Flywheel rotor temperature estimation method and system

Abstract

The application relates to the technical field of flywheel temperature control, and specifically discloses a flywheel rotor temperature estimation method and system. The flywheel rotor temperature estimation method comprises the following steps: obtaining a test temperature vector of a flywheel under multiple groups of preset operating conditions, and obtaining preset position temperature data of a rotor of the flywheel during operation; performing offline parameter calibration on a finite element thermal simulation model according to the test temperature vector to obtain a calibrated finite element thermal simulation model; constructing a lightweight reduced-order model based on the calibrated finite element thermal simulation model, inputting the preset position temperature data into the lightweight reduced-order model as input parameters, and outputting the current estimated temperature of the rotor at a target position. The application accurately captures local hot spots formed on the surface of the rotor due to uneven distribution of eddy current loss, improves the accuracy and real-time performance of temperature detection, and greatly reduces the calculation time after training of the lightweight reduced-order model, thereby realizing high-precision, low-delay real-time estimation of the global temperature of the flywheel.

Description

Technical Field

[0001] This invention relates to the field of flywheel temperature control technology, and in particular to a method and system for estimating the rotor temperature of a flywheel. Background Technology

[0002] In related technologies, infrared thermography and embedded sensors are the two main solutions for temperature detection of flywheel rotors in a vacuum environment. However, both solutions have significant limitations in practical applications: Firstly, while infrared thermography achieves non-contact measurement, it is susceptible to strong electromagnetic interference and limited by installation space, making it difficult to guarantee measurement accuracy. More importantly, this method can only obtain the average temperature of the rotor surface and cannot accurately reflect the local temperature gradient changes caused by the uneven distribution of eddy current losses. Secondly, the high-speed rotation of the rotor makes signal transmission of traditional wired sensors difficult to achieve; at the same time, the long-term reliability of embedded sensors in strong magnetic field environments is difficult to guarantee, and their installation can easily disrupt the dynamic balance performance of the flywheel structure.

[0003] Due to the limitations of the aforementioned hardware detection methods, traditional finite element thermal simulation models often require fixed parameter settings and cannot adapt to the dynamic changes in material performance parameters during actual operation. Furthermore, most existing temperature prediction models employ open-loop calculation methods, which easily lead to a gradual decrease in prediction accuracy over time. This lack of overall temperature prediction capability forces control systems to often only take reactive measures after detecting abnormal temperature increases. This post-event approach not only suffers from low model accuracy and delayed temperature detection but also fails to effectively prevent serious malfunctions such as thermal runaway. Summary of the Invention

[0004] The present invention aims to solve at least one of the technical problems existing in the prior art.

[0005] Therefore, one objective of this invention is to propose a method for estimating the rotor temperature of a flywheel. This method effectively overcomes the physical limitations of traditional detection methods, accurately captures local hot spots formed on the rotor surface due to uneven distribution of eddy current losses, improves the accuracy and real-time performance of temperature detection, and significantly reduces the computation time after training the lightweight reduced-order model. Furthermore, it fundamentally breaks the limitation of easy accuracy decay in traditional open-loop prediction modes, achieving high-precision, low-latency real-time estimation of the entire flywheel temperature.

[0006] Therefore, a second objective of the present invention is to provide a rotor temperature estimation system for a flywheel.

[0007] To achieve the above objectives, the first aspect of this invention discloses a method for estimating the rotor temperature of a flywheel, comprising: acquiring test temperature vectors of the flywheel under multiple preset operating conditions, and acquiring preset position temperature data of the flywheel rotor during operation; performing offline parameter calibration on a finite element thermal simulation model based on the test temperature vectors to obtain a calibrated finite element thermal simulation model; constructing a lightweight reduced-order model based on the calibrated finite element thermal simulation model, and inputting the preset position temperature data as input parameters into the lightweight reduced-order model, and outputting the current estimated temperature of the rotor at the target position.

[0008] Further, when performing offline parameter calibration on the finite element thermal simulation model based on the test temperature vector to obtain the calibrated finite element thermal simulation model, the process includes: acquiring the initial simulation results of the finite element thermal simulation model and comparing and verifying the initial simulation results with the test temperature vector point by point; performing partitioned evaluation of the point-by-point comparison verification according to preset component regions, and determining the partitioning error in each preset component region according to a preset algorithm; performing multi-condition calibration of the heat source compensation coefficient and / or iterative adjustment of the contact thermal resistance used to characterize the heat transfer path; after completing the multi-condition calibration of the heat source compensation coefficient and / or iterative adjustment of the contact thermal resistance used to characterize the heat transfer path, determining whether the partitioning error meets the preset convergence condition; if so, adjusting the convective heat transfer coefficient of the flywheel shell to obtain the calibrated finite element thermal simulation model; wherein, the preset convergence condition includes: the partitioning error is less than a preset convergence threshold. Therefore, it can effectively avoid the interference of low-sensitivity external heat dissipation boundary parameters such as heat source compensation coefficient and contact thermal resistance, improve the overall accuracy and convergence speed of the finite element thermal simulation model, and thus improve the accuracy and real-time performance of temperature detection.

