A high-voltage integrated and coordinated control method and system for energy storage systems

Abstract

The present invention discloses a method and system for high-voltage integration and collaborative control of an energy storage system. The method includes: based on the power level and heat distribution requirements of the energy storage system, a topology optimization algorithm is used to perform a three-dimensional spatial coupling design between the high-voltage box and the power conversion system (PCS), generating a physical layer architecture scheme for the high-voltage integrated device; based on the physical layer architecture scheme of the high-voltage integrated device, a collaborative optimization data set is constructed using timestamp alignment technology; the collaborative optimization data set is input into a model predictive control framework to generate dynamically adjusted PCS operating parameters; based on the PCS operating parameters, a three-dimensional thermal field reconstruction model is used to calculate the internal temperature distribution of the high-voltage integrated device in real time, dynamically adjust the cooling fan speed and power device switching frequency, and generate collaborative control instructions. The embodiments of the present invention can improve the integration level and collaborative control capabilities of the high-voltage energy storage system, and achieve efficient and stable operation of the energy storage system.

Description

Technical Field

[0001] The present invention belongs to the field of energy storage technology, and in particular to a high-voltage integrated and coordinated control method and system for an energy storage system. Background Art

[0002] With the rapid development of renewable energy power generation and the increasing demand for flexible regulation capabilities in power systems, energy storage systems, as an important means of achieving energy storage and allocation, have become a key component of modern power systems. In particular, efficient integration and coordinated control on the high-voltage side can not only significantly improve the operating efficiency of energy storage equipment, but also enhance the stability and responsiveness of the power system. Traditional energy storage systems usually adopt a decentralized or isolated design approach, managing high-voltage conversion equipment, cooling systems, and energy storage units as independent subsystems. In practical applications, due to the complex system structure and the coupling of heat dissipation and electrical connections, it is difficult to achieve efficient spatial layout and thermal management, resulting in insufficient energy conversion efficiency, increased equipment heat loss, and increased maintenance difficulties. Summary of the Invention

[0003] The purpose of the present invention is to provide a high-voltage integrated and coordinated control method and system for an energy storage system to address the deficiencies in the prior art, thereby improving the integration level and coordinated control capabilities of the high-voltage energy storage system and achieving efficient and stable operation of the energy storage system.

[0004] An embodiment of the present application provides a high-voltage integrated and coordinated control method for an energy storage system, the method comprising:

[0005] Based on the power level and heat distribution requirements of the energy storage system, a topology optimization algorithm is used to perform a three-dimensional spatial coupling design between the high-voltage box and the power conversion system (PCS), generating a physical layer architecture solution for the high-voltage integrated device. The topology optimization algorithm simultaneously calculates the optimal matching relationship between the electrical connection path and the heat dissipation channel.

[0006] Based on the physical layer architecture solution of the high-voltage integrated device, real-time battery status data from the battery management system (BMS) and grid dispatch instructions from the energy management system (EMS) are synchronously collected through a dynamic data bus, and timestamp alignment technology is used to construct a collaborative optimization data set. The collaborative optimization data set includes multi-dimensional coupling characteristics of battery health, charge and discharge rates, and grid demand response.

[0007] The collaborative optimization dataset is input into a model predictive control framework, and a rolling horizon optimization algorithm is used to solve the charge and discharge strategy and generate dynamically adjusted PCS operating parameters. The rolling horizon optimization algorithm achieves multi-objective collaborative optimization by constraining the battery life decay rate and the power grid power fluctuation range.

[0008] According to the PCS operating parameters, the internal temperature distribution of the high-voltage integrated machine is calculated in real time through a three-dimensional thermal field reconstruction model, the cooling fan speed and the switching frequency of the power device are dynamically adjusted, and coordinated control instructions are generated. At the same time, these instructions are fed back to the BMS and EMS to update the battery charging and discharging thresholds and the grid scheduling strategy, forming a closed-loop control link.

[0009] Optionally, based on the power level and heat distribution requirements of the energy storage system, a topology optimization algorithm is used to perform a three-dimensional spatial coupling design of the high-voltage box and the power conversion system PCS to generate a physical layer architecture solution for the high-voltage integrated device. The topology optimization algorithm simultaneously calculates the optimal matching relationship between the electrical connection path and the heat dissipation channel, including:

[0010] Based on the power level parameters and heat distribution requirements of the energy storage system, an initial 3D model of the high-voltage box and PCS is constructed, generating an initial 3D mesh model including electrical interfaces, heat dissipation vent locations, and internal cavities.

[0011] In the initial 3D mesh model, the electrical connection path is designed based on the minimum path principle. The busbar routing length and impedance are iteratively optimized using the ant colony algorithm to generate a candidate set of electrical connection paths and mark the coordinates of key nodes.

[0012] Simultaneously perform fluid dynamics simulation on the heat dissipation channel, optimize the heat dissipation fin layout based on the heat source distribution density, calculate the heat dissipation efficiency curve under different wind speeds, and generate the heat dissipation channel topology and fan deployment plan;

[0013] Input the candidate set of electrical connection paths and the heat dissipation channel topology into the multi-objective optimization algorithm. Using the electrical loss rate and heat dissipation efficiency as constraints, the Pareto optimal solution set is solved to generate the electrical-heat dissipation coupled topology solution.

[0014] Conduct electromagnetic-thermomechanical coupling simulation verification on the topology solution, detect local hotspots and areas with excessive electromagnetic interference, correct the connection paths and heat dissipation channel spacing, and output the final three-dimensional solution for the physical layer architecture of the high-voltage integrated machine.

[0015] Optionally, the physical layer architecture based on the high-voltage integrated device synchronously collects real-time battery status data of the battery management system BMS and grid dispatch instructions of the energy management system EMS through a dynamic data bus, and uses timestamp alignment technology to construct a collaborative optimization data set, wherein the collaborative optimization data set contains multi-dimensional coupling characteristics of battery health, charge and discharge rate, and grid demand response, including:

[0016] The battery status data of the BMS and the grid dispatch instructions of the EMS are collected through the dynamic data bus, and the data timestamps are aligned using hardware clock synchronization technology to generate a time-synchronized original data stream;

[0017] Clean outliers from the original data stream, compensate for missing data based on the battery cluster topology, and use the covariance matrix to eliminate sensor noise interference to generate a purified multidimensional monitoring data set;

[0018] Extract battery health characteristics, charge and discharge rate characteristics, and grid demand response characteristics from the multidimensional monitoring data set to construct a multidimensional coupling feature correlation matrix;

[0019] Based on the multi-dimensional coupling feature correlation matrix, the dynamic correlation between battery health, charge and discharge rate, and grid demand response is analyzed using the Pearson correlation coefficient to generate a multi-dimensional coupling feature mapping table containing weight labels.

[0020] The multi-dimensional coupling feature map is fused with the time series data, packaged into a collaborative optimization dataset with spatiotemporal labels according to the preset sampling period, and stored in the shared memory area of the edge computing node.

[0021] Optionally, the collaborative optimization data set is input into a model predictive control framework, and a rolling horizon optimization algorithm is used to solve the charging and discharging strategy to generate dynamically adjusted PCS operating parameters, wherein the rolling horizon optimization algorithm realizes multi-objective collaborative optimization by constraining the battery life attenuation rate and the grid power fluctuation range, including:

[0022] The collaborative optimization dataset is fed into the model predictive control framework, which predicts the power demand curve for a preset number of control cycles in the future based on the current battery status and grid instructions, and generates a reference trajectory for charge and discharge power.

[0023] A multi-objective optimization model was established, with reference to the charge and discharge power benchmark reference trajectory, battery life decay rate as the first constraint and grid power fluctuation range as the second constraint, and the objective function was defined as maximizing system efficiency.

[0024] ‌A rolling time domain optimization algorithm is used to solve the objective function. The optimal charge and discharge current values, PCS switching frequency, and power factor compensation parameters are iteratively calculated within each control cycle to generate a candidate set of dynamic PCS operating parameters.

[0025] Perform feasibility verification on candidate parameters in the dynamic PCS operating parameter candidate set, detect whether the parameters exceed the safety limit through the battery polarization voltage model and the grid harmonic analysis model, and output the verified dynamic adjustment PCS operating parameter instruction set.

[0026] Optionally, based on the PCS operating parameters, the internal temperature distribution of the high-voltage integrated device is calculated in real time through a three-dimensional thermal field reconstruction model, the cooling fan speed and the power device switching frequency are dynamically adjusted, and coordinated control instructions are generated. At the same time, these instructions are fed back to the BMS and EMS to update the battery charge and discharge thresholds and the grid scheduling strategy, forming a closed-loop control link, including:

[0027] Based on the PCS operating parameters, the finite element method is used to calculate the power device loss in real time, and a three-dimensional thermal field distribution model inside the high-voltage integrated machine is constructed using fluid mechanics equations.

[0028] Identify areas with excessive temperatures in the three-dimensional thermal field distribution model, dynamically adjust the speed of the cooling fan and the switching frequency of the power devices, and generate a draft of the cooling control instructions;

[0029] The draft heat dissipation control instructions are input into the digital twin system for simulation and execution, and the temperature change trend after adjustment is predicted. If the simulation results do not meet expectations, the instructions are re-planned until the thermal management requirements are met, and the final coordinated control instructions are obtained;

[0030] The final coordinated control instructions are sent to the PCS and cooling system for execution. At the same time, the temperature distribution data is fed back to the BMS to update the battery charge and discharge thresholds, and grid dispatch strategy correction suggestions are sent to the EMS to form a closed-loop control link.

[0031] Another embodiment of the present application provides a high-voltage integrated and coordinated control system for an energy storage system, the system comprising:

[0032] A design module is used to perform a three-dimensional spatial coupling design between the high-voltage box and the power conversion system (PCS) based on the power level and heat distribution requirements of the energy storage system using a topology optimization algorithm to generate a physical layer architecture solution for the high-voltage integrated device. The topology optimization algorithm simultaneously calculates the optimal matching relationship between the electrical connection path and the heat dissipation channel;

[0033] A construction module is used to synchronously collect real-time battery status data of the battery management system (BMS) and grid dispatch instructions of the energy management system (EMS) through a dynamic data bus based on the physical layer architecture solution of the high-voltage integrated device, and use timestamp alignment technology to construct a collaborative optimization data set, wherein the collaborative optimization data set includes multi-dimensional coupling characteristics of battery health, charge and discharge rates, and grid demand response;

[0034] a generation module, configured to input the collaborative optimization dataset into a model predictive control framework, solve the charge and discharge strategy using a rolling horizon optimization algorithm, and generate dynamically adjusted PCS operating parameters, wherein the rolling horizon optimization algorithm achieves multi-objective collaborative optimization by constraining the battery life decay rate and the grid power fluctuation range;

[0035] The control module is used to calculate the internal temperature distribution of the high-voltage integrated machine in real time through a three-dimensional thermal field reconstruction model based on the PCS operating parameters, dynamically adjust the cooling fan speed and the switching frequency of the power device, generate coordinated control instructions, and simultaneously feed back to the BMS and EMS to update the battery charging and discharging thresholds and the power grid scheduling strategy, forming a closed-loop control link.

