A method and device for modeling an MWh-level aqueous sodium-ion battery energy storage system architecture

Abstract

This invention discloses a modeling method and apparatus for MWh-level aqueous sodium-ion battery energy storage systems, aiming to improve the accuracy and engineering adaptability of modeling large-capacity aqueous sodium-ion battery energy storage systems. The method comprises four steps: first, decomposing the system into five core modules, extracting key parameters to establish a standardized database, and identifying core parameters such as battery internal resistance and PCS conversion efficiency through sensitivity analysis; second, constructing a dedicated quantitative model for each module's characteristics and embedding formulas to characterize dynamic properties; third, integrating the modular models based on digital twin bus coupling technology, and establishing a virtual-real bidirectional interactive channel to achieve precise synchronization; and finally, verifying the model performance based on actual operating data under multiple operating conditions. This method achieves full-dimensional and dynamic modeling of the system, solving the problems of fragmented modules and virtual-real disconnect in traditional modeling, providing support for the design, optimization, and safe operation of energy storage systems, and is suitable for engineering applications of MWh-level aqueous sodium-ion battery energy storage systems.

Description

Technical Field

[0001] This invention relates to the field of energy storage system modeling and virtual simulation technology, and more specifically, to a method and apparatus for modeling the architecture of an MWh-level aqueous sodium-ion battery energy storage system. Background Technology

[0002] With the rapid development of the new energy industry, MWh-level aqueous sodium-ion battery energy storage systems have been widely used in grid peak shaving and distributed energy storage scenarios due to their advantages such as high safety, low cost, and environmental friendliness. Accurate modeling of the energy storage system architecture is a core prerequisite for achieving system performance optimization, operational status prediction, and fault simulation analysis, directly determining the operational stability and energy efficiency level of the energy storage system.

[0003] Existing energy storage system modeling methods suffer from the following shortcomings: First, modular decomposition is incomplete, focusing primarily on battery units or single PCS modules while neglecting the coupling effects of integrated containers, temperature control, and fire suppression systems, resulting in insufficient model completeness. Second, parameter processing lacks standardized procedures, leading to mixed parameters of different dimensions, resulting in low accuracy in sensitivity analysis and poor targeting of core parameter selection, failing to provide accurate basis for model optimization. Third, module integration is low, with delays in data interaction between sub-models and a lack of a "physical-virtual" two-way linkage mechanism, making it difficult to achieve dynamic collaborative simulation. Fourth, verification conditions are limited, relying solely on rated operating condition data to verify model accuracy, failing to cover complex real-world operating scenarios and restricting model practicality.

[0004] Therefore, there is an urgent need for a modeling method and device for MWh-level aqueous sodium-ion battery energy storage system architecture that takes into account modular integrity, parameter accuracy, integration and synergy, and verification reliability, in order to overcome the shortcomings of existing technologies. Summary of the Invention

[0005] The purpose of this invention is to provide a modeling method and device for the architecture of an MWh-level aqueous sodium-ion battery energy storage system, which solves the problems of incomplete module decomposition, non-standard parameter processing, poor integration and coordination, and limited verification scenarios in existing modeling methods. This invention enables accurate modeling and dynamic simulation of the energy storage system architecture, providing reliable support for system optimization.

[0006] The technical solution of this invention is as follows:

[0007] A method for modeling the architecture of an MWh-level aqueous sodium-ion battery energy storage system includes the following steps:

[0008] S1: System Component Disassembly, Parameterization, and Sensitivity Analysis: The energy storage system is defined as consisting of a container integration module, an electrical wiring module, a battery energy storage unit module, an energy storage converter (PCS) module, and a temperature control and fire-fighting linkage module. Key technical parameters of each module are extracted and a standardized parameter database is established. The core parameters affecting system performance are determined, including the battery cell internal resistance R, open-circuit voltage U, PCS conversion efficiency η, temperature control threshold T, and fire response delay time t.

[0009] S2: Modular modeling: Based on the structural characteristics and working principles of each component, a container integrated structure mechanical-thermal coupling model, an electrical system wiring impedance loss model, a battery energy storage unit series-parallel active balancing model, a PCS bidirectional conversion sliding mode control model, a temperature control fuzzy adaptive PID model, and a fire multi-source sensing linkage response model are constructed respectively. Each model is embedded with a dedicated quantitative formula to achieve dynamic characteristic representation.

[0010] S3: System Integration Modeling: Based on digital twin bus coupling technology and heterogeneous signal interaction link construction, data interaction protocol and real-time mapping function are introduced to integrate the modular models into a complete energy storage system architecture model. Dynamic collaborative simulation between modules is realized through real-time data mapping, and a two-way interaction channel between "physical entity and virtual model" is constructed to achieve precise synchronization between physical and virtual quantities.

[0011] S4: Model Validation: The performance of the integrated model is validated based on actual operating data under multiple working conditions.

[0012] Furthermore, in S4, an improved genetic algorithm is used to iteratively adjust the core parameters, and weighted least squares fitting is used to correct the model error, so that the model accuracy meets the requirements of energy storage system design, operation and safety assessment, wherein the SOC estimation error is ≤2%, the temperature control accuracy error is ≤±1℃, and the system response delay error is ≤50ms.

[0013] Furthermore, S1 includes: breaking down the energy storage system into six core modules according to function: container integration, electrical wiring, battery energy storage unit, PCS, temperature control, and fire protection; extracting key technical parameters of each module and unifying parameter symbols; and establishing an unambiguous standardized parameter database.

[0014] Furthermore, the core parameters of S1 are determined as follows: All original parameters are converted into standardized parameters in the [0,1] interval using the min-max normalization formula to eliminate the influence of different dimensions. Negative parameters such as resistance and delay time are additionally converted to 1-x to unify them into positive indicators. A perturbation matrix is ​​constructed based on the standardized parameters. A fixed perturbation step size of 0.15 is set to generate perturbation matrices of corresponding orders. MATLAB / Simulink simulations are used to obtain the changes in three performance indicators—SOC estimation accuracy, temperature control accuracy, and response speed—under different perturbation combinations for each parameter. The parameters are then quantified and substituted into the sensitivity index calculation formula to solve for the global sensitivity mean μ and local sensitivity standard deviation σ of each parameter. μ represents the overall influence of the parameter on the system performance, and σ represents the nonlinear strength of the parameter's influence on the system performance in different disturbance ranges. A dual screening threshold of μ≥0.3 and σ≥0.1 is set to screen parameters that have a significant impact on the system performance and obvious nonlinear characteristics from all parameters. Finally, the battery cell internal resistance, open circuit voltage, PCS conversion efficiency, temperature control threshold, and fire response delay time are determined as the core parameters of the system, which provide clear target objects for subsequent modeling and optimization of the energy storage system.

[0015] Furthermore, S2 includes: a container integrated structure mechanics-thermal coupling model: based on the finite element analysis concept, a model is built using the ANSYS multiphysics coupling module. The model accurately simulates the heat exchange process between the battery array and the external environment through a three-dimensional unsteady-state heat conduction equation. Simultaneously, mechanical constraint equations limit the actual stress of the container structure under battery self-weight and external loads to not exceed the allowable stress of the material, achieving dual control over heat conduction characteristic simulation and structural load stability; an electrical system wiring impedance loss model: comprehensively considering the combined effects of wire resistance, terminal contact resistance, and loop inductance, a formula for calculating the total impedance of the AC loop is constructed. By extracting the real component of the impedance and combining it with the loop active power loss formula, the energy consumption of the electrical wiring links is accurately quantified, providing core data support for the overall system energy efficiency analysis; a battery energy storage unit series-parallel active balancing model: based on a series-parallel hybrid topology, the SOC value of each battery unit is first calculated with high precision using an ampere-hour integral + open-circuit voltage correction formula to reduce estimation errors; a 5% SOC difference is set as the active balancing trigger threshold, and the balancing current is adjusted using an active balancing current control formula to achieve rapid convergence of the SOC of each battery unit; PC S-shaped bidirectional sliding mode control model: Addressing the bidirectional charging and discharging requirements of the PCS, a mathematical model is constructed in the dq synchronous rotating coordinate system to achieve decoupled control of active and reactive power. First, a sliding mode switching function is defined to characterize the d-axis current tracking error. An exponential reaching law formula is used to rapidly converge the system state to the sliding surface and suppress chattering. Finally, the sliding mode control law formula is derived by combining the PCS voltage equation to adjust the d-axis control voltage, achieving accurate active power tracking. Temperature control fuzzy adaptive PID model: A model is designed to address the nonlinear and time-varying characteristics of the system temperature. Temperature control deviation and the rate of change of deviation are defined as input variables. The PID parameter adjustment amount is output through 5×5 fuzzy rule base inference, and then the PID control parameters are updated in real time through the PID parameter correction formula to adapt to the temperature control requirements of different working conditions; Fire multi-source sensing linkage response model: integrates temperature, smoke and voltage multi-source sensing modules, and quantifies the fire risk judgment value through multi-source sensing weighted fusion formula. The judgment value ≥0.8 is set as the trigger threshold. Once the threshold is met, a three-level linkage response is initiated. Combining hardware triggering and software control, the fire response delay is controlled within 0.5s, realizing accurate perception, quantitative judgment and rapid handling of fire risks.