[0009] Furthermore, obtaining the initial simulation results of the finite element thermal simulation model includes: assigning corresponding electromagnetic and thermal property parameters to each functional component within the finite element thermal simulation model; determining the losses of each functional component based on the electromagnetic parameters; using the losses of each functional component as the heat source input of the finite element thermal simulation model; and solving the problem according to a preset thermal field algorithm based on the thermal property parameters, the heat source input, and the preset boundary conditions of the finite element thermal simulation model to obtain the initial simulation results of the finite element thermal simulation model. Thus, through the above-mentioned refined material property definition and multi-dimensional boundary condition settings, the actual thermophysical behavior of the flywheel under complex working conditions can be accurately reproduced, thereby significantly improving the accuracy and reliability of the overall thermal field calculation, and further improving the accuracy of the finite element thermal simulation model calibration.

[0010] Furthermore, during the multi-condition calibration of the heat source compensation coefficient, the process includes: acquiring the rotor speed of the flywheel, the charging and discharging power of the flywheel, and the rotor operating temperature; and determining the heat source compensation coefficient based on the partitioning error, the rotor speed, the charging and discharging power, and the rotor operating temperature using a preset fitting algorithm and according to a preset differentiated operating condition calibration strategy. The heat source compensation coefficient includes a baseline coefficient, a speed influence factor, and a temperature influence factor. This effectively solves the simulation distortion problem caused by simple superposition under a single speed or temperature dimension in existing technologies, significantly improving the accuracy and robustness of the thermal field prediction of the flywheel under complex variable operating conditions, and thus improving the accuracy and real-time performance of temperature detection.

[0011] Furthermore, when performing iterative adjustments to the contact thermal resistance used to characterize the heat transfer path, the process includes: sequentially performing iterative calibration on the contact thermal resistance of each target contact interface according to a preset heat transfer path sequence; during the iterative calibration of the current target contact interface, obtaining the simulation temperature of the current target contact interface output by the finite element thermal simulation model, and the actual temperature of the current target contact interface; determining the temperature deviation between the actual temperature and the simulation temperature; using the temperature deviation as the iterative objective function, solving and correcting the contact pressure of the current target contact interface through a preset algorithm, so as to update the contact thermal resistance of the current target contact interface through changes in contact pressure; when the temperature deviation is less than a preset temperature difference threshold, determining that the contact thermal resistance calibration of the current target contact interface is complete, and jumping to the next target contact interface to continue performing the iterative calibration. This not only improves the convergence speed of the finite element thermal simulation model calibration but also avoids multi-parameter coupling interference, significantly improving the stability and calculation accuracy of the finite element thermal simulation model, thereby improving the accuracy and real-time performance of temperature estimation.

[0012] Further, when adjusting the convective heat transfer coefficient of the flywheel's outer shell to obtain a calibrated finite element thermal simulation model, the process includes: obtaining the simulated outer shell temperature output by the finite element thermal simulation model and obtaining the actual outer shell temperature; based on the simulated outer shell temperature and the actual outer shell temperature, determining whether the temperature difference change trend between the simulated outer shell temperature and the actual outer shell temperature is in the same direction; if so, determining the initial convective heat transfer coefficient of the outer shell based on a preset heat transfer formula, and iteratively adjusting the initial convective heat transfer coefficient of the outer shell according to a preset adjustment range; when iteratively adjusting the initial convective heat transfer coefficient of the outer shell according to the preset adjustment range, obtaining the maximum absolute error between the simulated outer shell temperature and the actual outer shell temperature; when the maximum absolute error is less than a preset fine-tuning convergence threshold, stopping the iterative adjustment of the initial convective heat transfer coefficient of the outer shell to obtain a calibrated finite element thermal simulation model. This improves the adaptability of the finite element thermal simulation model to complex assembly tolerances and variable air-cooling environments. At the same time, by solidifying a dedicated parameter set, the high-precision finite element thermal simulation model can be standardized and reused, reducing the repetitive calculation time for subsequent multi-condition finite element thermal simulation model analysis.

[0013] Further, when constructing a lightweight reduced-order model based on the calibrated finite element thermal simulation model, and inputting the preset position temperature data as input parameters into the lightweight reduced-order model to output the current estimated temperature of the rotor at the target position, the process includes: performing preset transient thermal field simulation on the calibrated finite element thermal simulation model based on multiple sets of preset operating conditions, collecting time-domain temperature response data of each target monitoring point within the calibrated finite element thermal simulation model, and constructing a multi-condition temperature snapshot sample set; based on the multi-condition temperature snapshot sample set, using multiple sets of preset operating conditions as input variables and the current estimated temperature of the rotor at the target position as the output variable, constructing the lightweight reduced-order model characterizing the mapping relationship between the preset operating conditions and the time-domain temperature response data; when the model deviation of the lightweight reduced-order model relative to the calibrated finite element thermal simulation model is less than or equal to a preset reduction error threshold, determining that the lightweight reduced-order model has been constructed; and inputting the preset position temperature data into the constructed lightweight reduced-order model to output the current estimated temperature of the rotor at the target position. This enables online temperature estimation of the flywheel rotor at the target location, effectively overcoming the physical limitations of traditional detection methods. It can accurately capture local hot spots formed on the rotor surface due to uneven distribution of eddy current losses, significantly improving the accuracy and real-time performance of temperature detection.