[0036] Yet another embodiment of the present application provides a storage medium, wherein the storage medium stores a computer program, wherein the computer program is configured to execute any of the above methods when run.

[0037] Yet another embodiment of the present application provides an electronic device, comprising a memory and a processor, wherein the memory stores a computer program, and the processor is configured to run the computer program to execute any of the above methods.

[0038] Compared with the existing technology, the present invention provides a high-voltage integration and collaborative control method for an energy storage system. According to the power level and heat distribution requirements of the energy storage system, the high-voltage box and the power conversion system PCS are designed to be coupled in three dimensions through a topology optimization algorithm to generate a physical layer architecture scheme for the high-voltage integrated machine; based on the physical layer architecture scheme of the high-voltage integrated machine, a collaborative optimization data set is constructed using timestamp alignment technology; the collaborative optimization data set is input into the model predictive control framework to generate dynamically adjusted PCS operating parameters; according to the PCS operating parameters, the internal temperature distribution of the high-voltage integrated machine is calculated in real time through a three-dimensional thermal field reconstruction model, the cooling fan speed and the power device switching frequency are dynamically adjusted, and collaborative control instructions are generated, thereby improving the integration level and collaborative control capability of the high-voltage energy storage system and realizing efficient and stable operation of the energy storage system. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] Figure 1 A hardware structure block diagram of a computer terminal for a high-voltage integrated and coordinated control method for an energy storage system provided by an embodiment of the present invention;

[0040] Figure 2 A schematic flow chart of a high-voltage integrated and coordinated control method for an energy storage system provided by an embodiment of the present invention;

[0041] Figure 3 A schematic structural diagram of a high-voltage integrated and coordinated control system for an energy storage system provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0042] The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention, and are not to be construed as limiting the present invention.

[0043] The embodiment of the present invention first provides a high-voltage integrated and coordinated control method for an energy storage system. The method can be applied to electronic devices, such as computer terminals, specifically ordinary computers.

[0044] The following describes it in detail by taking running on a computer terminal as an example. Figure 1The hardware structure block diagram of a computer terminal for a high-voltage integrated and coordinated control method of an energy storage system provided by an embodiment of the present invention. Figure 1 As shown, the computer device includes a processor, a memory, and a network interface connected via a system bus, wherein the memory may include a non-volatile storage medium and an internal memory.

[0045] The non-volatile storage medium can store an operating system and a computer program. The computer program includes program instructions that, when executed, enable a processor to execute any one of the high-voltage integrated and coordinated control methods for an energy storage system.

[0046] The processor is used to provide computing and control capabilities and support the operation of the entire computer equipment.

[0047] The internal memory provides an environment for the operation of the computer program in the non-volatile storage medium. When the computer program is executed by the processor, the processor can execute any one of the high-voltage integration and coordinated control methods of the energy storage system.

[0048] The network interface is used for network communication, such as sending assigned tasks, etc. Those skilled in the art will understand that Figure 1 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than shown in the figure, or combine certain components, or have a different component arrangement.

[0049] It should be understood that the processor may be a central processing unit (CPU), other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA), other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or any conventional processor, etc.

[0050] See also Figure 2 , an embodiment of the present invention provides a high-voltage integrated and coordinated control method for an energy storage system, which may include the following steps:

[0051] S201: Based on the power level and heat distribution requirements of the energy storage system, a topology optimization algorithm is used to perform a three-dimensional spatial coupling design of the high-voltage box and the power conversion system (PCS) to generate a physical layer architecture solution for the high-voltage integrated device. The topology optimization algorithm simultaneously calculates the optimal matching relationship between the electrical connection path and the heat dissipation channel. Specifically, the method may include:

[0052] Based on the power level parameters and heat distribution requirements of the energy storage system, an initial 3D model of the high-voltage box and PCS is constructed, generating an initial 3D mesh model including electrical interfaces, heat dissipation vent locations, and internal cavities.

[0053] Analysis of Power Rating and Thermal Distribution Requirements: Energy storage system power rating parameters include rated voltage (e.g., 1500V), maximum charge / discharge current (e.g., 300A), and peak power (e.g., 1MW). Thermal distribution requirements are determined by the battery cluster layout and the thermal dissipation characteristics of PCS power devices (e.g., IGBT modules). For example, the internal temperature rise of the high-voltage box must not exceed 40°C (maximum temperature 65°C at an ambient temperature of 25°C).

[0054] Heat source location: Use thermal imaging data to mark the main heating areas, such as the PCS DC / AC module (coordinates x=1.2m, y=0.8m) and the high-voltage box busbar connection point (coordinates x=0.5m, y=1.0m).

[0055] ‌3D Modeling Tools‌: Use SolidWorks or ANSYS SpaceClaim to build the geometric models of the high-voltage box and PCS, defining the electrical interfaces (such as DC input terminals and AC output terminals) and the location of the heat dissipation vents (200 mm × 100 mm, located on the top of the box).

[0056] ‌Initial 3D mesh model generation‌

[0057] Meshing: The model is discretized using tetrahedral elements. The mesh size is set to 5mm (for critical areas such as heat dissipation vents) to 20mm (for non-critical areas) based on accuracy requirements. For example, the internal cavity of the high-voltage box is divided into 500,000 mesh nodes, and the PCS module is divided into 300,000 nodes.

[0058] Mesh Optimization: Use the Mesh Quality Check function to detect deformed meshes (such as elements with aspect ratios > 10) and use the Laplacian smoothing algorithm to optimize mesh quality and ensure computational convergence.

[0059] In the initial 3D mesh model, the electrical connection path is designed based on the minimum path principle. The busbar routing length and impedance are iteratively optimized using the ant colony algorithm to generate a candidate set of electrical connection paths and mark the coordinates of key nodes.

[0060] Minimum Path Principle and Ant Colony Algorithm Parameter Setting

[0061] Path constraints: Busbar routing must meet safety spacing (e.g., spacing from heat dissipation channels ≥ 10 mm), bend radius (≥ 5 times the busbar thickness), and impedance limit (≤ 0.1 mΩ / m).

[0062] Ant Colony Algorithm Parameters: Set the pheromone evaporation rate to 0.3, the heuristic factor α=1, β=2, and the number of iterations to 500.

[0063] Busbar routing optimization process

[0064] Initial Path Generation: Ten initial paths are randomly generated from the DC input terminal of the high-voltage box (starting coordinates x=0, y=0) to the PCS module (ending coordinates x=1.2m, y=0.8m). The length range is 1.5m-2.0m, and the impedance range is 0.15mΩ-0.25mΩ.

[0065] Iterative Optimization: Each ant selects the next node based on pheromone concentration and path length. For example, after the 100th iteration, the optimal path length was shortened to 1.2m and the impedance dropped to 0.08mΩ.

[0066] Candidate set generation: retain the top five optimal paths (e.g., lengths of 1.2m, 1.3m, and 1.4m, impedances of 0.08mΩ, 0.09mΩ, and 0.10mΩ) and mark key nodes (e.g., inflection point coordinates x=0.8m, y=0.6m).

[0067] Simultaneously perform fluid dynamics simulation on the heat dissipation channel, optimize the heat dissipation fin layout based on the heat source distribution density, calculate the heat dissipation efficiency curve under different wind speeds, and generate the heat dissipation channel topology and fan deployment plan;

[0068] Fluid mechanics simulation and heat sink fin optimization

[0069] Heat source distribution density calculation: Determine the heat sink fin layout density based on the heat dissipation data of the 3D mesh model (e.g., the heat flux density of the PCS module is 5000W / m², and the heat flux density of the high-voltage box busbar connection point is 3000W / m²).

[0070] Fin parameters: Fin height 30mm, thickness 2mm, spacing 10mm, material aluminum alloy (thermal conductivity 237W / m·K). ANSYS Fluent simulations were used to compare the heat dissipation efficiency of different layouts (e.g., parallel and staggered).

[0071] Wind speed-heat dissipation efficiency curve: At a wind speed of 2 m / s, the heat dissipation efficiency is 60% (temperature rise reduced by 24°C), increasing to 75% (temperature rise reduced by 30°C) at a wind speed of 4 m / s. After the wind speed exceeds 6 m / s, the efficiency growth slows down (diminishing marginal benefits).

[0072] Wind turbine deployment plan design

[0073] Fan selection: Use axial fans (such as the EBM-papst model A3G450, with an air volume of 200 CFM and a speed of 2500 RPM). Determine the number of fans to deploy based on the heat dissipation efficiency curve (for example, four fans, one on the top and two on each side of the cabinet).

[0074] Topology output: The heat dissipation channel width is set to 50mm, and the fan installation coordinates (e.g., the top fan coordinates are x=0.5m, y=1.5m, and the side fan coordinates are x=0.2m, y=0.5m and x=1.8m, y=0.5m).

[0075] Input the candidate set of electrical connection paths and the heat dissipation channel topology into the multi-objective optimization algorithm. Using the electrical loss rate and heat dissipation efficiency as constraints, the Pareto optimal solution set is solved to generate the electrical-heat dissipation coupled topology solution.

[0076] Multi-objective optimization model construction

[0077] Objective function: Minimize electrical loss (target value ≤ 0.1mΩ) and maximize heat dissipation efficiency (target value ≥ 70%).

[0078] Constraints: Busbar path length ≤ 1.5m, heat dissipation channel flow velocity ≥ 3m / s, fan power consumption ≤ 200W.

[0079] NSGA-II algorithm for finding Pareto optimal solutions

[0080] Population initialization: Generate 100 individuals, each with a busbar path and a heat dissipation topology. For example, individual 1 has a path length of 1.2m and a heat dissipation efficiency of 75%, while individual 2 has a path length of 1.3m and a heat dissipation efficiency of 80%.

[0081] Iterative Evolution: After 200 generations of evolution, the Pareto Front contains 20 non-dominated solutions, of which the optimal solution has a path length of 1.25m, an impedance of 0.09mΩ, and a heat dissipation efficiency of 78%.

[0082] Solution selection: Based on project requirements (such as cost constraints), the solution that prioritizes heat dissipation efficiency is selected: path length 1.3m, impedance 0.10mΩ, heat dissipation efficiency 80%, and the number of fans is reduced to 3.

[0083] Conduct electromagnetic-thermomechanical coupling simulation verification on the topology solution, detect local hotspots and areas with excessive electromagnetic interference, correct the connection paths and heat dissipation channel spacing, and output the final three-dimensional solution for the physical layer architecture of the high-voltage integrated machine.