[0016] Furthermore, S3 includes: Basic construction of bus coupling and signal interaction: using a digital twin bus as the core of data interaction for the entire system, establishing the OPC UA 2.0 standardized data interaction protocol to achieve heterogeneous signal compatibility among mechanical, electrical, control, and safety module models; simultaneously building a three-level signal interaction link of perception-transmission-modeling, where the perception layer collects physical entity current, voltage, and other operating data at a 100Hz sampling frequency, the transmission layer uploads the data to the digital twin bus within 20ms via 5G / industrial Ethernet, and the modeling layer receives the data and drives the virtual model to run synchronously, completing real-time data linkage between the physical and virtual systems; high-precision bidirectional mapping and closed-loop iteration between physical and virtual systems: establishing a bidirectional mapping relationship between the parameters of the physical entity and the virtual model based on linear mapping + dynamic error compensation formula, calibrating the proportional coefficient α and offset coefficient β using historical data, and updating the dynamic error compensation in real time based on the deviation at the previous moment. The virtual model outputs system optimization instructions through simulation, which are then sent to the physical system for execution via the digital twin bus. The execution results are then fed back to the perception layer to complete the closed-loop iteration. Full module integration and dynamic collaborative simulation under all operating conditions: The battery unit, PCS, temperature control, fire protection and other sub-module models are coupled and integrated through the digital twin bus. The data types, 100Hz transmission frequency, trigger conditions and other interface specifications are uniformly defined to build a complete energy storage system architecture model and realize the dynamic collaborative simulation of each module. During the simulation, the rated charging and discharging, battery failure and fire alarm typical operating conditions are simulated in real time. The collaborative performance between modules is verified by comparing the virtual simulation data and the physical operation data to ensure that the virtual model accurately reproduces the real operating state of the physical system.

[0017] Furthermore, S4 includes: collecting sufficient actual operating sample data of four typical operating conditions of the energy storage system: rated power charging and discharging, partial load, extreme temperature, and battery fault simulation; extracting three core performance indicators: SOC, temperature, and system response time; and verifying whether the model meets the basic error requirements by comparing the actual values ​​of the indicators with the model simulation values ​​to calculate the error.

[0018] Furthermore, in S2, the mechanical-thermal coupling model of the container integrated structure includes the following three-dimensional unsteady heat conduction equation:

[0019]

[0020] In the formula: ρ is the density of the container material; c_p is the specific heat capacity of the material at constant pressure; k is the thermal conductivity of the material; q_v is the heat generation per unit volume of the battery array, which is calculated from the battery charging and discharging current and internal resistance; T(x,y,z,t) is the temperature (K) of a certain point (x,y,z) inside the container at time t. The Laplace operator characterizes the temperature gradient distribution;

[0021] Simultaneously, mechanical constraints are introduced to ensure the stability of the container structure under battery weight and external loads. The constraint equations are as follows:

[0022]

[0023] In the formula: σ(x,y,z) is the actual stress at a certain point of the container structure; [σ] is the allowable stress of the material. Through coupling analysis, the accurate simulation of heat conduction characteristics and the dual control of structural load stability are achieved.

[0024] Furthermore, in S2, the electrical system wiring impedance loss model includes: constructing an AC circuit total impedance model, where power loss is solved by the real part of the impedance, achieving accurate quantification of energy consumption in the electrical wiring links. The total impedance calculation formula is as follows:

[0025]

[0026] In the formula: R_w is the resistance of the conductor, calculated by the resistance law R_w=ρL3 / S2; R_j is the contact resistance of the terminal; ω is the AC angular frequency, ω=2πf, f is the system frequency; L2 is the loop inductance;

[0027] The formula for calculating the active power loss of a circuit is:

[0028]

[0029] In the formula: I is the circuit operating current; Re(Z) is the real part of the total impedance; P_loss is the active power loss of the electrical wiring circuit, providing the core quantitative basis for the overall energy efficiency analysis of the energy storage system.

[0030] A modeling device for an MWh-level aqueous sodium-ion battery energy storage system architecture includes a parameter processing and sensitivity analysis module, a modular modeling module, a system integration modeling module, and a model verification module. These modules work in tandem to achieve virtual modeling of the energy storage system architecture, as detailed below:

[0031] The parameter processing and sensitivity analysis module is used to identify the constituent modules of the MWh-level aqueous sodium-ion battery energy storage system. These modules include a container integration module, an electrical wiring module, a battery energy storage unit module, an energy storage converter (PCS) module, and a temperature control and fire-fighting linkage module. Key technical parameters of each constituent module are extracted, a standardized parameter database is established, and sensitivity analysis is used to determine the core parameters affecting system performance. These core parameters include the battery cell internal resistance R, open-circuit voltage U, PCS conversion efficiency η, temperature control threshold T, and fire-fighting response delay time t.

[0032] The modular modeling module is used to construct a dedicated dynamic model based on the structural characteristics and working principle of each component module. The dedicated dynamic model includes a container integrated structural mechanics-thermal coupling model, an electrical system wiring impedance loss model, a battery energy storage unit series-parallel active balancing model, a PCS bidirectional conversion sliding mode control model, a temperature control fuzzy adaptive PID model, and a fire protection multi-source sensing linkage response model. Each dedicated dynamic model embeds a corresponding quantitative formula to characterize the module's dynamic characteristics.

[0033] The system integration modeling module is used to build a heterogeneous signal interaction link based on digital twin bus coupling technology, introduce data interaction protocol and real-time mapping function, integrate the dedicated dynamic models built by the sub-module modeling module to form a complete virtual model of energy storage system architecture; realize dynamic collaborative simulation between modules through real-time data mapping, build a two-way interaction channel between "physical entity and virtual model" to ensure accurate synchronization between physical and virtual quantities.

[0034] The model verification module is used to acquire actual operating data of the energy storage system under multiple operating conditions, and to perform performance verification on the complete energy storage system architecture virtual model constructed by the system integration modeling module based on the actual operating data, and output the model verification results to determine whether the model accuracy meets the preset requirements.

[0035] This invention's modeling method focuses on the full-dimensional physical characteristics and dynamic operating laws of MWh-level large-capacity aqueous sodium-ion battery energy storage systems. Following a four-layer logical architecture of "partial decomposition - precise modeling - system coupling - empirical verification," it achieves precise, dynamic, and interactive modeling and characterization of the architecture of large-capacity aqueous sodium-ion battery energy storage systems through multi-dimensional parameter analysis, heterogeneous module-specific modeling, digital twin-driven system integration, and closed-loop design for actual operating condition verification. The core working logic and underlying mechanism are as follows:

[0036] First, the energy storage system is disassembled into modular components to clarify the functional boundaries of six core modules, including container integration, electrical wiring, and battery energy storage units. A standardized database is established by extracting key technical parameters. At the same time, sensitivity analysis is used to identify core parameters that play a decisive role in system performance, such as battery cell internal resistance, PCS conversion efficiency, and temperature control threshold. This defines the core research dimensions for subsequent modular modeling, eliminates redundant parameter interference, and achieves "focusing on the big picture and letting go of the small details" and precise focus in modeling.