[0014] Furthermore, when acquiring the test temperature vector of the flywheel under multiple preset operating conditions, the process includes: continuously collecting multi-source temperature data under multiple preset operating conditions, wherein the multi-source temperature data includes stator temperature, bearing temperature, rotor temperature, casing temperature, ambient temperature, and fluid medium temperature; preprocessing and time-synchronizing the collected multi-source temperature data to construct the test temperature vector under each preset operating condition. This not only allows for continuous acquisition of full-cycle data without stopping the flywheel or disrupting the vacuum environment, avoiding the natural cooling error caused by traditional disassembly temperature measurement, but also accurately captures instantaneous temperature rise peaks caused by operating condition fluctuations. This provides high-fidelity, high-temporal-resolution benchmark data for subsequent finite element thermal simulation model calibration and order reduction analysis, thereby preventing the continuous propagation of original simulation errors during the finite element thermal simulation model lightweighting process, and achieving high-precision, low-latency real-time estimation of the flywheel's full-domain temperature.

[0015] Furthermore, acquiring the temperature data of the flywheel at a preset position during operation includes: collecting initial temperature data of the flywheel at a preset position with a preset sampling period; filtering the initial temperature data to obtain an effective temperature measurement result; and using the effective temperature measurement result as the temperature data of the rotor at the preset position during operation. Thus, by filtering the initial temperature data, measurement errors caused by instantaneous electromagnetic spike noise can be effectively filtered out, resulting in an accurate and effective temperature measurement result. Finally, using this effective temperature measurement result as the temperature data of the rotor at the preset position during operation not only ensures the anti-interference capability of non-contact temperature measurement under complex working conditions but also ensures the authenticity and high fidelity of the temperature data, providing reliable data support for subsequent accurate capture of local hot spots on the rotor surface and calibration of the finite element thermal simulation model.

[0016] According to the flywheel rotor temperature estimation method of the present invention, the test temperature vectors of the flywheel under multiple preset operating conditions are obtained, and the rotor temperature data of the flywheel at preset positions during operation are obtained. Then, based on the test temperature vectors, the finite element thermal simulation model is calibrated offline to obtain the calibrated finite element thermal simulation model. This not only corrects the inherent deviation of the finite element thermal simulation model itself, but also overcomes the defect of traditional models that cannot adapt to actual operating conditions due to the use of fixed parameter settings. Thus, it avoids the continuous transmission of the original simulation error in the process of reducing the order of the finite element thermal simulation model from the source, and improves the reproduction accuracy of the finite element thermal simulation model of the complex thermal behavior of the flywheel. Based on this, a lightweight reduced-order model is constructed using the calibrated finite element thermal simulation model. Preset location temperature data is input into the lightweight reduced-order model, which outputs the estimated current temperature of the rotor at the target location. This effectively overcomes the physical limitations of traditional detection methods, accurately capturing local hotspots formed on the rotor surface due to uneven eddy current losses, improving the accuracy and real-time performance of temperature detection. Furthermore, the computation time after training the lightweight reduced-order model is significantly reduced, fundamentally overcoming the limitation of easy accuracy decay in traditional open-loop prediction modes. This achieves high-precision, low-latency real-time estimation of the flywheel's overall temperature. This enables the flywheel to shift from reactive, post-event response to proactive early warning, effectively preventing serious failures such as thermal runaway and improving the safety and reliability of flywheel operation.

[0017] To achieve the above objectives, a second aspect of the present invention discloses a flywheel rotor temperature estimation system, comprising: an acquisition module, configured to acquire test temperature vectors of the flywheel under multiple preset operating conditions, and acquire preset position temperature data of the flywheel rotor during operation; a calibration module, configured to perform offline parameter calibration on a finite element thermal simulation model based on the test temperature vectors, to obtain a calibrated finite element thermal simulation model; and an estimation module, configured to construct a lightweight reduced-order model based on the calibrated finite element thermal simulation model, input the preset position temperature data as input parameters into the lightweight reduced-order model, and output the current estimated temperature of the rotor at the target position.

[0018] According to the flywheel rotor temperature estimation system of the present invention, the acquisition module acquires the test temperature vectors of the flywheel under multiple preset operating conditions and acquires the rotor temperature data of the flywheel at preset positions during operation. Subsequently, the calibration module performs offline parameter calibration on the finite element thermal simulation model based on the test temperature vectors to obtain the calibrated finite element thermal simulation model. This not only corrects the inherent deviation of the finite element thermal simulation model itself, but also overcomes the defect of traditional models that cannot adapt to actual operating conditions due to the use of fixed parameter settings. Thus, it avoids the continuous transmission of the original simulation error in the process of reducing the order of the finite element thermal simulation model from the source, and improves the reproduction accuracy of the finite element thermal simulation model of the complex thermal behavior of the flywheel. Based on this, the estimation module constructs a lightweight reduced-order model using the calibrated finite element thermal simulation model. It inputs preset location temperature data as parameters into the lightweight reduced-order model and outputs the estimated current temperature of the rotor at the target location. This effectively overcomes the physical limitations of traditional detection methods, accurately capturing local hotspots formed on the rotor surface due to uneven eddy current losses, improving the accuracy and real-time performance of temperature detection. Furthermore, the computation time after training the lightweight reduced-order model is significantly reduced, fundamentally overcoming the limitation of easy accuracy decay in traditional open-loop prediction modes. This achieves high-precision, low-latency real-time estimation of the flywheel's overall temperature. This enables the flywheel to shift from reactive, post-event response to proactive early warning, effectively preventing serious failures such as thermal runaway and improving the safety and reliability of flywheel operation.