[0084] ‌Electromagnetic-thermomechanical coupled simulation execution‌

[0085] Simulation tool: Use COMSOL Multiphysics to load the 3D model of the topology solution and set boundary conditions (such as ambient temperature 25°C and bus current 300A).

[0086] Hotspot Detection: Simulation results show that the temperature below the PCS module (coordinates x=1.2m, y=0.8m, z=0.5m) reached 68°C (exceeding the threshold of 65°C). The magnetic field strength at the busbar inflection point (coordinates x=0.8m, y=0.6m) exceeded the limit (>50μT).

[0087] ‌Scheme revision and verification‌

[0088] Spacing adjustment: Increase the spacing between the busbar and the heat dissipation channel from 10mm to 15mm, and add an electromagnetic shielding layer (MuMetal material, 1mm thickness) at the inflection point of the busbar.

[0089] Secondary simulation: After adjustment, the maximum temperature dropped to 63°C and the magnetic field strength dropped to 45μT, meeting the design requirements.

[0090] Final Output: Generates a 3D model file in STEP format, including the corrected geometry, material properties, and performance parameters (e.g., electrical loss rate of 0.10 mΩ, heat dissipation efficiency of 80%).

[0091] ‌Technical Details and Parameter Examples‌

[0092] ‌3D Mesh Model Parameters‌

[0093] Total number of grids: 800,000 (high-voltage box) + 500,000 (PCS) = 1.3 million cells;

[0094] Minimum grid size: 5mm (for busbar connection points and heat sinks);

[0095] Computing resources: The simulation requires 64 GB of memory and takes 6 hours (based on an Intel Xeon 16-core processor).

[0096] Ant Colony Algorithm Optimization Effect

[0097] The average length of the initial path is 1.8m, and the shortest path after optimization is 1.2m.

[0098] Impedance reduction: Optimized from 0.25mΩ to 0.08mΩ (68% reduction).

[0099] ‌Relationship between heat dissipation efficiency and wind speed‌

[0100] Heat dissipation power at wind speed 2m / s: 5000W;

[0101] Heat dissipation power at wind speed 4m / s: 8000W;

[0102] Critical wind speed: 6m / s (further increasing the wind speed only increases the heat dissipation power to 8200W).

[0103] Comparison of multi-objective optimization results

[0104] Pareto solution 1: loss rate 0.09 mΩ, heat dissipation efficiency 78%, cost $1200;

[0105] Pareto solution 2: loss rate 0.10mΩ, heat dissipation efficiency 80%, cost $1000;

[0106] The final solution selected is Solution 2 (meets the cost constraint).

[0107] ‌Application Scenario Examples‌

[0108] Scenario Description: Design of a high-voltage integrated device for a 1500V / 1MW energy storage system.

[0109] ‌3D modeling‌: High-voltage box dimensions: 2m×1m×1.5m, PCS module dimensions: 0.8m×0.6m×0.4m;

[0110] Electrical path optimization: busbar length optimized from 1.8m to 1.3m, impedance 0.10mΩ;

[0111] Heat dissipation design: 3 axial flow fans are deployed, with a heat dissipation efficiency of 80%;

[0112] ‌Coupled simulation verification‌: Maximum temperature 63°C, magnetic field strength 45μT;

[0113] Final Solution: Submit 3D model files and performance reports, and pass customer acceptance testing.

[0114] This step uses a topology optimization algorithm to perform a three-dimensional spatial coupling design of the high-voltage box and the power conversion system (PCS), based on the energy storage system's power level (e.g., hundreds of kilowatts or megawatts) and thermal distribution requirements (e.g., heat density of power devices). Specifically, this involves optimizing the busbar routing using an ant colony algorithm to reduce electrical impedance, optimizing the heat sink fin layout in conjunction with computational fluid dynamics (CFD) simulation to improve heat dissipation efficiency, and balancing electrical losses and heat dissipation performance through multi-objective optimization algorithms (e.g., Pareto optimality). Ultimately, a three-dimensional physical architecture solution for the high-voltage integrated unit is generated, ensuring that electrical connection paths and heat dissipation channels are spatially conflict-free and maximized in efficiency. This addresses the volume redundancy and inefficiency issues caused by the separation of electrical and heat dissipation design in traditional energy storage systems, providing a compact, integrated hardware foundation for high-power density energy storage systems.

[0115] S202, based on the physical layer architecture solution of the high-voltage integrated device, synchronously collect real-time battery status data from the battery management system (BMS) and grid dispatch instructions from the energy management system (EMS) via a dynamic data bus, and use timestamp alignment technology to construct a collaborative optimization data set, wherein the collaborative optimization data set includes multi-dimensional coupling characteristics of battery health, charge and discharge rates, and grid demand response; specifically, it may include:

[0116] The battery status data of the BMS and the grid dispatch instructions of the EMS are collected through the dynamic data bus, and the data timestamps are aligned using hardware clock synchronization technology to generate a time-synchronized original data stream;

[0117] Dynamic Data Bus Selection and Configuration

[0118] The dynamic data bus uses the EtherCAT (Ethernet Control Automation Technology) or CAN FD (Controller Area Network Flexible Data-rate) protocols, supporting high-speed (≥100Mbps) and low-latency (<1ms) data transmission. For example, the BMS transmits data such as battery cell voltage (accuracy ±1mV), temperature (accuracy ±0.5°C), and SOC (State of Charge, accuracy ±2%) via the CAN FD bus at a 1kHz frequency. The EMS also transmits grid dispatch commands (such as charge and discharge power commands, frequency modulation signals, and peak and valley electricity price information) via the EtherCAT bus.

[0119] Bus topology design: A star topology is used, with edge computing nodes (such as NVIDIA Jetson AGXXavier) as the master station, and BMS and EMS as slave stations. The master station polling period is set to 2ms to ensure real-time data.

[0120] Hardware Clock Synchronization: Based on IEEE 1588 PTP (Precision Time Protocol), the master and slave stations deploy hardware clock synchronization modules (such as the Texas Instruments DP83640 chip) to calibrate timestamp accuracy to 0.1ms. For example, the time deviation between the BMS and EMS is reduced from the initial 5ms to 0.2ms.

[0121] Timestamp alignment and data stream generation

[0122] Timestamp: Each data packet is embedded with a local clock timestamp (e.g., Unix timestamp format: 1630453200.123456) at the sending end, and the master station clock is synchronized to the slave station through the PTP protocol.

[0123] Alignment Algorithm: Interpolation is used to compensate for clock deviation. For example, the voltage data sent by the BMS at timestamp t=1000ms and the scheduling instruction sent by the EMS at timestamp t=1000.2ms are linearly interpolated to generate the aligned data point t=1000.1ms.

[0124] ‌Raw Data Stream Format‌: Generates a structured data stream containing aligned timestamps (Timestamp), BMS data (Voltage, Temperature, SOC) and EMS commands (Power_Setpoint, Frequency_Command), stored in CSV or binary format.

[0125] Clean outliers from the original data stream, compensate for missing data based on the battery cluster topology, and use the covariance matrix to eliminate sensor noise interference to generate a purified multidimensional monitoring data set;

[0126] Outlier detection and cleaning

[0127] Rule-based filtering: Define threshold rules (e.g., cell voltage <2.5V or >4.2V is abnormal, temperature <-20°C or >60°C is abnormal), and automatically remove data points that are out of range.

[0128] Statistical Method: Use the Z-score algorithm to detect outliers. For example, calculate the mean and standard deviation of voltage data (e.g., μ = 3.6V, σ = 0.1V). A Z-score greater than 3 (i.e., voltage greater than 3.9V or less than 3.3V) is considered an outlier.

[0129] Sliding window repair: For continuously abnormal data (such as the abnormal voltage of a battery cell for three consecutive cycles), the median value of the adjacent window (window size = 10 sampling points) is used to replace it.

[0130] Missing Data Compensation

[0131] Battery cluster topology modeling: Based on the series-parallel structure of the battery cluster (e.g., 16S48P: 16 series modules, each module with 48 parallel cells), an electrical relationship model between cells is established.

[0132] Compensation strategy: If a cell's data is missing (for example, the voltage of ID 5 is missing), the data is filled in using the average value of other cells in the same module (for example, the average voltage of all 47 cells in the module is 3.65V).

[0133] Time series interpolation: For data with short-term missing (e.g., <5 sampling points), cubic spline interpolation is used to complete the missing data.

[0134] Sensor Noise Suppression

[0135] ‌Covariance Matrix Construction‌: Assuming that the noise follows a Gaussian distribution, construct the covariance matrix of voltage, temperature, and SOC. For example:

[0136] Cov(Voltage, Temperature) = -0.05 (negative correlation: increasing temperature causes decreasing voltage);

[0137] Cov(SOC, Voltage) = 0.8 (strong positive correlation: higher SOC means higher voltage).

[0138] Kalman filter noise reduction: A Kalman filter is designed based on the covariance matrix to predict the true state and correct for noise. For example, a voltage measurement of 3.7V ± 0.1V will be filtered to output 3.68V ± 0.03V.

[0139] Extract battery health characteristics, charge and discharge rate characteristics, and grid demand response characteristics from the multidimensional monitoring data set to construct a multidimensional coupling feature correlation matrix;

[0140] Battery State of Health (SOH) Feature Extraction

[0141] Capacity decay rate calculation: Based on historical charge and discharge data (e.g., capacity drops from 100Ah to 85Ah after 1000 cycles), calculate SOH = current capacity / initial capacity × 100% (SOH = 85%).

[0142] Internal resistance growth characteristics: The internal resistance value is obtained through the HPPC (hybrid pulse power characteristic) test (for example, the initial internal resistance R0 = 1mΩ, the current internal resistance R = 1.5mΩ). The internal resistance growth rate ΔR = (R-R0) / R0 × 100% = 50%.

[0143] Capacity-internal resistance coupling index: Define the SOH comprehensive score as 0.7 × capacity decay rate + 0.3 × internal resistance growth rate (weights are calibrated based on experiments). For example, the SOH score = 0.7 × 85% + 0.3 × 50% = 74.5%.

[0144] ‌Charge and discharge rate (C-rate) feature extraction‌

[0145] Real-time C-rate calculation: C-rate = charge / discharge current (A) / rated capacity (Ah). For example, if the current is 300A and the rated capacity is 100Ah, the C-rate = 3C.

[0146] ‌C-rate change rate‌: Calculate the C-rate change gradient (ΔC-rate / Δt) in adjacent time windows (e.g., 1 second). For example, if the temperature rises from 2C to 3C, the gradient is 1C / s.

[0147] Voltage Slope Characteristics: During constant current charge and discharge, the voltage change rate (dV / dt) reflects the battery's polarization state. For example, during charging, dV / dt = 0.05V / s, and during discharging, dV / dt = -0.08V / s.

[0148] Grid demand response feature extraction

[0149] Power Deviation: Calculates the difference between the EMS command power and the actual power (ΔP = Setpoint - Actual). For example, ΔP = 50kW means the discharge power needs to be increased by 50kW.