[0037] Dedicated models were designed for the structural characteristics, working principles, and dynamic response laws of the six major modules, breaking through the limitations of traditional single-dimensional modeling. Adaptive modeling methods and quantification formulas were adopted for different modules: a mechanical-thermal coupling model was constructed for the container integration module to characterize the interaction between the structure and the thermal environment; a wiring impedance loss model was constructed for the electrical system to quantify the energy loss of electrical transmission; a series-parallel active balancing model was constructed for the battery cells to match the series-parallel operation characteristics of aqueous sodium-ion batteries; advanced control models such as sliding mode control and fuzzy adaptive PID were constructed for the PCS and temperature control modules to characterize the dynamic control process of power conversion and temperature regulation; and a multi-source sensing linkage response model was constructed for the fire protection module to characterize the dynamic triggering and response laws of safety protection, achieving a full-dimensional quantitative characterization of the physical characteristics, operating characteristics, control characteristics, and safety characteristics of each module.

[0038] Based on digital twin bus coupling technology and heterogeneous signal interaction links, a unified data interaction protocol and real-time mapping function are introduced to break down the heterogeneity barriers between various sub-module models. This organically integrates sub-module models from different dimensions such as structure, electrical, battery, control, and safety, constructing a complete energy storage system architecture model. Real-time data mapping enables dynamic collaborative simulation between modules. Simultaneously, a two-way interactive channel between the physical entity and the virtual model is established, allowing the parameters of the virtual model to be precisely synchronized with the actual operating physical quantities of the physical system. This achieves real-time mapping and dynamic following of the virtual model to the physical entity, enabling the integrated model to truly reflect the coupling effects and overall operating rules between the modules of a large-capacity energy storage system.

[0039] Based on actual operating data under multiple operating conditions, the integrated system architecture model is fully validated. By comparing and analyzing the simulation results of the model with the actual operating data of the physical system, the accuracy, dynamism and adaptability of the model are verified, forming a closed loop of "modeling-simulation-verification-correction". This ensures that the model can match the actual performance of the MWh-level aqueous sodium-ion battery energy storage system under different operating conditions, and avoids the disconnect between theoretical modeling and engineering practice.

[0040] Overall, the core mechanism of this invention is based on modular disassembly, with dedicated quantitative modeling as the core, digital twin technology as the coupling means, and actual working condition verification as the guarantee. It achieves accurate modeling of the MWh-level aqueous sodium-ion battery energy storage system at all levels of "component-module-system", all dimensions of "structure-electrical-control-safety", and bidirectional interaction of "physical-virtual", so that the model can realistically, dynamically and comprehensively represent the overall architecture and operation law of the large-capacity aqueous sodium-ion battery energy storage system.

[0041] Advantages and effects of the present invention

[0042] This invention's modeling method addresses the engineering requirements of MWh-level large-capacity energy storage systems and the technical characteristics of aqueous sodium-ion batteries. It overcomes the pain points of traditional energy storage system modeling, such as "fragmented modules, single dimension, disconnect between virtual and real systems, and low accuracy." It boasts multiple advantages, including technical relevance, refined modeling, system integration, and engineering practicality, achieving significant results in model construction, engineering application, and system optimization.

[0043] 1. By identifying the core parameters affecting system performance through sensitivity analysis and eliminating redundant parameters, the computational load and complexity of modeling are reduced, improving modeling efficiency. This also allows the modeling process to focus on the key dimensions that play a decisive role in system operation, avoiding model distortion caused by redundant parameters and significantly improving modeling accuracy.

[0044] 2. Customized models are designed for the characteristics of different modules, covering all dimensions such as structural mechanics, thermal, electrical loss, battery balancing, power conversion control, temperature regulation control, and safety and fire protection linkage. At the same time, customized quantitative formulas are embedded in each model to achieve accurate quantification of the dynamic characteristics of each module. This breaks through the limitations of traditional modeling that only focuses on the single dimension of electrical or battery, and can fully characterize the characteristics and interactions of each module in a large-capacity energy storage system.

[0045] 3. A dedicated active balancing model for series and parallel operation and electrochemical characteristics of aqueous sodium-ion batteries was designed. This model differs from the modeling method of lithium battery energy storage systems, making the model more consistent with the actual operating rules of aqueous sodium-ion battery energy storage systems and filling the gap in dedicated modeling methods for MWh-level aqueous sodium battery energy storage systems.

[0046] 4. Based on digital twin bus coupling technology and heterogeneous signal interaction links, a unified data interaction protocol and real-time mapping function are introduced to solve the heterogeneity compatibility problem of different dimensions and types of modular models, realize the seamless integration of each module model, and enable the model to truly reflect the coupling effect and collaborative operation law between modules of the MWh-level large-capacity energy storage system, avoiding the distortion of the overall system characteristics caused by traditional "module splicing" modeling.

[0047] 5. Construct a two-way interactive channel between "physical entity and virtual model" to achieve precise synchronization between physical and virtual quantities. This allows the virtual model to map the actual operating state of the physical system in real time. At the same time, the simulation optimization of the virtual model can guide the operation adjustment of the physical system, realizing the two-way empowerment of "virtual simulation guiding physical design and physical operation correcting the virtual model". This lays the core model foundation for the digital twin application of energy storage systems.

[0048] 6. By enabling dynamic collaborative simulation between modules through real-time data mapping, the system can accurately characterize the dynamic response of each module and the overall operation changes of the MWh-level energy storage system under different scenarios such as charge / discharge switching, load fluctuations, ambient temperature changes, and fault triggering, allowing the model to simulate the dynamic operation process of the system under all operating conditions. Attached Figure Description

[0049] Figure 1 This is a flowchart of the modeling method of the present invention.

[0050] Figure 2 This is a schematic diagram of the electrical wiring of the present invention.

[0051] Figure 3 This is a schematic diagram of the energy storage system structure of the present invention.

[0052] Figure 4 This is a schematic diagram of the temperature control fire protection structure of the present invention. Detailed Implementation

[0053] See Figure 1 , 4 .

[0054] Example 1:

[0055] 1. A method for modeling the architecture of an MWh-level aqueous sodium-ion battery energy storage system, comprising the following steps:

[0056] S1: System Component Disassembly, Parameterization, and Sensitivity Analysis: The energy storage system is defined as consisting of a container integration module, an electrical wiring module, a battery energy storage unit module, an energy storage converter (PCS) module, and a temperature control and fire-fighting linkage module. Key technical parameters of each module are extracted and a standardized parameter database is established. The core parameters affecting system performance are determined, including the battery cell internal resistance R, open-circuit voltage U, PCS conversion efficiency η, temperature control threshold T, and fire response delay time t.

[0057] S2: Modular modeling: Based on the structural characteristics and working principles of each component, a container integrated structure mechanical-thermal coupling model, an electrical system wiring impedance loss model, a battery energy storage unit series-parallel active balancing model, a PCS bidirectional conversion sliding mode control model, a temperature control fuzzy adaptive PID model, and a fire multi-source sensing linkage response model are constructed respectively. Each model is embedded with a dedicated quantitative formula to achieve dynamic characteristic representation.

[0058] S3: System Integration Modeling: Based on digital twin bus coupling technology and heterogeneous signal interaction link construction, data interaction protocol and real-time mapping function are introduced to integrate the modular models into a complete energy storage system architecture model. Dynamic collaborative simulation between modules is realized through real-time data mapping, and a two-way interaction channel between "physical entity and virtual model" is constructed to achieve precise synchronization between physical and virtual quantities.

[0059] S4: Model Validation: The performance of the integrated model is validated based on actual operating data under multiple working conditions.

[0060] Furthermore, in S4, an improved genetic algorithm is used to iteratively adjust the core parameters, and weighted least squares fitting is used to correct the model error, so that the model accuracy meets the requirements of energy storage system design, operation and safety assessment, wherein the SOC estimation error is ≤2%, the temperature control accuracy error is ≤±1℃, and the system response delay error is ≤50ms.

[0061] Furthermore, in another embodiment of the present invention, S1 includes: decomposing the energy storage system into six core modules according to function: container integration, electrical wiring, battery energy storage unit, PCS, temperature control, and fire protection; extracting key technical parameters of each module and unifying parameter symbols; and establishing an unambiguous standardized parameter database.