[0019] Additional aspects and advantages of the invention will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention. Attached Figure Description

[0020] The above and / or additional aspects and advantages of the present invention will become apparent and readily understood from the description of the embodiments taken in conjunction with the following drawings, in which: Figure 1 This is a flowchart of a method for estimating the rotor temperature of a flywheel according to an embodiment of the present invention; Figure 2 This is a structural block diagram of a flywheel rotor temperature estimation system according to an embodiment of the present invention.

[0021] Explanation of reference numerals in the attached figures: 100-Flywheel Rotor Temperature Estimation System; 110 - Acquisition module; 120 - Calibration module; 130 - Estimation module. Detailed Implementation

[0022] The embodiments of the present invention are described in detail below. The embodiments described with reference to the accompanying drawings are exemplary. The embodiments of the present invention are described in detail below.

[0023] The following is for reference. Figure 1A method for estimating the rotor temperature of a flywheel according to an embodiment of the present invention is described.

[0024] like Figure 1 The diagram shows a flowchart of a flywheel rotor temperature estimation method according to an embodiment of the present invention. The flywheel rotor temperature estimation method includes at least steps S1-S3.

[0025] Step S1: Obtain the test temperature vector of the flywheel under multiple preset operating conditions, and obtain the preset position temperature data of the flywheel rotor during operation.

[0026] Among them, the preset location refers to an easy-to-measuring location where temperature monitoring is easy, and the preset location temperature refers to the real-time temperature of the easy-to-measuring location.

[0027] In this embodiment, the flywheel is, for example, a magnetically levitated flywheel. An array of contact temperature sensors is pre-arranged on the stationary components of the flywheel (such as the stator and bearings), while an infrared thermometer is used to non-contactly acquire the temperature distribution on the surface of the high-speed rotating flywheel rotor. By simultaneously collecting measured temperature data from these components under different preset operating conditions, multi-source temperature data is obtained. This measured temperature data is then integrated to generate multiple sets of test temperature vectors characterizing the overall thermal state of the flywheel. Furthermore, temperature data at preset positions of the flywheel rotor during operation is simultaneously acquired. By arranging multiple measuring points at different circumferential positions of the rotor, comprehensive and accurate temperature distribution information on the rotor surface is obtained, providing complete data support for the subsequent construction of a lightweight, reduced-order model.

[0028] Step S2: Based on the test temperature vector, perform offline parameter calibration on the finite element thermal simulation model to obtain the calibrated finite element thermal simulation model.

[0029] In this embodiment, the three-dimensional geometric model of the flywheel is simplified by removing small structures, sharp corners, rounded edges, and tiny protrusions—structures that have no impact on overall heat conduction—to reduce the difficulty of mesh generation and the overall computational load. This allows for the establishment of a finite element thermal simulation model of the flywheel, and the initial setting of key thermal parameters such as material properties, heat source compensation coefficient, contact thermal resistance, and shell convective heat transfer coefficient. Subsequently, the collected test temperature vector is compared point-by-point with the initial simulation results output by the finite element thermal simulation model to identify temperature deviations. Based on this, an iterative parameter tuning strategy is employed to optimize and adjust input parameters with high sensitivity or uncertainty in the finite element thermal simulation model within a physically reasonable range, and the simulation is rerun. Through this repeated calibration process, errors in the finite element thermal simulation model are gradually eliminated, ultimately controlling the overall simulation error to within 5%, thereby obtaining a high-fidelity finite element thermal simulation model that accurately reflects the true thermal characteristics of the flywheel. This calibration process effectively corrects the shortcomings of traditional models that cannot adapt to actual operating conditions due to the use of fixed parameter settings, significantly improves the accuracy of finite element thermal simulation models in reproducing the complex thermal behavior of flywheels, and provides an accurate physical benchmark for the subsequent construction of highly reliable lightweight reduced-order models.

[0030] Step S3: Construct a lightweight reduced-order model based on the calibrated finite element thermal simulation model, and input the preset position temperature data as input parameters into the lightweight reduced-order model to output the current estimated temperature of the rotor at the target position.

[0031] In the embodiment, although the simulation error of the calibrated finite element thermal simulation model can be controlled within 5%, and the offline accuracy fully meets the design verification requirements, its hundreds of thousands of degrees of freedom mean that the simulation of a single transient thermal field often takes several hours. This extremely high computational cost makes it unsuitable for the stringent requirements of flywheel embedded controllers in online operation scenarios such as millisecond-level real-time temperature monitoring, dynamic temperature rise early warning, and hardware-in-the-loop simulation.

