliminary decision; the model predictive control technology is used to establish a multivariate coupled physical model of stacking pressure-current-temperature under the reference band and the safety boundary constraints, with the preset objectives of maintaining the CSP platform area, improving coulombic efficiency, slowing down the growth of interface impedance, and controlling the temperature rise, and performing rolling optimization based on the preliminary decision and the multivariate coupled physical model to determine the optimal control sequence and corresponding backoff sequence candidate schemes, and to determine a new pressure-current-temperature linkage control strategy; The collaborative scheduling and fault diagnosis module is used to: combine the new pressure-current-temperature linkage control strategy and the state characteristics provided by the state estimation and CSP identification module, and comprehensively allocate the thermal management resources, stacking pressure regulation resources, and safety switching and monitoring resources of the battery system through a multi-objective coordination algorithm; and jointly optimize the thermal management strategy, stacking pressure regulation strategy, cell balancing strategy, and graded rollback strategy to obtain the final pressure, current, and temperature control signals, thereby achieving comprehensive optimization of efficiency and lifespan while meeting the constraints of the safety boundary; and perform anomaly / fault diagnosis based on the state characteristics provided by the state estimation and CSP identification module, and determine the corresponding execution measures and rollback decisions. The execution sensing layer is used to: execute the final pressure, current and temperature control signals, and collect multi-source real-time signals.

2. The battery stacking pressure adaptive control system according to claim 1, characterized in that, The orchestration layer includes a target parsing and evidence retrieval module, a rule and security library module, and an instruction compilation and auditing module; The target parsing and evidence retrieval module is used to: call and use a preset large language model to extract control targets from business and security requirements; perform structured parsing of the control targets and convert them into performance indicators and constraints of various devices in the battery system; and perform evidence retrieval from a preset knowledge base based on the performance indicators and constraints to generate a reference band. The rules and security library module is used to: store and maintain the security boundaries and pressure-current-temperature linkage control strategies of various devices, and integrate emergency response SOPs; and to perform version management and permission management for each version of the security boundaries, the pressure-current-temperature linkage control strategies and the emergency response SOPs. The instruction compilation and auditing module is used to: receive the reference band, the safety boundary, and the pressure-current-temperature linkage control strategy; compile the pressure-current-temperature linkage control strategy into instruction sets corresponding to different devices; track and record the execution process of each instruction in the instruction set in real time; and send the reference band, the safety boundary, and the instruction set to the decision layer.

3. The battery stacking pressure adaptive control system according to claim 2, characterized in that, The safety boundaries include at least the upper and lower limits of stacking pressure, upper and lower limits of current, upper and lower limits of temperature, overpotential threshold, and acoustic emission and impedance anomaly threshold corresponding to different battery chemical systems, module structures, stacking methods, and cooling structures.

4. The battery stacking pressure adaptive control system according to claim 1, characterized in that, The execution perception layer includes a pressurization execution module, a charging / discharging and thermal management module, and a multimodal acquisition and synchronization module; The pressurization execution module is used to: receive pressure and displacement signals from the multimodal acquisition and synchronization module; and, based on the final pressure control signal, perform closed-loop control of the internal pressure of the battery system according to the pressure and displacement signals. The charge / discharge and thermal management module is used to: coordinately regulate the charging and discharging process and temperature of the battery system based on the final current control signal and the final temperature control signal; The multimodal acquisition and synchronization module is used to: configure at least a stacked pressure sensor, displacement sensor, voltage sensor, current sensor, temperature sensor, acoustic sensor and online impedance testing unit, and synchronously acquire stacked pressure, displacement, terminal voltage, operating current, temperature, acoustic emission signal and impedance spectrum at a preset sampling frequency.

5. The battery stacking pressure adaptive control system according to claim 1, characterized in that, The orchestration layer is deployed on servers or edge computing nodes; The decision-making layer is deployed in a battery test bench, high-voltage control unit, or energy management controller; The execution perception layer is deployed in the battery test bench, module / cluster level control unit, and sensor network.