[0062] Furthermore, in another embodiment of the present invention, the core parameters of S1 are determined as follows: All original parameters are converted into standardized parameters in the [0,1] interval using a min-max normalization formula to eliminate the influence of different dimensions of parameters. Negative parameters such as resistance and delay time are additionally converted to 1-x to unify them into positive indicators. A perturbation matrix is ​​constructed based on the standardized parameters, and a fixed perturbation step size of 0.15 is set to generate a perturbation matrix of the corresponding order. Simulation using MATLAB / Simulink is used to obtain the system SOC estimation accuracy, temperature control accuracy, and response speed under different perturbation combinations of each parameter. The changes in performance indicators are then substituted into the sensitivity index calculation formula to solve for the global sensitivity mean μ and local sensitivity standard deviation σ of each parameter. μ represents the overall impact of the parameter on system performance, and σ represents the nonlinear strength of the parameter's impact on system performance in different disturbance ranges. A dual screening threshold of μ≥0.3 and σ≥0.1 is set to screen parameters that have a significant impact on system performance and obvious nonlinear characteristics from all parameters. Finally, the battery cell internal resistance, open circuit voltage, PCS conversion efficiency, temperature control threshold, and fire response delay time are determined as the core parameters of the system, which provide clear target objects for subsequent modeling and optimization of the energy storage system.

[0063] Furthermore, in another embodiment of the present invention, S2 includes: a container integrated structure mechanics-thermal coupling model: based on the finite element analysis concept, a model is built using the ANSYS multiphysics coupling module. The three-dimensional unsteady-state heat conduction equation accurately simulates the heat exchange process between the battery array and the external environment. Simultaneously, mechanical constraint equations limit the actual stress of the container structure under battery self-weight and external loads to not exceed the allowable stress of the material, achieving dual control over heat conduction characteristic simulation and structural load stability; an electrical system wiring impedance loss model: comprehensively considering the combined effects of wire resistance, terminal contact resistance, and loop inductance, a formula for calculating the total impedance of the AC loop is constructed. By extracting the real component of the impedance and combining it with the loop active power loss formula, the energy consumption of the electrical wiring links is accurately quantified, providing core data support for the overall system energy efficiency analysis; a battery energy storage unit series-parallel active balancing model: based on a series-parallel hybrid topology, the SOC value of each battery unit is first calculated with high precision using the ampere-hour integral + open-circuit voltage correction formula to reduce estimation errors. A 5% SOC difference is set as the active balancing trigger threshold, and the balancing current magnitude is adjusted using the active balancing current control formula to achieve rapid SOC convergence of each battery unit. To achieve consistency; PCS bidirectional sliding mode control model: For the bidirectional charging and discharging requirements of PCS, a mathematical model is constructed in the dq synchronous rotating coordinate system to achieve decoupled control of active and reactive power; firstly, a sliding mode switching function is defined to characterize the d-axis current tracking error. An exponential reaching law formula is used to make the system state quickly converge to the sliding surface and suppress chattering. Finally, the sliding mode control law formula is derived by combining the PCS voltage equation to adjust the d-axis control voltage, achieving accurate active power tracking; Temperature control fuzzy adaptive PID model: A model is designed for the nonlinear and time-varying characteristics of system temperature. Firstly, the temperature control deviation and the deviation change rate are defined as inputs. The variables are inferred from a 5×5 fuzzy rule base to output PID parameter adjustment amounts, and then the PID control parameters are updated in real time through the PID parameter correction formula to adapt to the temperature control requirements of different operating conditions; Fire multi-source sensing linkage response model: integrates temperature, smoke, and voltage multi-source sensing modules, and quantifies the fire risk judgment value through a multi-source sensing weighted fusion formula. The judgment value ≥ 0.8 is set as the trigger threshold. Once the threshold is met, a three-level linkage response is initiated. Combining hardware triggering and software control, the fire response delay is controlled within 0.5s, realizing accurate perception, quantitative judgment, and rapid handling of fire risks.

[0064] Furthermore, in another embodiment of the present invention, S3 includes: Building the foundation for bus coupling and signal interaction: using a digital twin bus as the core of data interaction for the entire system, establishing the OPC UA 2.0 standardized data interaction protocol to achieve heterogeneous signal compatibility among mechanical, electrical, control, and safety module models; simultaneously building a three-level signal interaction link of perception-transmission-modeling, where the perception layer collects physical entity current, voltage, and other operating data at a 100Hz sampling frequency, the transmission layer uploads the data to the digital twin bus within 20ms via 5G / industrial Ethernet, and the modeling layer receives the data and drives the virtual model to run synchronously, completing real-time data linkage between the physical and virtual systems; High-precision bidirectional mapping and closed-loop iteration between physical and virtual systems: establishing a bidirectional mapping relationship between the parameters of the physical entity and the virtual model based on a linear mapping + dynamic error compensation formula, calibrating the proportional coefficient α and offset coefficient β using historical data, and updating the dynamic error compensation in real time based on the deviation at the previous moment. The virtual model outputs system optimization instructions through simulation, which are then sent to the physical system for execution via the digital twin bus. The execution results are then fed back to the perception layer to complete the closed-loop iteration. Full module integration and dynamic collaborative simulation under all operating conditions: The battery unit, PCS, temperature control, fire protection and other sub-module models are coupled and integrated through the digital twin bus. The data types, 100Hz transmission frequency, trigger conditions and other interface specifications are uniformly defined to build a complete energy storage system architecture model and realize the dynamic collaborative simulation of each module. During the simulation, the rated charging and discharging, battery failure and fire alarm typical operating conditions are simulated in real time. The collaborative performance between modules is verified by comparing the virtual simulation data and the physical operation data to ensure that the virtual model accurately reproduces the real operating state of the physical system.

[0065] Furthermore, in another embodiment of the present invention, S4 includes: collecting sufficient actual operating sample data of four typical operating conditions of the energy storage system: rated power charging and discharging, partial load, extreme temperature, and battery fault simulation; extracting three core performance indicators: SOC, temperature, and system response time; and verifying whether the model meets the basic error requirements by comparing the actual values ​​of the indicators with the model simulation values ​​to calculate the error.

[0066] Furthermore, in another embodiment of the present invention, in S2, the mechanical-thermal coupling model of the container integrated structure includes the following three-dimensional unsteady heat conduction equation:

[0067]

[0068] In the formula: ρ is the density of the container material; c_p is the specific heat capacity of the material at constant pressure; k is the thermal conductivity of the material; q_v is the heat generation per unit volume of the battery array, which is calculated from the battery charging and discharging current and internal resistance; T(x,y,z,t) is the temperature (K) of a certain point (x,y,z) inside the container at time t. The Laplace operator characterizes the temperature gradient distribution;

[0069] Simultaneously, mechanical constraints are introduced to ensure the stability of the container structure under battery weight and external loads. The constraint equations are as follows:

[0070]

[0071] In the formula: σ(x,y,z) is the actual stress at a certain point of the container structure; [σ] is the allowable stress of the material. Through coupling analysis, the accurate simulation of heat conduction characteristics and the dual control of structural load stability are achieved.

[0072] Furthermore, in another embodiment of the present invention, in S2, the electrical system wiring impedance loss model includes: constructing an AC circuit total impedance model, where power loss is solved by the real part of the impedance, thereby achieving accurate quantification of energy consumption in the electrical wiring links. The total impedance calculation formula is as follows:

[0073]

[0074] In the formula: R_w is the resistance of the conductor, calculated by the resistance law R_w=ρL3 / S2; R_j is the contact resistance of the terminal; ω is the AC angular frequency, ω=2πf, f is the system frequency; L2 is the loop inductance;

[0075] The formula for calculating the active power loss of a circuit is:

[0076]

[0077] In the formula: I is the circuit operating current; Re(Z) is the real part of the total impedance; P_loss is the active power loss of the electrical wiring circuit, providing the core quantitative basis for the overall energy efficiency analysis of the energy storage system.