[0032] To address this, the calibrated finite element thermal simulation model was downgraded to a lightweight, reduced-order model. This model takes the real-time temperatures of several easily measurable locations as input, i.e., preset location temperature data, and can output real-time temperature estimates for any target location within the entire domain, with the overall estimation error strictly controlled within 10%. Subsequently, this lightweight, reduced-order model was compiled into low-level code executable by the controller and deployed, enabling online temperature estimation of the flywheel rotor at the target location. This effectively overcomes the physical limitations of traditional detection methods, accurately capturing local hotspots formed on the rotor surface due to uneven eddy current losses, significantly improving the accuracy and real-time performance of temperature detection.

[0033] Furthermore, this lightweight reduced-order model possesses full-domain spatial extrapolation capabilities, enabling flexible estimation of the temperature at any target location inside the flywheel without the need for additional physical sensors. This provides comprehensive and reliable data support for the flywheel's thermal management and safety early warning, thus successfully achieving a closed-loop combination of offline high-precision simulation and online real-time estimation.

[0034] Therefore, by correcting the temperature vector through testing to eliminate the bias in the finite element thermal simulation model itself, and then performing order reduction processing based on the calibrated finite element thermal simulation model, the original simulation error is prevented from being continuously propagated during the lightweighting process of the finite element thermal simulation model. This processing not only significantly reduces the computation time after the lightweight and reduced-order model is trained, meeting the millisecond-level real-time requirements of the embedded controller, but also fundamentally breaks through the limitation of easy accuracy decay in traditional open-loop prediction modes, achieving high-precision, low-latency real-time estimation of the flywheel's full-domain temperature. This enables the flywheel to shift from reactive, post-event response to proactive early warning, effectively preventing serious failures such as thermal runaway, and significantly improving the safety and reliability of flywheel operation.

[0035] Therefore, the aforementioned method for estimating the rotor temperature of a flywheel obtains the test temperature vectors of the flywheel under multiple preset operating conditions and acquires the rotor temperature data at preset positions during operation. Subsequently, based on the test temperature vectors, the finite element thermal simulation model is calibrated offline to obtain the calibrated finite element thermal simulation model. This not only corrects the inherent biases of the finite element thermal simulation model itself but also overcomes the shortcomings of traditional models that cannot adapt to actual operating conditions due to fixed parameter settings. This avoids the continuous transmission of original simulation errors during the reduction process of the finite element thermal simulation model from the source, thereby improving the accuracy of the finite element thermal simulation model in reproducing the complex thermal behavior of the flywheel. Based on this, a lightweight reduced-order model is constructed using the calibrated finite element thermal simulation model. Preset location temperature data is input into the lightweight reduced-order model, which outputs the estimated current temperature of the rotor at the target location. This effectively overcomes the physical limitations of traditional detection methods, accurately capturing local hotspots formed on the rotor surface due to uneven eddy current losses, improving the accuracy and real-time performance of temperature detection. Furthermore, the computation time after training the lightweight reduced-order model is significantly reduced, fundamentally overcoming the limitation of easy accuracy decay in traditional open-loop prediction modes. This achieves high-precision, low-latency real-time estimation of the flywheel's overall temperature. This enables the flywheel to shift from reactive, post-event response to proactive early warning, effectively preventing serious failures such as thermal runaway and improving the safety and reliability of flywheel operation.

[0036] In one embodiment of the present invention, when performing offline parameter calibration on the finite element thermal simulation model based on the test temperature vector to obtain the calibrated finite element thermal simulation model, the process includes: obtaining the initial simulation results of the finite element thermal simulation model and comparing and verifying the initial simulation results with the test temperature vector point by point; performing partitioned evaluation of the point-by-point comparison verification according to preset component regions, and determining the partitioning error in each preset component region according to a preset algorithm; performing multi-condition calibration of the heat source compensation coefficient and / or iterative adjustment of the contact thermal resistance used to characterize the heat transfer path; after completing the multi-condition calibration of the heat source compensation coefficient and / or iterative adjustment of the contact thermal resistance used to characterize the heat transfer path, determining whether the partitioning error meets the preset convergence condition; if so, adjusting the convective heat transfer coefficient of the flywheel shell to obtain the calibrated finite element thermal simulation model; wherein, the preset convergence condition includes: the partitioning error is less than a preset convergence threshold.

[0037] In this embodiment, the initial simulation results of the finite element thermal simulation model are obtained, and then compared and verified point by point with the test temperature vector at the same time and location. This verification process is performed by evaluating preset component regions (such as stator region, bearing region, rotor surface region, and housing region), and the root mean square error and maximum absolute error of each preset component region are calculated as the partition error. This allows for precise location of specific physical regions where the deviations of the finite element thermal simulation model are concentrated, thus providing a clear calibration direction for subsequent targeted adjustments to the heat source compensation coefficient or contact thermal resistance of the corresponding preset component regions. This effectively avoids error coupling caused by blindly adjusting parameters globally, significantly improving the convergence speed and accuracy of the finite element thermal simulation model calibration.

[0038] Subsequently, a local sensitivity analysis was conducted using the controlled variable method. By shifting a single parameter by ±10% and observing the temperature change at each measurement point, the sensitivity levels of various heat source parameters of the flywheel were classified. The analysis results show that the sensitivity of each heat source parameter, from highest to lowest, is as follows: heat source compensation coefficient, contact thermal resistance, material thermal conductivity, shell convective heat transfer coefficient, and surface radiation coefficient.