[0150] Frequency modulation response delay: This counts the delay from when the command is issued to when the power adjustment is completed (e.g., average delay = 200ms).

[0151] Economic efficiency: Based on the time-of-use electricity price (e.g., 1.2 yuan / kWh during peak hours and 0.5 yuan / kWh during off-peak hours), the difference in charging and discharging costs and benefits is calculated (e.g., discharging benefit = 1.2 × 500kWh = 600 yuan, charging cost = 0.5 × 500kWh = 250 yuan, net benefit = 350 yuan).

[0152] ‌Construction of multidimensional coupling feature correlation matrix‌

[0153] Matrix dimension design: Rows represent time series samples (e.g., one sample per second), and columns include 10 features, including capacity decay rate, internal resistance growth, SOH score, C-rate, C-rate change rate, voltage slope, ΔP, frequency modulation response delay, and economic indicators.

[0154] Normalization: Min-Max normalization is performed on the feature values. For example, the SOH score range [60%, 100%] is mapped to [0, 1], and the C-rate range [0C, 5C] is mapped to [0, 1].

[0155] Matrix storage structure: Numpy array or Pandas DataFrame is used for storage. For example, the matrix size of a 1-hour dataset (3600 samples) is 3600×10.

[0156] Based on the multi-dimensional coupling feature correlation matrix, the dynamic correlation between battery health, charge and discharge rate, and grid demand response is analyzed using the Pearson correlation coefficient to generate a multi-dimensional coupling feature mapping table containing weight labels.

[0157] Pearson correlation coefficient calculation

[0158] Calculation steps: For each feature pair (e.g., SOH score and C-rate), calculate the covariance divided by the product of the standard deviations. For example, r(SOH, C-rate) = Cov(SOH, C-rate) / (σ_SOH × σ_C-rate). Assuming Cov = -0.15, σ_SOH = 0.2, and σ_C-rate = 0.3, then r = -0.15 / (0.2×0.3) = -2.5 (correlation outside the [-1, 1] range is necessary; the actual value should be -0.25).

[0159] Dynamic Window Analysis: A sliding window (window size = 300 samples, i.e., 5 minutes of data) is used to calculate real-time correlation coefficients. For example, the correlation coefficient between SOH and C-rate at t = 10:00 is -0.6, indicating that a high C-rate accelerates SOH decay.

[0160] Weighted label assignment‌

[0161] ‌Correlation Coefficient Grading‌: Defines the strength of association based on the |r| value:

[0162] |r| ≥ 0.8: strong association (weight = 1.0); 0.5 ≤ |r| < 0.8: moderate association (weight = 0.7); |r| < 0.5: weak association (weight = 0.3).

[0163] Dynamic weight adjustment: For example, the correlation coefficient between SOH and grid power deviation ΔP has a weight of 0.7 during peak hours (r=0.7) and a weight of 0.3 during off-peak hours (r=0.4).

[0164] ‌Multi-dimensional coupling feature map generation‌

[0165] ‌Table structure‌: Contains fields such as feature pair name, correlation coefficient, weight, timestamp, etc. For example, Table 1:

[0166]

[0167] Real-time update mechanism: The mapping table is updated every 5 minutes (i.e., the window slides once) and stored in the Redis cache database.

[0168] The multi-dimensional coupling feature map is fused with the time series data, packaged into a collaborative optimization dataset with spatiotemporal labels according to the preset sampling period, and stored in the shared memory area of the edge computing node.

[0169] ‌Spatiotemporal Tag Definition‌

[0170] Time stamp: Uses the ISO 8601 standard (e.g., 2023-10-01T10:00:00.000Z), accurate to milliseconds.

[0171] Spatial tag: Coded according to the energy storage system deployment location, such as site ID (Site_ID=CN_BJ_001) and battery cluster ID (Cluster_ID=BC_01).

[0172] Data fusion and packaging

[0173] Sampling period setting: The preset sampling period is 1 second, and the multi-dimensional coupling feature map is merged with the original time series data (such as voltage and temperature) every second.

[0174] Data format conversion: Converts data to the Parquet columnar storage format, supporting efficient compression (compression ratio ≥ 5:1) and fast queries.

[0175] ‌Metadata addition‌: Add fields such as data version number (Version=1.2), checksum (CRC32), etc.

[0176] ‌Store to shared memory area‌

[0177] Edge computing node configuration: Use a multi-core ARM processor (such as the NXP Layerscape LX2160A) and allocate a shared memory area size of 4GB.

[0178] In-memory database selection: Use Apache Arrow or Redis as an in-memory database, supporting high-speed read and write (throughput ≥ 100,000 records / second).

[0179] Data partitioning strategy: Partition storage by time range (e.g., one partition per hour) and spatial label (site ID) to facilitate parallel access.

[0180] ‌Application Scenario Examples‌

[0181] Scenario description: Collaborative optimization of the BMS and EMS of a 100MWh energy storage power station.

[0182] Data Collection: The BMS uploads voltage and temperature data of 192 battery cells at a frequency of 1kHz, and the EMS issues grid dispatch instructions once a second.

[0183] Data cleaning: Repaired five abnormal voltage points (3.8V→3.65V) and compensated for two missing temperature data points (using the average value of 25°C for the same module).

[0184] Feature extraction: Calculates the SOH score (82%), C-rate (2.5C), and power deviation (ΔP = 30kW).

[0185] Correlation analysis: We found a strong negative correlation between SOH and C-rate (r=-0.7), triggering an optimization strategy to reduce C-rate.

[0186] Dataset storage: Data is packaged in 1-second cycles and stored in edge node memory for real-time control algorithm calls.

[0187] Using a dynamic data bus (such as CAN or EtherCAT), the system collects real-time data on battery cell voltage, temperature, and SOC from the BMS, as well as grid frequency and peak-shaving instructions from the EMS. Hardware clock synchronization is used to align millisecond timestamps, and sensor noise interference is eliminated using a covariance matrix. Features of battery health (such as capacity decay), charge and discharge rates (such as C-rate fluctuations), and grid demand response (such as frequency modulation power change rate) are extracted to construct a multidimensional coupled feature matrix, forming a collaborative optimization dataset with spatiotemporal labels. This breaks down data silos between the BMS and EMS, enabling millisecond-level cross-system data synchronization and feature fusion. This provides high-precision input for subsequent model predictive control, preventing inaccurate control instructions caused by data delays or noise.

[0188] S203: Input the collaborative optimization data set into a model predictive control framework, use a rolling horizon optimization algorithm to solve the charging and discharging strategy, and generate dynamically adjusted PCS operating parameters, wherein the rolling horizon optimization algorithm achieves multi-objective collaborative optimization by constraining the battery life attenuation rate and the power grid power fluctuation range; specifically, it may include:

[0189] The collaborative optimization dataset is fed into the model predictive control framework, which predicts the power demand curve for a preset number of control cycles in the future based on the current battery status and grid instructions, and generates a reference trajectory for charge and discharge power.

[0190] ‌Model Predictive Control (MPC) Framework Construction‌

[0191] Input Data Preprocessing: The collaborative optimization dataset includes features such as battery health (SOH), charge / discharge rate (C-rate), and grid power deviation (ΔP). For example, SOH = 85%, C-rate = 2.5C, and ΔP = 30kW. Edge computing nodes (such as Intel Xeon D-2145NT processors) normalize the data (Min-Max normalization to the range [0, 1]) and divide it into a training set (80%) and a test set (20%).

[0192] Forecasting Model Selection: Use a Long Short-Term Memory (LSTM) or Autoregressive Integrated Moving Average (ARIMA) model to forecast future power demand. For example, the LSTM network configuration uses three hidden layers (128 neurons per layer), a time window length of 10 control cycles (5 seconds per cycle), and a prediction step length of 3 cycles (15 seconds).

[0193] Power Demand Curve Generation: Based on the current battery SOC (e.g., 60%) and EMS dispatch instructions (e.g., a discharge power demand of 500kW), the system predicts power changes every 5 seconds for the next 15 seconds. For example, the prediction result shows that the discharge power needs to increase to 520kW at t=5s and decrease to 490kW at t=10s, forming a reference trajectory.

[0194] Dynamic Baseline Adjustment Mechanism

[0195] Real-time feedback correction: At the end of each control cycle (e.g., 5 seconds), the actual power is compared with the predicted value, and the residual error (Residual Error) is calculated. If the residual error exceeds a threshold (e.g., ±5%), an online update of the model parameters is triggered. For example, if the actual discharge power is 515kW (predicted value 520kW), the residual error = -5kW (-1%), and no correction is triggered. If the residual error reaches -25kW (-5%), the LSTM weights are adjusted using the Recursive Least Squares (RLS) method.

[0196] Multi-scenario prediction fusion: For different scenarios, such as frequency regulation and peak shaving, multiple reference trajectories are generated (e.g., frequency regulation requires rapid power fluctuations, while peak shaving requires smooth changes). The optimal trajectory is selected through weighted voting. For example, the frequency regulation scenario has a weight of 0.7, while the peak shaving scenario has a weight of 0.3. The final trajectory is a linear combination of the two.

[0197] A multi-objective optimization model was established, with reference to the charge and discharge power benchmark reference trajectory, battery life decay rate as the first constraint and grid power fluctuation range as the second constraint, and the objective function was defined as maximizing system efficiency.

[0198] Multi-objective optimization model definition

[0199] ‌Objective function design‌: System efficiency = charging and discharging efficiency × grid response efficiency × battery health factor. For example:

[0200] Charge and discharge efficiency = actual output energy / battery released energy (typical value 95%);

[0201] Grid response efficiency = 1 - |ΔP| / Setpoint (ΔP = actual power deviation);

[0202] Battery health factor = 1 - 0.1×(1-SOH) (factor = 0.985 when SOH = 85%).

[0203] Objective function = 0.95 × 0.98 × 0.985 ≈ 0.913, which needs to be maximized.

[0204] ‌Constraint Quantification‌:

[0205] Battery life decay rate: Limit the capacity decay within each control cycle to ≤ 0.01% (corresponding to a 0.01% decrease in SOH);

[0206] Grid power fluctuation: limit |ΔP|≤5% Setpoint (e.g., when Setpoint=500kW, ΔP≤25kW);

[0207] PCS hardware limitations: Charge and discharge current ≤ rated value (e.g., 1000A), switching frequency ≤ 20kHz.

[0208] ‌Multi-objective optimization algorithm implementation‌

[0209] Algorithm selection: Use the Non-Dominated Sorting Genetic Algorithm (NSGA-II) or Particle Swarm Optimization (PSO). For example, NSGA-II is configured with a population size of 50, a crossover probability of 0.8, a mutation probability of 0.1, and 100 iterations.