[0078] Example 2

[0079] A modeling device for an MWh-level aqueous sodium-ion battery energy storage system architecture includes a parameter processing and sensitivity analysis module, a modular modeling module, a system integration modeling module, and a model verification module. These modules work in tandem to achieve virtual modeling of the energy storage system architecture, as detailed below:

[0080] The parameter processing and sensitivity analysis module is used to identify the constituent modules of the MWh-level aqueous sodium-ion battery energy storage system. These modules include a container integration module, an electrical wiring module, a battery energy storage unit module, an energy storage converter (PCS) module, and a temperature control and fire-fighting linkage module. Key technical parameters of each constituent module are extracted, a standardized parameter database is established, and sensitivity analysis is used to determine the core parameters affecting system performance. These core parameters include the battery cell internal resistance R, open-circuit voltage U, PCS conversion efficiency η, temperature control threshold T, and fire-fighting response delay time t.

[0081] The modular modeling module is used to construct a dedicated dynamic model based on the structural characteristics and working principle of each component module. The dedicated dynamic model includes a container integrated structural mechanics-thermal coupling model, an electrical system wiring impedance loss model, a battery energy storage unit series-parallel active balancing model, a PCS bidirectional conversion sliding mode control model, a temperature control fuzzy adaptive PID model, and a fire protection multi-source sensing linkage response model. Each dedicated dynamic model embeds a corresponding quantitative formula to characterize the module's dynamic characteristics.

[0082] The system integration modeling module is used to build a heterogeneous signal interaction link based on digital twin bus coupling technology, introduce data interaction protocol and real-time mapping function, integrate the dedicated dynamic models built by the sub-module modeling module to form a complete virtual model of energy storage system architecture; realize dynamic collaborative simulation between modules through real-time data mapping, build a two-way interaction channel between "physical entity and virtual model" to ensure accurate synchronization between physical and virtual quantities.

[0083] The model verification module is used to acquire actual operating data of the energy storage system under multiple operating conditions, and to perform performance verification on the complete energy storage system architecture virtual model constructed by the system integration modeling module based on the actual operating data, and output the model verification results to determine whether the model accuracy meets the preset requirements.

[0084] Application example:

[0085] See Figure 1 , 4 .

[0086] A method for modeling the architecture of an MWh-level aqueous sodium-ion battery energy storage system includes the following steps:

[0087] S1: System component decomposition, parameterization, and sensitivity analysis

[0088] 1.1 Component Disassembly and Parameter Extraction

[0089] The MWh-level aqueous sodium-ion battery energy storage system is functionally broken down into 6 core modules. Key technical parameters for each module are extracted and a standardized database is established. The specific parameters are as follows:

[0090] Container integrated module: Dimensional parameters (length L1, width W1, height H1), insulation layer thickness δ, heat dissipation area S1, structural load-bearing strength F;

[0091] Electrical wiring module: conductor cross-sectional area S2, conductor resistivity ρ, terminal contact resistance R_j, loop inductance L2;

[0092] Battery energy storage unit module: cell internal resistance R_d, open circuit voltage U_oc, rated capacity C_r, charge / discharge rate C_rate, number of series and parallel connections (N_s series × N_p parallel);

[0093] PCS module: Conversion efficiency η, rated power P_pcs, input / output voltage range U_pcs, switching frequency f_s;

[0094] Temperature control module: temperature control threshold T_set, heat dissipation power P_cool, heating power P_heat, fan speed n_fan;

[0095] Firefighting module: fire response delay time t_delay, sensor detection range S3, extinguishing medium release rate v_ext.

[0096] 1.2 Parameter Standardization Processing

[0097] All extracted parameters are standardized using the min-max normalization formula to eliminate the influence of dimensions, resulting in a standardized parameter x with a value range of [0,1], ensuring the accuracy of subsequent sensitivity analysis. This method is suitable for positive parameters (the larger the parameter value, the better the performance). For negative parameters (such as resistance and delay time), 1-x can be added to the formula to convert them into positive indicators.

[0098]

[0099] In the formula: x is the original parameter value; x_min is the minimum value of the parameter in the database; x_max is the maximum value of the parameter in the database; The standardized parameter values ​​(0≤x≤1).

[0100] 1.3 Improved Morris Sensitivity Analysis

[0101] A perturbation matrix is ​​constructed based on standardized parameters, with a perturbation step size Δ=0.15, generating an N×(N+1) order perturbation matrix B (N being the total number of parameters). Each parameter is quantitatively perturbed within its value range. System performance indicators (SOC estimation accuracy ε_SOC, temperature control accuracy ε_T, and response speed t_resp) under different perturbation combinations are obtained through MATLAB / Simulink simulation. These indicators are then substituted into the sensitivity index calculation formula to obtain the global sensitivity mean μ (reflecting the overall impact of the parameter on system performance) and the local sensitivity standard deviation σ (reflecting the difference in the parameter's impact across different perturbation intervals).

[0102]

[0103] In the formula: M is the number of simulation experiments (matching the dimension of the perturbation matrix); ΔY_i is the change in the system performance index under the i-th perturbation; Δx_i is the standardized perturbation increment of the parameter in the i-th perturbation (here Δx_i=0.15); μ is the global sensitivity mean (the larger μ is, the more significant the overall influence of the parameter); σ is the local sensitivity standard deviation (the larger σ is, the stronger the nonlinearity of the parameter influence).

[0104] Parameters with μ≥0.3 and σ≥0.1 were selected as core parameters. Finally, the battery cell internal resistance R_d, open circuit voltage U_oc, PCS conversion efficiency η, temperature control threshold T_set, and fire response delay time t_delay were determined as core parameters to provide targeted objects for subsequent modeling and optimization.

[0105] S2: Modular Modeling

[0106] Each module model is constructed based on multiphysics coupling, control theory and signal processing technology. The core relationships are quantified by formulas to achieve accurate characterization and control of module performance.

[0107] 2.1 Mechanical-Thermal Coupled Model of Container Integrated Structure

[0108] Based on the finite element method (FEA) approach, a coupled model of the container structure and heat conduction was constructed using the ANSYS multiphysics coupling module, simultaneously considering the heat exchange (conduction, convection, and radiation) between the battery array and the external environment. The three-dimensional unsteady-state heat conduction equations are as follows:

[0109]

[0110] In the formula: ρ is the density of the container material (kg / m³) 3 c_p is the specific heat capacity at constant pressure of the material (J / (kg·K)); k is the thermal conductivity of the material (W / (m·K)); q_v is the heat generation per unit volume of the battery array (W / m³). 3 The value q_v is calculated from the battery charging and discharging current and internal resistance (q_v = I²R_d / V_cell, where V_cell is the volume of a single battery cell); T(x,y,z,t) is the temperature (K) of a point (x,y,z) inside the container at time t; ∇ is the Laplace operator, which characterizes the temperature gradient distribution.

[0111] Simultaneously, mechanical constraints are introduced to ensure the stability of the container structure under battery weight (total weight G = N_s × N_p × m_cell, where m_cell is the mass of a single battery cell) and external loads (such as wind and seismic loads). The constraint equations are as follows:

[0112]

[0113] In the formula: σ(x,y,z) is the actual stress (MPa) at a certain point in the container structure; [σ] is the allowable stress of the material (MPa), such as [σ]=235MPa for Q235 steel. Coupled analysis enables precise simulation of heat conduction characteristics and dual control of structural load stability.

[0114] 2.2 Electrical System Wiring Impedance Loss Model

[0115] Taking into account the combined effects of conductor resistance, terminal contact resistance, and circuit inductance in an electrical circuit, a total impedance model for the AC circuit is constructed. Power loss is solved by the real part of the impedance, achieving accurate quantification of energy consumption in electrical wiring. The formula for calculating the total impedance is as follows:

[0116]

[0117] In the formula: R_w is the resistance of the conductor (Ω), calculated by the resistance law R_w=ρL3 / S2 (L3 is the length of the conductor, m; S2 is the cross-sectional area of ​​the conductor, m²). 2 R_j is the contact resistance of the terminal block (Ω), typically taken as 10. -3 ~10 -2 The value is in the order of Ω; ω is the AC angular frequency (rad / s), ω=2πf, f is the system frequency (Hz, usually 50Hz); L2 is the loop inductance (H).

[0118] The formula for calculating the active power loss in a circuit (generated solely by the resistive component) is as follows:

[0119]

[0120] In the formula: I is the operating current of the circuit (A); Re(Z) is the real part of the total impedance (i.e., the resistance component, Ω); P_loss is the active power loss of the electrical wiring circuit (W), which provides the core quantitative basis for the overall energy efficiency analysis of the energy storage system.