[0039] Specifically, the heat source compensation coefficient has the highest sensitivity because it directly determines the magnitude of the eddy current loss heat source of the entire flywheel, and thus the temperature reference value of all temperature measurement points. Next is the contact thermal resistance of key contact interfaces such as the stator and shell, and the bearing housing and shell, which mainly affects the efficiency of the core heat transfer path. The material thermal conductivity is usually a factory-measured parameter and has low sensitivity under normal operating conditions, only having a significant impact when there are large local temperature differences. The shell convection heat transfer coefficient has relatively low sensitivity and is usually used as the last adjustment item to eliminate deviations at all measurement points. The surface emissivity is fixed as a constant during calibration because the flywheel is uniformly coated with a high emissivity thermally conductive paint (emissivity ≥ 0.85), and the cavity is a vacuum environment that does not easily cause surface oxidation. Based on the above sensitivity ranking, the calibration process follows the principle of "core first, then boundary," prioritizing the adjustment of the heat source compensation coefficient and contact thermal resistance, which are highly sensitive, and finally adjusting the shell convection heat transfer coefficient, which is less sensitive. This effectively avoids the adjustment of low-sensitivity parameters from interfering with the calibration accuracy of the heat source compensation coefficient and contact thermal resistance.

[0040] Specifically, when performing multi-condition calibration of the heat source compensation coefficient and / or iterative adjustment of the contact thermal resistance used to characterize the heat transfer path, the adjustment follows a priority strategy from high to low based on sensitivity levels. Specifically, the heat source compensation coefficient with the highest sensitivity is calibrated first; if the contact thermal resistance also has deviations requiring adjustment, iterative correction of the contact thermal resistance is performed after the optimization of the heat source compensation coefficient, avoiding coupling interference between low-sensitivity parameter adjustments and the core heat source calibration.

[0041] Furthermore, after completing the multi-condition calibration of the heat source compensation coefficient and / or the iterative adjustment of the contact thermal resistance used to characterize the heat transfer path, it is determined whether the partition error is less than the preset convergence threshold (e.g., 5%). If the partition error is less than 5%, it is determined that the preset convergence condition is met. Then, the convective heat transfer coefficient of the flywheel shell is finely adjusted to obtain the calibrated finite element thermal simulation model.

[0042] In addition, if the partitioning error is greater than 5%, the heat source parameters corresponding to the preset component area where the partitioning error is located will be adjusted. For example, if the partitioning error of the bearing area is large, the contact thermal resistance of the bearing will be adjusted first; if the partitioning error of the rotor surface area is large, the heat source compensation coefficient will be adjusted first. Through the above iterative optimization process, the partitioning error of all preset component areas will meet the preset convergence conditions.

[0043] Therefore, it can effectively avoid the interference of low-sensitivity external heat dissipation boundary parameters such as heat source compensation coefficient and contact thermal resistance, improve the overall accuracy and convergence speed of the finite element thermal simulation model, and thus improve the accuracy and real-time performance of temperature detection.

[0044] In one embodiment of the present invention, obtaining the initial simulation results of the finite element thermal simulation model includes: assigning corresponding electromagnetic parameters and thermal property parameters to each functional component in the finite element thermal simulation model; determining the loss of each functional component based on the electromagnetic parameters; using the loss of each functional component as the heat source input of the finite element thermal simulation model; and solving the problem according to a preset thermal field algorithm based on the thermal property parameters, the heat source input, and the preset boundary conditions of the finite element thermal simulation model to obtain the initial simulation results of the finite element thermal simulation model.

[0045] In this embodiment, each functional component in the finite element thermal simulation model is assigned corresponding material properties, including electromagnetic parameters and thermal properties. The functional components include, but are not limited to, stators, rotors, rotor sheaths, magnets, housings, bearings, end caps, and energy storage bodies. The electromagnetic parameters include, but are not limited to, electrical conductivity, iron loss curves, magnetization curves, remanence, and coercivity. The thermal properties include, but are not limited to, density, thermal conductivity, and constant-pressure heat capacity. The thermal conductivity needs to be set to isotropic or anisotropic based on the actual winding and lamination structure.

[0046] Subsequently, based on the different working conditions of the actual test, the stator iron loss, stator copper loss, rotor sheath and permanent magnet eddy current loss obtained from the electromagnetic simulation solution were used as the heat source input of the finite element thermal simulation model. Considering that the initial values ​​of rotor sheath and permanent magnet eddy current loss may be deviated due to current harmonics and material nonlinearity, they will be corrected by heat source compensation coefficient in the future.

[0047] Finally, various preset boundary conditions were reasonably set: contact thermal resistance was set for the contact surfaces of each functional component, and the values ​​were estimated and assigned using classical contact theory formulas. Specifically, corresponding contact thermal resistances were set at the contact interfaces between the stator and the housing, the bearing and the end cover, the end cover and the housing, the rotor sheath and the magnet, the rotor sheath and the energy storage body, and the magnet and the energy storage body; the surface emissivity was set for the outer surfaces of each functional component, with the surface emissivity of the area coated with special thermally conductive paint set to be greater than or equal to 0.85, and the values ​​for the remaining exposed metal surfaces determined according to the actual state such as polishing or oxidation, for example, the value range for polished surfaces is 0.3~0.5; given that the flywheel is in a vacuum environment, the convective heat transfer between functional components can be ignored, and only heat conduction and heat radiation are considered; the convective heat transfer coefficient of the flywheel housing is calculated using an empirical formula based on actual air-cooling conditions; at the same time, the ambient temperature is set based on the measured value of the room temperature sensor. After completing the above settings, the initial simulation results of the temperature of each node are obtained by solving the preset thermal field algorithm.