[0210] Constraint handling strategy: The constraints are converted into penalty terms (Penalty Function). For example, if the capacity decay rate exceeds the standard, the objective function value is multiplied by the decay coefficient (such as 0.8); if the power fluctuation exceeds the standard, the objective function value is multiplied by 0.9.

[0211] Pareto Front Solution: NSGA-II generates a non-dominated solution set (Pareto Front), for example, 10 sets of solutions (charging and discharging current, switching frequency, and power factor compensation parameter combinations) for the decision-maker to choose from.

[0212] ‌A rolling time domain optimization algorithm is used to solve the objective function. The optimal charge and discharge current values, PCS switching frequency, and power factor compensation parameters are iteratively calculated within each control cycle to generate a candidate set of dynamic PCS operating parameters.

[0213] Rolling Horizon Optimization (RHC) Process

[0214] Time Windowing: Decompose the optimization problem into rolling sub-problems, each covering the next N control cycles (e.g., N=3, or 15 seconds). For example, at time t=0, optimize the parameters from t=0 to 15 seconds, and at time t=5, roll over and optimize the parameters from t=5 to 20 seconds.

[0215] Dynamic Parameter Adjustment: Within each window, the objective function and constraints are recalculated based on the latest data. For example, at t = 0, the calculated current is 800A and the switching frequency is 15kHz. At t = 5 seconds, due to rising temperature, the current is reduced to 750A and the switching frequency is reduced to 12kHz.

[0216] Candidate set generation: Each solution set contains the charge and discharge current (I), switching frequency (f_sw), and power factor compensation angle (θ). For example:

[0217] Candidate 1: I=800A, f_sw=15kHz, θ=0.95 (capacitive compensation);

[0218] Candidate 2: I=780A, f_sw=14kHz, θ=0.98.

[0219] Real-time computing and hardware acceleration

[0220] Computing resource allocation: FPGAs (such as the Xilinx Alveo U280) are used to accelerate optimization algorithms, reducing the time required for a single iteration from 10ms to 0.5ms.

[0221] Parallel Computing Optimization: Calculate multiple candidate solutions simultaneously (e.g., 50 solutions) and select the optimal solution by prioritizing it. For example, priority = objective function value × weight (efficiency weight = 0.6, lifespan weight = 0.4).

[0222] Parameter Smooth Transition: To prevent sudden changes in parameters between cycles, a rate of change limit is imposed on parameters such as current and frequency (e.g., current rate of change ≤ 50A / s). For example, if the current current is 800A, the maximum allowable value in the next cycle is 800 + 50×5 = 1050A.

[0223] Perform feasibility verification on candidate parameters in the dynamic PCS operating parameter candidate set, detect whether the parameters exceed the safety limit through the battery polarization voltage model and the grid harmonic analysis model, and output the verified dynamic adjustment PCS operating parameter instruction set.

[0224] Battery polarization voltage model verification

[0225] Model Building: Calculate the polarization voltage (V_pol) of the battery at a specific current based on an equivalent circuit model (e.g., a second-order RC model). For example, at a current of 800A, V_pol = 0.05V (initial value) + 0.01V (temperature compensation).

[0226] Safety Margin Detection: Limits polarization voltage to ≤ 0.1V (to prevent lithium deposition). If the limit is exceeded, the candidate parameter is deemed invalid. For example, if Candidate 1's V_pol = 0.12V, the elimination mechanism is triggered.

[0227] Dynamic Parameter Correction: Linearly adjust parameters that exceed the specified value. For example, if the current is reduced from 800A to 750A, V_pol will be recalculated to 0.09V, which meets the requirement.

[0228] Grid harmonic analysis and verification

[0229] Harmonic distortion calculation: The harmonic components of the PCS output current are analyzed using a Fourier transform (FFT) to calculate the total harmonic distortion (THD). For example, when the switching frequency f_sw = 15kHz, THD = 4.5% (the standard requires ≤ 5%).

[0230] Filter parameter adjustment: If THD exceeds the standard (e.g., THD = 5.2%), increase the LC filter inductance (e.g., from 100μH to 120μH) or adjust the switching frequency (e.g., from 15kHz to 16kHz).

[0231] Electromagnetic compatibility (EMC) testing: This test checks whether high-frequency harmonics (e.g., 150kHz-30MHz) exceed emission limits (e.g., EN 55022 Class A). If so, a magnetic ring or shielding layer is added.

[0232] Instruction set generation and issuance

[0233] Verify by command encapsulation: Encapsulate parameters such as current, frequency, and power factor into Modbus TCP commands, for example: { "CMD": "SET_PCS_PARAM", "I": 750, "f_sw": 14, "theta": 0.98}.

[0234] Based on the collaborative optimization dataset, the Model Predictive Control (MPC) framework establishes a multi-objective optimization function with a rolling time horizon of 5-15 minutes. Using battery life decay (e.g., capacity loss ≤ 0.01% / cycle) as a hard constraint and grid power fluctuation (e.g., frequency deviation ≤ 0.2Hz) as a soft constraint, the function dynamically solves for the PCS's optimal charge and discharge currents, switching frequency, and power factor compensation parameters. Through iterative optimization using the Lagrange multiplier method, dynamic control commands are generated that balance battery life and grid stability. This overcomes the limitations of traditional single-objective control, achieving a dynamic balance between battery life and grid demand, and improving the energy storage system's economic efficiency and grid compatibility.

[0235] S204: Based on the PCS operating parameters, the internal temperature distribution of the high-voltage integrated unit is calculated in real time using a three-dimensional thermal field reconstruction model. The cooling fan speed and power device switching frequency are dynamically adjusted to generate coordinated control instructions. These instructions are then fed back to the BMS and EMS to update the battery charge and discharge thresholds and grid dispatch strategy, forming a closed-loop control link. Specifically, this may include:

[0236] Based on the PCS operating parameters, the finite element method is used to calculate the power device loss in real time, and a three-dimensional thermal field distribution model inside the high-voltage integrated machine is constructed using fluid mechanics equations.

[0237] Power Device Loss Modeling and Finite Element Analysis

[0238] Loss Decomposition and Parameter Collection: Power device losses (such as IGBT modules) are divided into conduction loss and switching loss. For example, when the PCS operating parameters are charge / discharge current I = 800A, switching frequency f_sw = 15kHz, and bus voltage V_dc = 1000V, the conduction loss is proportional to the square of the current (P_cond = I² × R_ds(on). When R_ds(on) = 0.5mΩ, P_cond = 800² × 0.0005 = 320W). Switching loss is related to frequency and voltage (P_sw = 0.5 × V_dc × I × t_sw × f_sw. When t_sw = 100ns, P_sw = 0.5 × 1000 × 800 × 100e-9 × 15e3 = 600W). The total loss P_total = 320 + 600 = 920W.

[0239] Finite Element Meshing: Based on the 3D geometric model of the high-voltage integrated circuit (e.g., generated in SolidWorks or ANSYS SpaceClaim as a 3D thermal field reconstruction model), key components such as power devices, busbars, and heat sinks are meshed. The mesh size is set to 0.5 mm to balance accuracy and computational speed. For example, an IGBT module is meshed into 500,000 hexahedral elements, and the heat sink baseplate is meshed into 300,000 tetrahedral elements.

[0240] Thermal-Electrical Coupling Simulation: Using ANSYS Mechanical or COMSOL Multiphysics, load the losses as heat sources onto the mesh nodes to solve for the steady-state temperature field. For example, the IGBT junction temperature, T_junction, equals losses × thermal resistance (R_th = 0.1°C / W) + ambient temperature (T_amb = 40°C). Thus, T_junction = 920 × 0.1 + 40 = 132°C, which needs to be further reduced to a safe value (e.g., <110°C) through fluid cooling.

[0241] Fluid mechanics equations and heat dissipation channel modeling

[0242] Parametric Modeling of the Cooling Channel: Based on the fan deployment plan (e.g., number of axial fans = 4, maximum airflow per fan Q = 500 CFM), a fluid domain model of the cooling channel was constructed. The inlet boundary condition was set to a wind speed v_in = 5 m / s (corresponding to Q = 500 CFM), the outlet was a pressure outlet (P_out = 1 atm), and the heat sink fin surface was set to a no-slip wall.

[0243] Turbulence Model Selection: Use the k-ε turbulence model (for high Reynolds number flows) or the SST k-ω model (for near-wall flows) to solve the Navier-Stokes equations. For example, using Fluent to calculate the convective heat transfer coefficient on the heat sink fin surface, h = 50 W / (m²·K), the fin temperature drops from 132°C to 98°C.

[0244] Thermal and flow field coupling iteration: The temperature field calculated by finite element analysis is used as input for fluid simulation. Physical parameters such as air density and viscosity are updated, and iteration is performed until the residual converges (for example, the energy equation residual is less than 1e-6). Finally, a 3D temperature distribution cloud map of the HV series unit is output as a 3D thermal field distribution model, with key hotspots marked (for example, the IGBT module center temperature = 112°C, requiring further optimization).

[0245] Real-time computing and hardware acceleration

[0246] GPU-accelerated solution: NVIDIA A100 GPUs are used for parallel computing of finite element and fluid dynamics equations, reducing the single simulation time from minutes to seconds (e.g., a complete thermal field update can be completed in 10 seconds).

[0247] Application of reduced-order model (ROM): For highly repetitive operating conditions (such as constant current charging and discharging), the thermal field characteristic modes are extracted through principal component analysis (PCA) to construct a ROM to replace the full-order model, boosting the calculation speed to the millisecond level (e.g., predicting temperature changes within 1 ms).

[0248] Online calibration mechanism: The actual temperature data is collected in real time through an infrared thermal imager (such as FLIR A655sc) and compared with the simulation results. If the deviation exceeds ±5°C, an adaptive correction of the model parameters is triggered (such as adjusting the convective heat transfer coefficient or material thermal conductivity).

[0249] Identify the areas with excessive temperature in the three-dimensional thermal field distribution model, dynamically adjust the speed gears of the cooling fans and the switching frequencies of power devices, and generate a draft of the heat dissipation control instructions;

[0250] Detection and prioritization of areas with excessive temperature

[0251] Threshold determination and area marking: Define the temperature safety threshold (e.g., T_safe = 110°C), and mark all areas where the temperature exceeds the threshold by traversing the nodes of the three-dimensional thermal field grid (e.g., the temperatures of IGBT modules A1, A2, and B1 are 112°C, 115°C, and 108°C respectively).

[0252] Classification of thermal hazard levels: Classify according to the over-temperature amplitude and component importance. For example:

[0253] Level 1 over-temperature (T > 120°C): Immediately reduce the load or shut down;

[0254] Level 2 over-temperature (115°C < T ≤ 120°C): Forcefully enhance the heat dissipation intensity;

[0255] Level 3 over-temperature (110°C < T ≤ 115°C): Locally adjust the fan and switching frequency.