[0121] 2.3 Active Balancing Model for Series-Parallel Connection of Battery Energy Storage Units

[0122] A series-parallel hybrid topology is adopted to construct an active balancing model based on the SOC difference, calculate the SOC value of each battery cell, and reduce the estimation error. A 5% SOC difference (ΔSOC≥5%) is set as the active balancing trigger threshold, and the balancing current is adjusted by the balancing current control formula to achieve rapid consistency of SOC among the cells.

[0123]

[0124] In the formula: SOC(t) is the battery SOC value at time t; SOC(t0) is the SOC value at the initial time (t0); C_r is the battery rated capacity (Ah); I(t) is the charging and discharging current at time t (A, positive for charging and negative for discharging); k_ocv is the open-circuit voltage correction coefficient (% / V), which is determined experimentally (usually taken as 5~8% / V); U_oc is the actual open-circuit voltage (V); U_oc,ref is the open-circuit voltage under standard SOC (V, such as U_oc when SOC=50%).

[0125] Active equalization current control formula:

[0126]

[0127] In the formula: I_bal is the balancing current (A); k_bal is the balancing coefficient (0.1-0.2), corresponding to a balancing rate of 0.1C-0.2C; ΔSOC is the SOC difference between adjacent battery cells (%); C_r is the rated capacity of the battery (Ah). By adjusting I_bal, the SOC of each cell can be quickly made consistent, thereby improving the cycle life and charge / discharge stability of the battery pack.

[0128] Table 1-1 Parameters of 1125 kW Energy Storage Converter

[0129] 2.4 PCS Two-Way Transform Sliding Mode Control Model

[0130] To address the bidirectional charging and discharging requirements of PCS, a mathematical model is constructed in the dq synchronous rotating coordinate system to achieve decoupled control of active power (d-axis current) and reactive power (q-axis current). A sliding mode switching function and an exponential reaching law are designed to suppress sliding mode chattering. The d-axis voltage is adjusted by a control law to achieve precise tracking of active power.

[0131] First, define the sliding mode switching function (d-axis current tracking error):

[0132]

[0133] In the formula: i_d,ref is the d-axis reference current (corresponding to the target active power); i_d is the d-axis actual current.

[0134] The exponential reaching law formula is used to make the system state converge quickly to the sliding surface, while suppressing chattering.

[0135]

[0136] In the formula: ε is the approach coefficient (>0, usually taken as 510); k is the exponential coefficient (>0, usually taken as 0.52); sign(s) is the sign function (1 when s>0, -1 when s<0, and 0 when s=0).

[0137] Based on the PCS voltage equation in the dq coordinate system, derive the sliding mode control law formula and adjust the d-axis control voltage u_d:

[0138]

[0139] In the formula: L is the PCS filter inductance (H); R_i is the equivalent resistance of the inductor (Ω); ω is the grid angular frequency (rad / s); i_q is the q-axis current; C is the filter capacitor (F); i_grid,d is the grid d-axis current (A). This control law enables precise tracking of active power, ensures the stability of PCS charging and discharging mode switching, and controls the conversion efficiency fluctuation within ±2%.

[0140] 2.5 Temperature Control Fuzzy Adaptive PID Model

[0141] To address the nonlinear and time-varying temperature characteristics of energy storage systems, a fuzzy adaptive PID model is designed. This model adjusts PID parameters in real-time through fuzzy inference to adapt to temperature control requirements under different operating conditions. Using the temperature control deviation *e* and the rate of change of deviation *ec* as inputs, the model outputs the PID parameter adjustment to achieve precise temperature control.

[0142] First, define the input variables:

[0143]

[0144] In the formula: T_set is the target temperature control threshold (°C); T_act is the actual temperature (°C); t_k and t_{k-1} are adjacent sampling times (s).

[0145] Using a 5×5 fuzzy rule base (where e and ec are each divided into 5 fuzzy subsets: negative large (NB), negative small (NS), zero (ZO), positive small (PS), and positive large (PB), the PID parameter adjustment values ​​ΔK_p, ΔK_i, and ΔK_d are inferred and output. The corrected PID parameter formulas are as follows:

[0146]

[0147] In the formula: K_p0, K_i0, and K_d0 are the initial parameters of the PID controller (determined through engineering tuning); ΔK_p, ΔK_i, and ΔK_d are the parameter adjustment values ​​output by fuzzy inference. Ultimately, this ensures that the temperature control accuracy error is controlled within ±1.5℃, laying the foundation for subsequent optimization.

[0148] 2.6 Firefighting Multi-Source Sensing and Linkage Response Model

[0149] Integrate a multi-source perception module for temperature, smoke, and voltage monitoring, adopt a weighted fusion algorithm to quantify the fire risk, and set a judgment threshold. When F≥0.8, a hierarchical linkage response is triggered. The multi-source perception weighted fusion formula is as follows:

[0150]

[0151] In the formula: F is the fire response judgment value (0≤F≤1); ω1, ω2, and ω3 are the weights of each perception module (satisfying ω1 + ω2 + ω3 = 1, determined by the analytic hierarchy process, usually ω1 = 0.4, ω2 = 0.3, ω3 = 0.3); T is the normalized temperature perception value (converted by the formula, calculated based on the temperature over-standard amplitude); S is the normalized smoke concentration value (0 means no smoke, 1 means the alarm threshold concentration); is the normalized value of voltage abnormality (calculated based on the battery voltage deviation amplitude).

[0152] When F≥0.8, a three-level hierarchical linkage response is triggered. By combining hardware triggering and software control, the fire response delay time t_delay is controlled within 0.5s, realizing multi-source accurate perception, quantitative judgment, and rapid hierarchical disposal.

[0153] S3: System integration modeling

[0154] Taking the digital twin bus as the core, realize heterogeneous compatibility of sub-module models, physical-virtual two-way mapping, and dynamic collaborative simulation, build a digital twin model with all elements and all processes, and provide a virtual carrier for system optimization.

[0155] 3.1 Bus coupling and signal interaction construction

[0156] Adopt the digital twin bus as the core of data interaction, formulate the OPC UA 2.0 standardized data interaction protocol, and realize the heterogeneous signal compatibility of sub-module models (mechanical, electrical, control, safety). Build a three-level signal interaction link of "perception layer - transmission layer - modeling layer": The perception layer collects real-time operation data of physical entities (current, voltage, temperature, SOC, etc., sampling frequency 100Hz) through sensors; The transmission layer uploads the data to the digital twin bus through 5G / industrial Ethernet, and the transmission delay is controlled within 20ms; The modeling layer receives the data and drives the virtual model to run synchronously, realizing the real-time linkage of physical and virtual systems.

[0157] 3.2 Physical-virtual two-way mapping

[0158] Based on the linear mapping + dynamic error compensation formula, establish a two-way mapping relationship between the parameters of physical entities and virtual models. By dynamically adjusting the mapping coefficient, ensure that the two-way mapping accuracy is ≥98%.

[0159]

[0160] In the formula: V(t) is the virtual model parameter value at time t; P(t) is the physical entity parameter value at time t; α and β are mapping coefficients (calibrated by historical data, α is the proportional coefficient, and β is the offset coefficient); ε(t) is the dynamic error compensation amount (ε(t) = P(t) - V(t) deviation at the previous time, updated in real time).

[0161] Based on this mapping relationship, a closed-loop link of "physical acquisition - virtual simulation - optimization instructions - physical execution" is constructed: the virtual model outputs optimization instructions through simulation, which are then sent to the physical system for execution via the digital twin bus. The execution results are fed back to the perception layer to complete the closed-loop iteration and realize the reverse guidance of the virtual to the physical.

[0162] 3.3 System Integration and Dynamic Co-simulation

[0163] By coupling and integrating the modular models through a digital twin bus, a complete energy storage system architecture model is built. A unified interface specification (data type, 100Hz transmission frequency, trigger conditions) is defined to achieve dynamic collaborative simulation of modules such as battery cells, PCS, temperature control, and fire protection. During the simulation, typical operating conditions such as rated charge / discharge, charge / discharge switching, extreme temperature fluctuations, battery faults (sudden internal resistance changes), and fire alarms are simulated in real time. By comparing virtual simulation data with physical operating data, the collaborative performance between modules is verified, ensuring that the virtual model can accurately reproduce the real operating state of the physical system, with a mapping error ≤2%.