[0048] By defining the material properties and setting the boundary conditions in a refined manner, the actual thermophysical behavior of the flywheel under complex working conditions can be accurately reproduced, thereby significantly improving the accuracy and reliability of the overall thermal field calculation and thus improving the accuracy of the finite element thermal simulation model calibration.

[0049] In one embodiment of the present invention, when performing multi-condition calibration of the heat source compensation coefficient, the process includes: acquiring the rotor speed of the flywheel, the charging and discharging power of the flywheel, and the rotor operating temperature; and determining the heat source compensation coefficient based on the partition error, rotor speed, charging and discharging power, and rotor operating temperature by means of a preset fitting algorithm and in accordance with a preset differential operating condition calibration strategy. The heat source compensation coefficient includes a reference coefficient, a speed influence factor, and a temperature influence factor.

[0050] For example, the benchmark coefficient is denoted as The speed influence factor is denoted as The temperature influence factor is denoted as The rotor speed is denoted as n, and the rotor operating temperature is denoted as... .

[0051] In this embodiment, the heat source compensation coefficient is used to accurately correct the initial values ​​of rotor sheath and permanent magnet eddy current losses. The rotor speed, charge / discharge power, and rotor operating temperature of the flywheel are obtained; based on the partitioning error, rotor speed, charge / discharge power, and rotor operating temperature, a preset fitting algorithm and a preset differential operating condition calibration strategy are used to determine the baseline coefficient, speed influence factor, and temperature influence factor.

[0052] Specifically, for the first operating condition (low speed, low load reference condition): under this condition, the rotor's eddy current loss accounts for a very small proportion of the total loss, while the stator copper loss dominates, unaffected by eddy current losses caused by rotor speed and temperature rise. The first operating condition is used to calibrate the reference coefficient under no-coupling interference. This is to accurately compensate for the systematic deviations caused by the inherent material conductivity parameters and simulation mesh generation.

[0053] Second operating condition (rated speed, standard operating condition at normal temperature): Under the second operating condition, the rotor operating speed reaches the rated value, the rotor eddy current loss increases rapidly, and the amplification effect of rotor speed on eddy current loss becomes prominent. The second operating condition is used to separately calibrate the speed influence factor. The two satisfy the following relationship: ,in This is the rated speed of the flywheel.

[0054] Third operating condition (rated speed, high-temperature extreme condition): The flywheel is controlled to run continuously at rated speed until the measured temperature on the rotor surface reaches a stable value above 100°C. The purpose of this third operating condition is to isolate the interference from speed variables and specifically capture the eddy current loss deviation caused by the decrease in material conductivity under high-temperature conditions, for the purpose of calibrating the temperature influence factor. Finally, the compensation coefficient formula for the two-way coupling of rotational speed and temperature is constructed as follows: ,in, The room temperature is the standard room temperature in the laboratory.

[0055] Unlike single-temperature-point fitting methods, this invention simultaneously collects measured temperature data from three key heat source points on the stator, bearing, and rotor surfaces under three different operating conditions. It incorporates charging / discharging power and zoning errors into the fitting objective and simultaneously solves for the baseline coefficients using a weighted least squares fitting algorithm. Rotation speed influencing factor Temperature influence factors Three core parameters. The calibration of the heat source compensation coefficient is completed in one go during the flywheel offline testing phase, eliminating the need for online real-time iterative correction. The calibrated continuous function can directly adapt to the flywheel's full speed and full temperature range operating conditions, significantly reducing the error in the global electromagnetic thermal simulation. This effectively solves the simulation distortion problem caused by simple superposition under a single speed or temperature dimension in existing technologies, significantly improving the accuracy of thermal field prediction and model robustness of the flywheel under complex variable operating conditions, and thus improving the accuracy and real-time performance of temperature detection.

[0056] In one embodiment of the present invention, when performing iterative adjustment of contact thermal resistance for characterizing heat transfer paths, the method includes: sequentially performing iterative calibration on the contact thermal resistance of each target contact interface according to a preset heat transfer path sequence; during the iterative calibration process for the current target contact interface, obtaining the simulation temperature of the current target contact interface output by the finite element thermal simulation model, and the actual temperature of the current target contact interface; determining the temperature deviation between the actual temperature and the simulation temperature; using the temperature deviation as the iterative objective function, solving and correcting the contact pressure of the current target contact interface through a preset algorithm, so as to update the contact thermal resistance of the current target contact interface through the change of contact pressure; when the temperature deviation is less than a preset temperature difference threshold, determining that the contact thermal resistance calibration of the current target contact interface is complete, and jumping to the next target contact interface to continue performing iterative calibration.