[0256] Dynamic priority assignment: The weighted scoring method is adopted, where the score = over-temperature amplitude × weight (e.g., IGBT weight = 0.7, capacitor weight = 0.3), and the area with the highest score is processed first. For example, the score of IGBT module A2 = (115 - 110) × 0.7 = 3.5, which needs to be adjusted first.

[0257] Coordinated control of cooling fans and switching frequencies

[0258] Fan speed adjustment strategy: This strategy matches the fan to the overtemperature zone. For example, if fan FAN1 is cooling IGBT module A1, its speed will be increased from 2000 RPM to 2500 RPM (corresponding to an increase in air volume from 400 CFM to 500 CFM), while limiting the maximum speed to 3000 RPM to prevent excessive noise.

[0259] Dynamic Switching Frequency Adjustment: This reduces the switching frequency of power devices in over-temperature zones to reduce losses. For example, reducing the switching frequency of IGBT module A2 from 15kHz to 12kHz reduces switching losses from 600W to 480W (a 20% reduction) and the junction temperature from 115°C to 109°C.

[0260] Multivariable Coupling Optimization: A fuzzy logic controller (FLC) is used to coordinate fan speed and switching frequency. The input variables are the temperature deviation (ΔT = T - T_safe) and its rate of change (dΔT / dt), and the output variables are the fan speed increment (ΔRPM) and the switching frequency adjustment (Δf_sw). For example, the rule base definition is:

[0261] If ΔT = 5°C and dΔT / dt = +1°C / s, then ΔRPM = +500, Δf_sw = -3kHz;

[0262] If ΔT = 3°C and dΔT / dt = -0.5°C / s, then ΔRPM = +300 and Δf_sw = -1kHz.

[0263] Draft instruction generation and conflict resolution

[0264] Command Conflict Detection: If multiple adjustment commands conflict (for example, increasing the speed of a particular fan might cause overcooling in adjacent areas), a compromise solution is found using game theory models (such as Nash equilibrium). For example, increasing the speed of fan FAN1 to 2500 RPM would reduce the temperature of capacitor C1 from 95°C to 88°C, but this might increase system noise. The optimal solution must be determined after careful consideration.

[0265] ‌Instruction encoding and encapsulation‌: Encode the adjustment parameters into a JSON format instruction draft. For example:

[0266] {

[0267] "cmd_id": "THERMAL_CTRL_001",

[0268] "timestamp": "2023-10-01T14:30:00Z",

[0269] "actions": [

[0270] {"target": "FAN1", "param": "rpm", "value": 2500},

[0271] {"target": "IGBT_A2", "param": "f_sw", "value": 12} ]

[0273] }.

[0274] Risk Assessment and Alternative Solutions: Conduct a Failure Mode and Effects Analysis (FMEA) on the draft directive. For example, if wind turbine FAN1 fails and cannot be accelerated, an alternative solution will be implemented (such as shutting down secondary loads to reduce overall losses).

[0275] The draft heat dissipation control instructions are input into the digital twin system for simulation and execution, and the temperature change trend after adjustment is predicted. If the simulation results do not meet expectations, the instructions are re-planned until the thermal management requirements are met, and the final coordinated control instructions are obtained;

[0276] Digital twin system modeling and real-time synchronization

[0277] Twin Model Construction: A digital twin of the HV series is created using Unity3D or Siemens NX, mapping the physical entity's 3D thermal field, electrical parameters, and mechanical state in real time. For example, parameters such as IGBT module temperature, fan speed, and switching frequency are synchronized once per second.

[0278] Simulation Engine Configuration: Integrate Modelica or Simulink as the backend simulation engine and define the coupled thermal-electrical-fluid multiphysics equations. For example, set the simulation step length Δt to 1 second to predict temperature changes over the next 60 seconds.

[0279] ‌Command injection and scenario triggering‌: Inject draft commands (such as increasing FAN1 speed to 2500 RPM) into the digital twin to activate preset simulation scenarios (such as "thermal management test under extreme frequency modulation load").

[0280] Predictive simulation and results evaluation

[0281] Short-term trend prediction: Run a simulation for 5-10 seconds and output key metrics (such as IGBT junction temperature, heat sink outlet air velocity, and total system power consumption). For example, the prediction results show that after FAN1 speed increases, the IGBT_A2 temperature drops from 115°C to 108°C within 5 seconds, but FAN1 power consumption increases from 200W to 250W.

[0282] Multi-objective compliance check: Check whether the prediction results meet all constraints:

[0283] Temperature constraint: T≤110℃;

[0284] Noise restriction: fan noise ≤ 75dB(A);

[0285] Efficiency constraint: system efficiency ≥ 90%.

[0286] If any of the conditions is not met (e.g. noise = 78dB(A)), the draft directive is deemed invalid.

[0287] Sensitivity Analysis: Monte Carlo simulations are used to assess the impact of parameter uncertainty. For example, if the wind speed fluctuates by ±10%, the predicted temperature may fluctuate by ±2°C. It is necessary to ensure that the temperature does not exceed the specified value even in the worst-case scenario.

[0288] Instruction replanning and iterative optimization

[0289] Root Cause Analysis and Parameter Adjustment: If the predicted results fall short of the target, identify the cause and adjust the command. For example, if excessive noise levels are caused by increased fan speed, the switch frequency can be reduced (from 15kHz to 10kHz) while the speed can be slightly increased (from 2000 to 2200 RPM).

[0290] Multi-scenario parallel testing: Multiple candidate instruction drafts are run simultaneously in the digital twin (e.g., Scenario A: frequency reduction + medium-speed fan; Scenario B: shutting down secondary circuits + high-speed fan). The solution with the highest overall score is selected. Example scoring formula: Score = 0.6 × (temperature reduction) + 0.2 × (efficiency loss) + 0.2 × (noise reduction).

[0291] Convergence conditions and termination mechanism: Set a maximum number of iterations (e.g., 10) or an error tolerance (e.g., temperature deviation ≤ 1°C). Output the final command when the condition is met. For example, after three iterations, a feasible solution is obtained: FAN1 = 2300 RPM, IGBT_A2 f_sw = 11kHz, predicted temperature = 109°C, and noise = 73dB(A).

[0292] The final coordinated control instructions are sent to the PCS and cooling system for execution. At the same time, the temperature distribution data is fed back to the BMS to update the battery charge and discharge thresholds, and grid dispatch strategy correction suggestions are sent to the EMS to form a closed-loop control link.

[0293] ‌Command issuance and device control‌

[0294] Protocol and interface adaptation: Send instructions to the PCS and wind turbine controller via the CAN bus or EtherCAT protocol. For example, to send CAN frame data:

[0295] ID: 0x301, Data: [0x12 0x34 0x56] / / 0x12 = fan FAN1, 0x34 = speed 2300 RPM;

[0296] ID: 0x302, Data: [0xA2 0x0B 0x00] / / 0xA2 = IGBT_A2, 0x0B = switching frequency 11 kHz.

[0297] Execution status monitoring: Real-time reading of device feedback (such as actual fan speed and IGBT temperature). If the deviation between the actual parameter and the command exceeds ±5% (such as a fan speed error >100 RPM), an abnormal alarm is triggered and redundant control is initiated (such as switching to a backup fan).

[0298] Timing Synchronization and Queuing: Priority queues are used to manage the order in which commands are executed. For example, temperature control commands have high priority, while power factor compensation commands have medium priority, ensuring that urgent operations are executed first.

[0299] ‌Data Feedback and Threshold Update‌

[0300] Dynamic adjustment of BMS charge and discharge thresholds: Calculates the maximum allowable battery charge and discharge current based on temperature distribution data (e.g., maximum temperature = 109°C). For example, if the original threshold I_max = 1000A, based on the temperature-current derating curve (5% derating for every 10°C increase in temperature), I_max is adjusted to 1000 × (1-(109-25) / 10 × 0.05) = 1000 × 0.58 = 580A.

[0301] State of Health (SOH) compensation: If the battery SOH is less than 80%, the threshold is further lowered (e.g., I_max = 580 × 0.8 = 464A) and the parameters are updated through the BMS's UDS protocol (ISO 14229).

[0302] Historical data storage and analysis: Store temperature, current, SOH, and other data in a time series database (such as InfluxDB) for subsequent life prediction and maintenance decision-making.

[0303] Collaborative optimization of power grid dispatching strategies

[0304] ‌EMS policy modification suggestion generation‌: Based on the heat dissipation capacity limit (such as the current maximum sustainable power = 500kW), a scheduling suggestion is sent to the EMS. For example:

[0305] {

[0306] "type": "POWER_LIMIT_ADJUST",

[0307] "new_max_power": 500,

[0308] "duration": 300,

[0309] "reason": "thermal_constraint"

[0310] }.

[0311] Adjusting Demand Response (DR) Participation: If the grid needs to call upon the energy storage system during peak hours, but the system's heat dissipation has reached its limit, the EMS is advised to reduce its frequency regulation participation (e.g., from 100% to 70%) or request a delayed response (e.g., after 10 minutes).

[0312] Closed-loop self-verification: By comparing key indicators before and after command execution (such as temperature, efficiency, and grid response error), the closed-loop control effect is evaluated and control parameters (such as PID gain and fuzzy rule base) are automatically optimized.

[0313] ‌Technical Details and Parameter Examples‌

[0314] Finite element simulation parameters

[0315] Grid quantity: 500,000 cells for IGBT modules and 300,000 cells for heat sinks;

[0316] Material properties: Copper busbar thermal conductivity 401 W / (m·K), aluminum heat sink 237 W / (m·K);

[0317] Boundary conditions: ambient temperature 40°C, initial wind speed 5m / s.

[0318] Digital Twin Simulation Configuration

[0319] Simulation step: 1 second;

[0320] Prediction duration: 60 seconds;

[0321] Number of Monte Carlo samples: 1000 times.

[0322] ‌Control Command Parameters‌

[0323] Fan speed range: 1000-3000 RPM;

[0324] Switching frequency range: 5-20kHz;

[0325] Temperature sampling frequency: 10Hz.

[0326] ‌Application Scenario Examples‌

[0327] Scenario description: While participating in grid frequency regulation, an energy storage power station experiences a sudden increase in ambient temperature (40°C to 45°C), causing the IGBT module to overheat.

[0328] Thermal Field Modeling: FEM calculation of IGBT losses of 920W and predicted junction temperature of 132°C;

[0329] Command generation: Increase fan FAN1 to 2500 RPM and reduce switching frequency to 12kHz;

[0330] Digital Twin Verification: The predicted temperature dropped to 109°C, but the noise level exceeded the standard by 76 dB(A).

[0331] ‌Command reprogramming‌: Adjust to fan speed 2300 RPM, switching frequency 11kHz, temperature = 110℃, noise = 73dB(A);

[0332] Closed-loop execution: Update the BMS charge and discharge thresholds to 580A. It is recommended that the EMS power be limited to 500kW.