[0164] S4: Model Validation

[0165] 4.1 Performance Verification of Multi-condition Model

[0166] One hundred sets of actual operating sample data (each set containing 20 parameters, sampling interval 1s) of energy storage system were collected under four typical operating conditions: ① rated power charge / discharge condition; ② partial load (50% rated power) operation condition; ③ extreme temperature (-20℃ / 45℃) condition; ④ battery fault simulation condition. The actual operating data were compared with the model simulation data to calculate the error of core performance indicators.

[0167]

[0168] In the formula: SOC_act and SOC_sim are the actual and simulated SOC values, respectively; T_act and T_sim are the actual and simulated temperature values, respectively; t_resp,act and t_resp,sim are the actual and simulated response time values, respectively. This preliminary verification checks whether the model meets the basic requirements (error ≤ 5%).

[0169] The present invention also includes:

[0170] 4.2 Optimization of core parameters based on improved genetic algorithm

[0171] Using the core parameters (R_d, U_oc, η, T_set, t_delay) as optimization variables, a multi-objective fitness function is designed, with the optimization objective being to minimize the three types of core errors. The multi-objective fitness function formula (weighted sum method) is as follows:

[0172]

[0173] In the formula: J is the fitness value (the smaller the better); ω_a, ω_b, ω_c are error weights (satisfying ω_a+ω_b+ω_c=1, set according to performance priority, usually ω_a=0.4, ω_b=0.3, ω_c=0.3).

[0174] The improved genetic algorithm operation process is as follows: Initialize the population (population size N=50, iteration count G=100, variable value range is set based on the actual interval of the core parameters) → Calculate individual fitness values ​​→ Selection operation (roulette wheel method, retaining superior individuals) → Crossover operation (single-point crossover, crossover probability P_c=0.7) → Mutation operation (site mutation, mutation probability P_m=0.05) → Determine the termination condition (fitness value convergence or iteration count reached). Through iterative adjustment of the core parameters, the error indicators are initially reduced by more than 30%.

[0175] 4.3 Model Error Correction Based on Least Squares Fitting

[0176] To address the systematic errors that still exist in the model after optimization using the genetic algorithm, a least squares fitting method is used for linear correction to minimize the deviation between the model output value and the actual value.

[0177]

[0178] In the formula: y_corr is the corrected model output value; y_sim is the optimized model simulation output value; a and b are fitting coefficients (a is the proportional coefficient, and b is the offset coefficient).

[0179] Based on n=100 sets of sample data, construct an objective function to minimize the sum of squared errors, and solve for a and b:

[0180]

[0181] Taking the partial derivatives with respect to a and b and setting them to 0, we obtain the normal equation system, which can be solved to obtain:

[0182]

[0183] After the correction is completed, the model performance is verified again to ensure that the SOC estimation error is ≤2%, the temperature control accuracy error is ≤±1℃, and the system response delay error is ≤50ms, so as to meet the actual application requirements.

[0184] 4.4 Model Iterative Update Mechanism

[0185] Establish a long-term mechanism for model iteration and updates. Collect new operating data of the energy storage system on a quarterly basis (covering different seasons and different load conditions), repeat steps S4.1-S4.3, and dynamically adjust the model parameters (a, b, initial values ​​of PID, etc.) and structure (such as the aging battery internal resistance correction model) to adapt to model deviations caused by system aging, changes in operating conditions, and environmental fluctuations, so as to ensure that the model maintains high accuracy and applicability in the long term.

Claims

1. A method for modeling the architecture of an MWh-level aqueous sodium-ion battery energy storage system, characterized in that, Includes the following steps: S1: System Component Disassembly, Parameterization, and Sensitivity Analysis: The energy storage system is defined as consisting of a container integration module, an electrical wiring module, a battery energy storage unit module, an energy storage converter (PCS) module, and a temperature control and fire-fighting linkage module. Key technical parameters of each module are extracted and a standardized parameter database is established. The core parameters affecting system performance are determined, including the battery cell internal resistance R, open-circuit voltage U, PCS conversion efficiency η, temperature control threshold T, and fire response delay time t. S2: Modular modeling: Based on the structural characteristics and working principles of each component, a container integrated structure mechanical-thermal coupling model, an electrical system wiring impedance loss model, a battery energy storage unit series-parallel active balancing model, a PCS bidirectional conversion sliding mode control model, a temperature control fuzzy adaptive PID model, and a fire multi-source sensing linkage response model are constructed respectively. Each model is embedded with a dedicated quantitative formula to achieve dynamic characteristic representation. S3: System Integration Modeling: Based on digital twin bus coupling technology and heterogeneous signal interaction link construction, data interaction protocol and real-time mapping function are introduced to integrate the modular models into a complete energy storage system architecture model. Dynamic collaborative simulation between modules is realized through real-time data mapping, and a two-way interaction channel between "physical entity and virtual model" is constructed to achieve precise synchronization between physical and virtual quantities. S4: Model Validation: The performance of the integrated model is validated based on actual operating data under multiple working conditions.

2. The method for modeling the architecture of an MWh-level aqueous sodium-ion battery energy storage system according to claim 1, characterized in that, In S4, an improved genetic algorithm is used to iteratively adjust the core parameters, and weighted least squares fitting is used to correct the model error, so that the model accuracy meets the requirements of energy storage system design, operation and safety assessment. Among them, the SOC estimation error is ≤2%, the temperature control accuracy error is ≤±1℃, and the system response delay error is ≤50ms.

3. The method for modeling the architecture of an MWh-level aqueous sodium-ion battery energy storage system according to claim 1, characterized in that, S1 includes: breaking down the energy storage system into six core modules according to function: container integration, electrical wiring, battery energy storage unit, PCS, temperature control, and fire protection; extracting key technical parameters of each module and unifying parameter symbols; and establishing an unambiguous standardized parameter database.

4. The method for modeling the architecture of an MWh-level aqueous sodium-ion battery energy storage system according to claim 1, characterized in that, The core parameters of S1 are determined as follows: all original parameters are converted into standardized parameters in the [0,1] interval using the min-max normalization formula to eliminate the influence of different dimensions of parameters. Negative parameters such as resistance and delay time are additionally converted to 1-x to unify them into positive indicators. A perturbation matrix is ​​constructed based on the standardized parameters. A fixed perturbation step size of 0.15 is set to generate a perturbation matrix of the corresponding order. The changes in three performance indicators of the system SOC estimation accuracy, temperature control accuracy, and response speed under different perturbation combinations of parameters are obtained through MATLAB / Simulink simulation. Substituting the values ​​into the sensitivity index calculation formula, the global sensitivity mean μ and local sensitivity standard deviation σ of each parameter are calculated. μ represents the overall impact of the parameter on system performance, while σ represents the nonlinear strength of the parameter's impact on system performance in different disturbance ranges. A dual screening threshold of μ≥0.3 and σ≥0.1 is set to screen parameters that have a significant impact on system performance and exhibit obvious nonlinear characteristics from all parameters. Finally, the battery cell internal resistance, open-circuit voltage, PCS conversion efficiency, temperature control threshold, and fire response delay time are determined as the core parameters of the system, providing clear target objects for subsequent modeling and optimization of the energy storage system.