6. The battery stacking pressure adaptive control system according to claim 2, characterized in that, The rules and security library module includes a rules and handling SOP management unit and a version governance and traceability unit; The rules and handling SOP management unit is used to: store the rules corresponding to the pressure-current-temperature linkage control strategy, including at least the anomaly / fault identification rules, degraded operation rules, and rollback strategy rules; and store emergency handling SOPs to guide the decision-making level to perform operations such as disconnecting faulty branches, degrading operation, enhancing cooling, data preservation, and alarm level classification according to the pre-approved steps when abnormal conditions are triggered. The version governance and traceability unit is used to: manage the version and permissions of each version of the safety boundary, the pressure-current-temperature linkage control strategy and the emergency response SOP, and record at least the creation time, effective time, modifier and applicable scope; and transmit and store the current version validity information of the safety boundary, the pressure-current-temperature linkage control strategy and the emergency response SOP to the instruction compilation and auditing module.

7. The battery stacking pressure adaptive control system according to claim 1, characterized in that, The hybrid intelligent control module also includes a reference band consistency determination and perturbation optimization unit; The reference band consistency determination and perturbation optimization unit is used to: compare the state features provided by the state estimation and CSP identification module with the reference band to determine whether the current operating point is within the reference band; when the current operating point deviates from the reference band, perform a preset small perturbation on the pressure within the safety boundary, and update the critical stacking pressure and form a pressure fine-tuning suggestion based on the performance index changes fed back by the state estimation and CSP identification module; and transmit the pressure fine-tuning suggestion and anomaly flag to the collaborative scheduling and fault diagnosis module; wherein, the anomaly flag indicates that the current operating point deviates from the reference band.

8. A battery stacking pressure adaptive control method, applied to the battery stacking pressure adaptive control system according to any one of claims 1-7, characterized in that, The method includes: The orchestration layer determines control objectives based on business and security requirements; based on the control objectives, a reference band, a safety boundary, and a pressure-current-temperature linkage control strategy are generated; the pressure-current-temperature linkage control strategy is compiled into instruction sets corresponding to different devices; and the reference band, the safety boundary, and the instruction sets are distributed to the decision layer. The system collects real-time signals from multiple sources through the perception layer and uploads them to the decision layer. Under the constraints of the reference band and the safety boundary, the decision layer utilizes multi-source real-time signals to complete state estimation, critical stacking pressure identification, and pressure-current-temperature collaborative optimization and scheduling management by combining expert systems and model predictive control technology. This enables adaptive coordinated control under different rates, temperatures, and aging stages to obtain the final pressure, current, and temperature control signals. The final pressure, current, and temperature control signals are then sent down to the execution perception layer. The final pressure, current, and temperature control signals are executed through the execution sensing layer, and multi-source real-time signals are continuously acquired.

9. The battery stacking pressure adaptive control method according to claim 8, characterized in that, The decision-making layer includes a state estimation and CSP identification module, a hybrid intelligent control module, and a collaborative scheduling and fault diagnosis module. The method further includes: The state estimation and CSP identification module is used to perform quality assessment and data repair, state estimation and critical stacking pressure identification on the multi-source real-time signal. After the quality assessment is qualified, the hybrid intelligent control module adopts a combination of expert system and model predictive control technology. Under the safety boundary constraints, it performs rolling optimization control on stacking pressure, charging and discharging current and temperature according to the operation corresponding to the instruction set, so as to update the pressure-current-temperature linkage control strategy in real time. By combining the collaborative scheduling and fault diagnosis module with the new pressure-current-temperature linkage control strategy and the state characteristics provided by the state estimation and CSP identification module, a multi-objective coordination algorithm is used to comprehensively allocate the thermal management resources, stacking pressure regulation resources, and safety switching and monitoring resources of the battery system. Furthermore, the thermal management strategy, stacking pressure regulation strategy, cell balancing strategy, and graded rollback strategy are jointly optimized to obtain the final pressure, current, and temperature control signals. This achieves optimal efficiency and lifespan while meeting the constraints of the safety boundary. Based on the state characteristics provided by the state estimation and CSP identification module, anomaly / fault diagnosis is performed, and corresponding execution measurement and rollback decisions are determined.

Images