[0333] Based on PCS operating parameters (such as IGBT switching frequency and current), the finite element method is used to calculate power device losses in real time. Combined with wind speed data from the cooling channel, a three-dimensional thermal field distribution model (with a resolution of 1 mm³) is constructed. After identifying areas with excessive temperatures (e.g., >85°C), the system dynamically adjusts the fan speed (e.g., from 2000 rpm to 3500 rpm) and switching frequency (e.g., from 10 kHz to 8 kHz). The control effect is verified through a digital twin, ultimately generating coordinated control instructions. Temperature data is also fed back to the BMS to adjust charge and discharge thresholds (e.g., current limiting by 10%), and grid dispatch correction suggestions (e.g., reducing frequency regulation response speed) are sent to the EMS. This achieves closed-loop coordination of thermal, electrical, and control functions, preventing equipment failures caused by localized overheating. This real-time feedback mechanism enhances the system's adaptability, ensuring the safety and response accuracy of the energy storage system under complex operating conditions.

[0334] It can be seen that according to the power level and heat distribution requirements of the energy storage system, the high-voltage box and the power conversion system PCS are three-dimensionally coupled through the topology optimization algorithm to generate the physical layer architecture scheme of the high-voltage integrated machine; based on the physical layer architecture scheme of the high-voltage integrated machine, the timestamp alignment technology is used to construct a collaborative optimization data set; the collaborative optimization data set is input into the model predictive control framework to generate dynamically adjusted PCS operating parameters; according to the PCS operating parameters, the internal temperature distribution of the high-voltage integrated machine is calculated in real time through the three-dimensional thermal field reconstruction model, the cooling fan speed and the power device switching frequency are dynamically adjusted, and collaborative control instructions are generated, thereby improving the integration level and collaborative control capabilities of the high-voltage energy storage system and realizing efficient and stable operation of the energy storage system.

[0335] Another embodiment of the present invention provides a high voltage integrated and coordinated control system for an energy storage system, see Figure 3 , the system may include:

[0336] Design module 301 is used to perform a three-dimensional spatial coupling design of the high-voltage box and the power conversion system (PCS) based on the power level and heat distribution requirements of the energy storage system using a topology optimization algorithm to generate a physical layer architecture solution for the high-voltage integrated device. The topology optimization algorithm simultaneously calculates the optimal matching relationship between the electrical connection path and the heat dissipation channel.

[0337] A construction module 302 is configured to synchronously collect real-time battery status data from a battery management system (BMS) and grid dispatch instructions from an energy management system (EMS) via a dynamic data bus based on the physical layer architecture solution of the high-voltage integrated device, and to construct a collaborative optimization dataset using timestamp alignment technology, wherein the collaborative optimization dataset includes multi-dimensional coupling characteristics of battery health, charge and discharge rates, and grid demand response;

[0338] A generation module 303 is configured to input the collaborative optimization data set into a model predictive control framework, employ a rolling horizon optimization algorithm to solve the charge and discharge strategy, and generate dynamically adjusted PCS operating parameters, wherein the rolling horizon optimization algorithm achieves multi-objective collaborative optimization by constraining the battery life decay rate and the grid power fluctuation range;

[0339] The control module 304 is used to calculate the internal temperature distribution of the high-voltage integrated machine in real time through a three-dimensional thermal field reconstruction model based on the PCS operating parameters, dynamically adjust the cooling fan speed and the switching frequency of the power device, generate coordinated control instructions, and simultaneously feed back to the BMS and EMS to update the battery charging and discharging thresholds and the power grid scheduling strategy, forming a closed-loop control link.

[0340] It can be seen that according to the power level and heat distribution requirements of the energy storage system, the high-voltage box and the power conversion system PCS are three-dimensionally coupled through the topology optimization algorithm to generate the physical layer architecture scheme of the high-voltage integrated machine; based on the physical layer architecture scheme of the high-voltage integrated machine, the timestamp alignment technology is used to construct a collaborative optimization data set; the collaborative optimization data set is input into the model predictive control framework to generate dynamically adjusted PCS operating parameters; according to the PCS operating parameters, the internal temperature distribution of the high-voltage integrated machine is calculated in real time through the three-dimensional thermal field reconstruction model, the cooling fan speed and the power device switching frequency are dynamically adjusted, and collaborative control instructions are generated, thereby improving the integration level and collaborative control capabilities of the high-voltage energy storage system and realizing efficient and stable operation of the energy storage system.

[0341] An embodiment of the present invention further provides a storage medium storing a computer program, wherein the computer program is configured to execute the steps of any one of the above method embodiments when running.

[0342] Specifically, in this embodiment, the above-mentioned storage medium may be configured to store a computer program for performing the following steps:

[0343] S201: Based on the power level and heat distribution requirements of the energy storage system, a topology optimization algorithm is used to perform a three-dimensional spatial coupling design between the high-voltage box and the power conversion system (PCS) to generate a physical layer architecture solution for the high-voltage integrated device. The topology optimization algorithm simultaneously calculates the optimal matching relationship between the electrical connection path and the heat dissipation channel.

[0344] S202, based on the physical layer architecture solution of the high-voltage integrated device, synchronously collect real-time battery status data from the battery management system (BMS) and grid dispatch instructions from the energy management system (EMS) via a dynamic data bus, and use timestamp alignment technology to construct a collaborative optimization data set, wherein the collaborative optimization data set includes multi-dimensional coupling characteristics of battery health, charge and discharge rates, and grid demand response;

[0345] S203: Inputting the collaborative optimization data set into a model predictive control framework, using a rolling horizon optimization algorithm to solve the charge and discharge strategy and generate dynamically adjusted PCS operating parameters, wherein the rolling horizon optimization algorithm achieves multi-objective collaborative optimization by constraining the battery life decay rate and the grid power fluctuation range;

[0346] S204, based on the PCS operating parameters, calculate the internal temperature distribution of the high-voltage integrated machine in real time through a three-dimensional thermal field reconstruction model, dynamically adjust the cooling fan speed and the power device switching frequency, generate collaborative control instructions, and simultaneously feed back to the BMS and EMS to update the battery charge and discharge thresholds and grid scheduling strategy, forming a closed-loop control link.

[0347] It can be seen that according to the power level and heat distribution requirements of the energy storage system, the high-voltage box and the power conversion system PCS are three-dimensionally coupled through the topology optimization algorithm to generate the physical layer architecture scheme of the high-voltage integrated machine; based on the physical layer architecture scheme of the high-voltage integrated machine, the timestamp alignment technology is used to construct a collaborative optimization data set; the collaborative optimization data set is input into the model predictive control framework to generate dynamically adjusted PCS operating parameters; according to the PCS operating parameters, the internal temperature distribution of the high-voltage integrated machine is calculated in real time through the three-dimensional thermal field reconstruction model, the cooling fan speed and the power device switching frequency are dynamically adjusted, and collaborative control instructions are generated, thereby improving the integration level and collaborative control capabilities of the high-voltage energy storage system and realizing efficient and stable operation of the energy storage system.

[0348] An embodiment of the present invention further provides an electronic device, comprising a memory and a processor, wherein the memory stores a computer program, and the processor is configured to run the computer program to perform the steps in any one of the above method embodiments.

[0349] Specifically, the electronic device may further include a transmission device and an input / output device, wherein the transmission device is connected to the processor, and the input / output device is connected to the processor.

[0350] Specifically, in this embodiment, the processor may be configured to execute the following steps through a computer program:

[0351] S201: Based on the power level and heat distribution requirements of the energy storage system, a topology optimization algorithm is used to perform a three-dimensional spatial coupling design between the high-voltage box and the power conversion system (PCS) to generate a physical layer architecture solution for the high-voltage integrated device. The topology optimization algorithm simultaneously calculates the optimal matching relationship between the electrical connection path and the heat dissipation channel.

[0352] S202, based on the physical layer architecture solution of the high-voltage integrated device, synchronously collect real-time battery status data from the battery management system (BMS) and grid dispatch instructions from the energy management system (EMS) via a dynamic data bus, and use timestamp alignment technology to construct a collaborative optimization data set, wherein the collaborative optimization data set includes multi-dimensional coupling characteristics of battery health, charge and discharge rates, and grid demand response;

[0353] S203: Inputting the collaborative optimization data set into a model predictive control framework, using a rolling horizon optimization algorithm to solve the charge and discharge strategy and generate dynamically adjusted PCS operating parameters, wherein the rolling horizon optimization algorithm achieves multi-objective collaborative optimization by constraining the battery life decay rate and the grid power fluctuation range;

[0354] S204, based on the PCS operating parameters, calculate the internal temperature distribution of the high-voltage integrated machine in real time through a three-dimensional thermal field reconstruction model, dynamically adjust the cooling fan speed and the power device switching frequency, generate collaborative control instructions, and simultaneously feed back to the BMS and EMS to update the battery charge and discharge thresholds and grid scheduling strategy, forming a closed-loop control link.

[0355] It can be seen that according to the power level and heat distribution requirements of the energy storage system, the high-voltage box and the power conversion system PCS are three-dimensionally coupled through the topology optimization algorithm to generate the physical layer architecture scheme of the high-voltage integrated machine; based on the physical layer architecture scheme of the high-voltage integrated machine, the timestamp alignment technology is used to construct a collaborative optimization data set; the collaborative optimization data set is input into the model predictive control framework to generate dynamically adjusted PCS operating parameters; according to the PCS operating parameters, the internal temperature distribution of the high-voltage integrated machine is calculated in real time through the three-dimensional thermal field reconstruction model, the cooling fan speed and the power device switching frequency are dynamically adjusted, and collaborative control instructions are generated, thereby improving the integration level and collaborative control capabilities of the high-voltage energy storage system and realizing efficient and stable operation of the energy storage system.

[0356] The above describes in detail the structure, features and effects of the present invention based on the embodiments shown in the drawings. The above is only a preferred embodiment of the present invention, but the scope of implementation of the present invention is not limited to what is shown in the drawings. Any changes made in accordance with the concept of the present invention, or modifications to equivalent embodiments with equivalent changes, which do not exceed the spirit covered by the description and drawings, should be within the scope of protection of the present invention.