5. The method for modeling the architecture of an MWh-level aqueous sodium-ion battery energy storage system according to claim 1, characterized in that, The S2 mentioned above includes a container integrated structural mechanics-thermal coupling model: based on the finite element analysis concept, the model is built with the help of the ANSYS multiphysics coupling module. The three-dimensional unsteady heat conduction equation accurately simulates the heat exchange process between the battery array and the outside world. At the same time, the mechanical constraint equation limits the actual stress of the container structure under the battery self-weight and external load to not exceed the allowable stress of the material, so as to achieve dual control of heat conduction characteristic simulation and structural load stability. Electrical system wiring impedance loss model: Taking into account the combined effects of conductor resistance, terminal contact resistance, and loop inductance, a formula for calculating the total impedance of the AC loop is constructed. By extracting the real components of the impedance and combining them with the active power loss formula of the loop, the energy consumption of electrical wiring links is accurately quantified, providing core data support for the overall energy efficiency analysis of the system. Battery energy storage unit series-parallel active balancing model: Based on a series-parallel hybrid topology, the SOC value of each battery unit is first calculated with high precision using an ampere-hour integral + open-circuit voltage correction formula to reduce estimation errors. A 5% SOC difference is set as the active balancing trigger threshold, and the balancing current is adjusted using an active balancing current control formula to achieve rapid convergence of the SOC of each battery unit. PCS bidirectional switching sliding mode control model: For the bidirectional charging and discharging requirements of PCS, a mathematical model is constructed in the dq synchronous rotating coordinate system to achieve decoupled control of active and reactive power. A sliding mode switching function is first defined to characterize the d-axis current tracking error, and then a sliding mode switching function is used to characterize the d-axis current tracking error. The approach law formula enables the system state to converge quickly to the sliding mode surface and suppresses chattering. Finally, the sliding mode control law formula is derived by combining the PCS voltage equation to adjust the d-axis control voltage, achieving accurate active power tracking. The temperature control fuzzy adaptive PID model is designed for the nonlinear and time-varying characteristics of system temperature. The temperature control deviation and the deviation change rate are defined as input variables. The PID parameter adjustment is output through inference from a 5×5 fuzzy rule base. Then, the PID control parameters are updated in real time through the PID parameter correction formula to adapt to the temperature control requirements of different operating conditions. The fire protection multi-source sensing linkage response model integrates temperature, smoke, and voltage multi-source sensing modules. The fire risk judgment value is quantitatively calculated through a multi-source sensing weighted fusion formula. The judgment value ≥ 0.8 is set as the trigger threshold. Once the threshold is met, a three-level linkage response is initiated. The fire response delay is controlled within 0.5s by combining hardware triggering and software control, achieving accurate perception, quantitative judgment, and rapid handling of fire risks.

6. The method for modeling the architecture of an MWh-level aqueous sodium-ion battery energy storage system according to claim 1, characterized in that, The S3 mentioned above includes: Basic construction of bus coupling and signal interaction: using a digital twin bus as the core of data interaction for the entire system, establishing the OPC UA 2.0 standardized data interaction protocol to achieve heterogeneous signal compatibility among mechanical, electrical, control, and safety module models; simultaneously building a three-level signal interaction link of perception-transmission-modeling. The perception layer collects operating data such as current and voltage of the physical entity at a sampling frequency of 100Hz; the transmission layer uploads the data to the digital twin bus within 20ms via 5G / industrial Ethernet; the modeling layer receives the data and drives the virtual model to run synchronously, completing real-time data linkage between the physical and virtual systems; high-precision bidirectional mapping and closed-loop iteration between physical and virtual systems: establishing a bidirectional mapping relationship between the parameters of the physical entity and the virtual model based on linear mapping + dynamic error compensation formula, calibrating the proportional coefficient α and offset coefficient β using historical data, and updating the dynamic error compensation in real time based on the deviation at the previous moment. The virtual model outputs system optimization instructions through simulation, which are then sent to the physical system for execution via the digital twin bus. The execution results are then fed back to the perception layer to complete the closed-loop iteration. Full module integration and dynamic collaborative simulation under all operating conditions: The battery unit, PCS, temperature control, fire protection and other sub-module models are coupled and integrated through the digital twin bus. The data types, 100Hz transmission frequency, trigger conditions and other interface specifications are uniformly defined to build a complete energy storage system architecture model and realize the dynamic collaborative simulation of each module. During the simulation, the rated charging and discharging, battery failure and fire alarm typical operating conditions are simulated in real time. The collaborative performance between modules is verified by comparing the virtual simulation data and the physical operation data to ensure that the virtual model accurately reproduces the real operating state of the physical system.

7. The method for modeling the architecture of an MWh-level aqueous sodium-ion battery energy storage system according to claim 1, characterized in that, The S4 mentioned above includes: collecting sufficient actual operating sample data of four typical operating conditions of energy storage system rated power charging and discharging, partial load, extreme temperature, and battery fault simulation; extracting three core performance indicators: SOC, temperature, and system response time; and verifying whether the model meets the basic error requirements by comparing the actual values ​​of the indicators with the model simulation values ​​to calculate the error.

8. The method for modeling the architecture of an MWh-level aqueous sodium-ion battery energy storage system according to claim 1, characterized in that, In S2, the mechanical-thermal coupling model of the container integrated structure includes the following three-dimensional unsteady heat conduction equation: In the formula: ρ is the density of the container material; c_p is the specific heat capacity of the material at constant pressure; k is the thermal conductivity of the material; q_v is the heat generation per unit volume of the battery array, which is calculated from the battery charging and discharging current and internal resistance; T(x,y,z,t) is the temperature (K) of a certain point (x,y,z) inside the container at time t. The Laplace operator characterizes the temperature gradient distribution; Simultaneously, mechanical constraints are introduced to ensure the stability of the container structure under battery weight and external loads. The constraint equations are as follows: In the formula: σ(x,y,z) is the actual stress at a certain point of the container structure; [σ] is the allowable stress of the material. Through coupling analysis, the accurate simulation of heat conduction characteristics and the dual control of structural load stability are achieved.

9. The method for modeling the architecture of an MWh-level aqueous sodium-ion battery energy storage system according to claim 1, characterized in that, In S2, the electrical system wiring impedance loss model includes: constructing an AC circuit total impedance model, solving for power loss through the real part of the impedance, achieving accurate quantification of energy consumption in electrical wiring links, and the total impedance calculation formula is as follows: In the formula: R_w is the resistance of the conductor, calculated by the resistance law R_w=ρL3 / S2; R_j is the contact resistance of the terminal; ω is the AC angular frequency, ω=2πf, f is the system frequency; L2 is the loop inductance; The formula for calculating the active power loss of a circuit is: In the formula: I is the circuit operating current; Re(Z) is the real part of the total impedance; P_loss is the active power loss of the electrical wiring circuit, providing the core quantitative basis for the overall energy efficiency analysis of the energy storage system.

10. A modeling device for an MWh-level aqueous sodium-ion battery energy storage system architecture, characterized in that, It includes a parameter processing and sensitivity analysis module, a modular modeling module, a system integration modeling module, and a model verification module. These modules work together sequentially to achieve virtual modeling of the energy storage system architecture, as detailed below: The parameter processing and sensitivity analysis module is used to identify the constituent modules of the MWh-level aqueous sodium-ion battery energy storage system. These modules include a container integration module, an electrical wiring module, a battery energy storage unit module, an energy storage converter (PCS) module, and a temperature control and fire-fighting linkage module. Key technical parameters of each constituent module are extracted, a standardized parameter database is established, and sensitivity analysis is used to determine the core parameters affecting system performance. These core parameters include the battery cell internal resistance R, open-circuit voltage U, PCS conversion efficiency η, temperature control threshold T, and fire-fighting response delay time t. The modular modeling module is used to construct a dedicated dynamic model based on the structural characteristics and working principle of each component module. The dedicated dynamic model includes a container integrated structural mechanics-thermal coupling model, an electrical system wiring impedance loss model, a battery energy storage unit series-parallel active balancing model, a PCS bidirectional conversion sliding mode control model, a temperature control fuzzy adaptive PID model, and a fire protection multi-source sensing linkage response model. Each dedicated dynamic model embeds a corresponding quantitative formula to characterize the module's dynamic characteristics. The system integration modeling module is used to build a heterogeneous signal interaction link based on digital twin bus coupling technology, introduce data interaction protocol and real-time mapping function, integrate the dedicated dynamic models built by the sub-module modeling module to form a complete virtual model of energy storage system architecture; realize dynamic collaborative simulation between modules through real-time data mapping, build a two-way interaction channel between "physical entity and virtual model" to ensure accurate synchronization between physical and virtual quantities; The model verification module is used to acquire actual operating data of the energy storage system under multiple operating conditions, and to perform performance verification on the complete energy storage system architecture virtual model constructed by the system integration modeling module based on the actual operating data, and output the model verification results to determine whether the model accuracy meets the preset requirements.

Images