Claims

1. A high-voltage integrated and coordinated control method for an energy storage system, characterized in that: The method comprises: Based on the power level and heat distribution requirements of the energy storage system, a topology optimization algorithm is used to perform a three-dimensional spatial coupling design between the high-voltage box and the power conversion system (PCS), generating a physical layer architecture solution for the high-voltage integrated device. The topology optimization algorithm simultaneously calculates the optimal matching relationship between the electrical connection path and the heat dissipation channel. Based on the physical layer architecture solution of the high-voltage integrated device, real-time battery status data from the battery management system (BMS) and grid dispatch instructions from the energy management system (EMS) are synchronously collected through a dynamic data bus, and timestamp alignment technology is used to construct a collaborative optimization data set. The collaborative optimization data set includes multi-dimensional coupling characteristics of battery health, charge and discharge rates, and grid demand response. The collaborative optimization dataset is input into a model predictive control framework, and a rolling horizon optimization algorithm is used to solve the charge and discharge strategy to generate dynamically adjusted PCS operating parameters. The rolling horizon optimization algorithm achieves multi-objective collaborative optimization by constraining the battery life decay rate and the grid power fluctuation range. The PCS operating parameters include: optimal charge and discharge current values, PCS switching frequency, and power factor compensation parameters. According to the PCS operating parameters, the internal temperature distribution of the high-voltage integrated machine is calculated in real time through a three-dimensional thermal field reconstruction model, the cooling fan speed and the switching frequency of the power device are dynamically adjusted, and coordinated control instructions are generated. At the same time, these instructions are fed back to the BMS and EMS to update the battery charging and discharging thresholds and the grid scheduling strategy, forming a closed-loop control link.

2. The method according to claim 1, characterized in that Based on the power level and heat distribution requirements of the energy storage system, a topology optimization algorithm is used to perform a three-dimensional spatial coupling design between the high-voltage box and the power conversion system PCS to generate a physical layer architecture solution for the high-voltage integrated device. The topology optimization algorithm simultaneously calculates the optimal matching relationship between the electrical connection path and the heat dissipation channel, including: Based on the power level parameters and heat distribution requirements of the energy storage system, an initial 3D model of the high-voltage box and PCS is constructed, generating an initial 3D mesh model including electrical interfaces, heat dissipation vent locations, and internal cavities. In the initial 3D mesh model, the electrical connection path is designed based on the minimum path principle. The busbar routing length and impedance are iteratively optimized using the ant colony algorithm to generate a candidate set of electrical connection paths and mark the coordinates of key nodes. Simultaneously perform fluid dynamics simulation on the heat dissipation channel, optimize the heat dissipation fin layout based on the heat source distribution density, calculate the heat dissipation efficiency curve under different wind speeds, and generate the heat dissipation channel topology and fan deployment plan; Input the candidate set of electrical connection paths and the heat dissipation channel topology into the multi-objective optimization algorithm. Using the electrical loss rate and heat dissipation efficiency as constraints, the Pareto optimal solution set is solved to generate the electrical-heat dissipation coupled topology solution. Conduct electromagnetic-thermomechanical coupling simulation verification on the topology solution, detect local hotspots and areas with excessive electromagnetic interference, correct the connection paths and heat dissipation channel spacing, and output the final three-dimensional solution for the physical layer architecture of the high-voltage integrated machine.

3. The method according to claim 2, characterized in that The physical layer architecture based on the high-voltage integrated device synchronously collects real-time battery status data of the battery management system (BMS) and grid dispatch instructions of the energy management system (EMS) through a dynamic data bus, and uses timestamp alignment technology to construct a collaborative optimization data set, wherein the collaborative optimization data set contains multi-dimensional coupling characteristics of battery health, charge and discharge rate, and grid demand response, including: The battery status data of the BMS and the grid dispatch instructions of the EMS are collected through the dynamic data bus, and the data timestamps are aligned using hardware clock synchronization technology to generate a time-synchronized original data stream; Clean outliers from the original data stream, compensate for missing data based on the battery cluster topology, and use the covariance matrix to eliminate sensor noise interference to generate a purified multidimensional monitoring data set; Extract battery health characteristics, charge and discharge rate characteristics, and grid demand response characteristics from the multidimensional monitoring data set to construct a multidimensional coupling feature correlation matrix; Based on the multi-dimensional coupling feature correlation matrix, the dynamic correlation between battery health, charge and discharge rate, and grid demand response is analyzed using the Pearson correlation coefficient to generate a multi-dimensional coupling feature mapping table containing weight labels. The multi-dimensional coupling feature map is fused with the time series data, packaged into a collaborative optimization dataset with spatiotemporal labels according to the preset sampling period, and stored in the shared memory area of the edge computing node.

4. The method according to claim 3, characterized in that The collaborative optimization data set is input into the model predictive control framework, and a rolling horizon optimization algorithm is used to solve the charging and discharging strategy to generate dynamically adjusted PCS operating parameters, wherein the rolling horizon optimization algorithm realizes multi-objective collaborative optimization by constraining the battery life attenuation rate and the power grid power fluctuation range, including: The collaborative optimization dataset is fed into the model predictive control framework, which predicts the power demand curve for a preset number of control cycles in the future based on the current battery status and grid instructions, and generates a reference trajectory for charge and discharge power. A multi-objective optimization model was established, with reference to the charge and discharge power benchmark reference trajectory, battery life decay rate as the first constraint and grid power fluctuation range as the second constraint, and the objective function was defined as maximizing system efficiency. ‌A rolling time domain optimization algorithm is used to solve the objective function. The optimal charge and discharge current values, PCS switching frequency, and power factor compensation parameters are iteratively calculated within each control cycle to generate a candidate set of dynamic PCS operating parameters. Perform feasibility verification on candidate parameters in the dynamic PCS operating parameter candidate set, detect whether the parameters exceed the safety limit through the battery polarization voltage model and the grid harmonic analysis model, and output the verified dynamic adjustment PCS operating parameter instruction set.

5. The method according to claim 4, characterized in that According to the PCS operating parameters, the internal temperature distribution of the high-voltage integrated machine is calculated in real time through a three-dimensional thermal field reconstruction model, the cooling fan speed and the power device switching frequency are dynamically adjusted, and coordinated control instructions are generated. At the same time, these instructions are fed back to the BMS and EMS to update the battery charge and discharge thresholds and the grid dispatch strategy, forming a closed-loop control link, including: Based on the PCS operating parameters, the finite element method is used to calculate the power device loss in real time, and a three-dimensional thermal field distribution model inside the high-voltage integrated machine is constructed using fluid mechanics equations. Identify areas with excessive temperatures in the three-dimensional thermal field distribution model, dynamically adjust the speed of the cooling fan and the switching frequency of the power devices, and generate a draft of the cooling control instructions; The draft heat dissipation control instructions are input into the digital twin system for simulation and execution, and the temperature change trend after adjustment is predicted. If the simulation results do not meet expectations, the instructions are re-planned until the thermal management requirements are met, and the final coordinated control instructions are obtained; The final coordinated control instructions are sent to the PCS and cooling system for execution. At the same time, the temperature distribution data is fed back to the BMS to update the battery charge and discharge thresholds, and grid dispatch strategy correction suggestions are sent to the EMS to form a closed-loop control link.

6. A high-voltage integrated and coordinated control system for an energy storage system, characterized in that: The system comprises: A design module is used to perform a three-dimensional spatial coupling design between the high-voltage box and the power conversion system (PCS) based on the power level and heat distribution requirements of the energy storage system using a topology optimization algorithm to generate a physical layer architecture solution for the high-voltage integrated device. The topology optimization algorithm simultaneously calculates the optimal matching relationship between the electrical connection path and the heat dissipation channel; A construction module is used to synchronously collect real-time battery status data of the battery management system (BMS) and grid dispatch instructions of the energy management system (EMS) through a dynamic data bus based on the physical layer architecture solution of the high-voltage integrated device, and use timestamp alignment technology to construct a collaborative optimization data set, wherein the collaborative optimization data set includes multi-dimensional coupling characteristics of battery health, charge and discharge rates, and grid demand response; a generation module, configured to input the collaborative optimization data set into a model predictive control framework, employ a rolling horizon optimization algorithm to solve the charge and discharge strategy, and generate dynamically adjusted PCS operating parameters, wherein the rolling horizon optimization algorithm achieves multi-objective collaborative optimization by constraining the battery life attenuation rate and the grid power fluctuation range, and the PCS operating parameters include: optimal charge and discharge current values, PCS switching frequency, and power factor compensation parameters; The control module is used to calculate the internal temperature distribution of the high-voltage integrated machine in real time through a three-dimensional thermal field reconstruction model based on the PCS operating parameters, dynamically adjust the cooling fan speed and the switching frequency of the power device, generate coordinated control instructions, and simultaneously feed back to the BMS and EMS to update the battery charging and discharging thresholds and the power grid scheduling strategy, forming a closed-loop control link.

7. The system according to claim 6, characterized in that The design module is specifically used for: Based on the power level parameters and heat distribution requirements of the energy storage system, an initial 3D model of the high-voltage box and PCS is constructed, generating an initial 3D mesh model including electrical interfaces, heat dissipation vent locations, and internal cavities. In the initial 3D mesh model, the electrical connection path is designed based on the minimum path principle. The busbar routing length and impedance are iteratively optimized using the ant colony algorithm to generate a candidate set of electrical connection paths and mark the coordinates of key nodes. Simultaneously perform fluid dynamics simulation on the heat dissipation channel, optimize the heat dissipation fin layout based on the heat source distribution density, calculate the heat dissipation efficiency curve under different wind speeds, and generate the heat dissipation channel topology and fan deployment plan; Input the candidate set of electrical connection paths and the heat dissipation channel topology into the multi-objective optimization algorithm. Using the electrical loss rate and heat dissipation efficiency as constraints, the Pareto optimal solution set is solved to generate the electrical-heat dissipation coupled topology solution. Conduct electromagnetic-thermomechanical coupling simulation verification on the topology solution, detect local hotspots and areas with excessive electromagnetic interference, correct the connection paths and heat dissipation channel spacing, and output the final three-dimensional solution for the physical layer architecture of the high-voltage integrated machine.

8. The system according to claim 7, characterized in that The building blocks are specifically used for: The battery status data of the BMS and the grid dispatch instructions of the EMS are collected through the dynamic data bus, and the data timestamps are aligned using hardware clock synchronization technology to generate a time-synchronized original data stream; Clean outliers from the original data stream, compensate for missing data based on the battery cluster topology, and use the covariance matrix to eliminate sensor noise interference to generate a purified multidimensional monitoring data set; Extract battery health characteristics, charge and discharge rate characteristics, and grid demand response characteristics from the multidimensional monitoring data set to construct a multidimensional coupling feature correlation matrix; Based on the multi-dimensional coupling feature correlation matrix, the dynamic correlation between battery health, charge and discharge rate, and grid demand response is analyzed using the Pearson correlation coefficient to generate a multi-dimensional coupling feature mapping table containing weight labels. The multi-dimensional coupling feature map is fused with the time series data, packaged into a collaborative optimization dataset with spatiotemporal labels according to the preset sampling period, and stored in the shared memory area of the edge computing node.

9. A storage medium, characterized in that: The storage medium stores a computer program, wherein the computer program is configured to execute the method according to any one of claims 1 to 5 when executed.

10. An electronic device comprising a memory and a processor, characterized in that: A computer program is stored in the memory, and the processor is configured to run the computer program to perform the method according to any one of claims 1 to 5.

Images