[0057] In this embodiment, the accuracy of the flywheel's thermal simulation is highly dependent on the contact thermal resistance of the assembly interface. To address the shortcomings of traditional global parallel iteration methods, which easily lead to mutual coupling interference among multiple contact thermal resistance parameters, severe oscillations in the iteration curve, slow convergence speed, and distortion of calibration results, a point-by-point serial iteration locking strategy based on the physical path of heat transfer is employed to achieve decoupled, high-precision independent calibration of the contact thermal resistance of each interface.

[0058] Specifically, the initial value of the contact thermal resistance is solved directly using the classical microscopic contact mechanism model, and its core physical expression is: ,in, For contact thermal resistance, Let A be the equivalent contact heat transfer coefficient at the interface, and A be the actual effective contact area. The input parameters of this classical microscopic contact mechanism model include the surface roughness of the contact surface, the hardness of the substrate material, and the initial assembly contact pressure. It does not rely on measured temperature data, thus ensuring that the preset boundary conditions of the finite element thermal simulation model have sufficient physical basis.

[0059] Contact pressure and contact thermal resistance exhibit a strong negative correlation and a monotonic variation law: Where P is the contact pressure, i.e., the contact thermal resistance is a single-valued function of the contact pressure. The greater the contact pressure, the tighter the contact interface fits, and the smaller the contact thermal resistance. During the calibration iteration process, for example, using a bearing as the current target contact interface, the simulated temperature of the bearing output from the finite element thermal simulation model, as well as the actual temperature of the bearing, are obtained. The temperature deviation between the actual and simulated temperatures of the bearing is used as the iterative objective function, and preset algorithms such as the bisection method or Newton's iteration method are selected for inverse closed-loop solution. Unlike the conventional method of directly modifying the contact thermal resistance value, this method corrects the contact pressure of the current target contact interface through inverse iteration. The change in contact pressure is used to equivalently correct the bolt preload deviation, thereby indirectly matching the contact thermal resistance under actual working conditions. The contact thermal resistance of the current target contact interface is updated by the change in contact pressure.

[0060] To completely eliminate the thermal coupling oscillation problem caused by multi-parameter synchronous iteration, the contact thermal resistance of each target contact interface will be adjusted step by step according to a preset heat conduction path sequence, that is, following the inherent heat transfer path from the heat source to the outer casing inside the flywheel. The fixed iteration sequence is as follows: stator-outer casing contact surface, bearing housing-end cover contact surface, and outer casing-end cover contact surface. Each iteration adjusts only the contact thermal resistance parameter of a single contact interface, without multi-parameter synchronous linkage modification throughout the entire process.

[0061] Simultaneously, a unified iterative convergence threshold is set, i.e., a preset temperature difference threshold, for example, 1℃. Specifically, when the temperature deviation between the actual temperature and the simulated temperature is less than 1℃, the contact thermal resistance calibration of the current target contact interface is considered complete. The contact thermal resistance of the current target contact interface is permanently locked and no longer participates in subsequent iterations. Then, the process jumps to the next target contact interface to continue iterative calibration. This not only improves the convergence speed of the finite element thermal simulation model calibration but also avoids multi-parameter coupling interference, significantly improving the stability and calculation accuracy of the finite element thermal simulation model, thereby enhancing the accuracy and real-time performance of temperature estimation.

[0062] In one embodiment of the present invention, when adjusting the convective heat transfer coefficient of the flywheel casing to obtain a calibrated finite element thermal simulation model, the process includes: obtaining the casing simulation temperature output by the finite element thermal simulation model and obtaining the actual casing temperature; based on the casing simulation temperature and the actual casing temperature, determining whether the temperature difference change trend between the casing simulation temperature and the actual casing temperature is in the same direction; if so, determining the initial casing convective heat transfer coefficient based on a preset heat transfer formula, and iteratively adjusting the initial casing convective heat transfer coefficient according to a preset adjustment range; when iteratively adjusting the initial casing convective heat transfer coefficient according to the preset adjustment range, obtaining the maximum absolute error between the casing simulation temperature and the actual casing temperature; when the maximum absolute error is less than a preset fine-tuning convergence threshold, stopping the iterative adjustment of the initial casing convective heat transfer coefficient to obtain the calibrated finite element thermal simulation model.

[0063] In the embodiment, after completing the multi-condition calibration of the heat source compensation coefficient and / or the iterative adjustment of the contact thermal resistance used to characterize the heat transfer path, and after the partition error meets the preset convergence condition, but there is still a systematic overall deviation between the measured temperature and the simulated temperature of all thermocouples on the flywheel housing (i.e., the simulated temperature of all measuring points is uniformly higher or lower than the measured temperature), rather than a single point random local error, the fine-tuning operation of the convective heat transfer coefficient of the housing is initiated to avoid prematurely adjusting the heat dissipation boundary and interfering with the calibration accuracy of the internal heat source and contact thermal resistance.

[0064] Specifically, the simulated shell temperature output from the finite element thermal simulation model and the actual shell temperature are obtained. Based on the simulated and actual shell temperatures, it is determined whether the temperature difference between the simulated and actual shell temperatures follows the same trend. If the temperature difference follows the same trend, the initial shell convective heat transfer coefficient is determined based on a preset heat transfer formula. This initial shell convective heat transfer coefficient is calculated based on the actual air-cooled flywheel operation, using the classic Gnielinski formula for turbulent heat transfer in pipes, which is widely used in the field of heat transfer. This ensures that the initial heat transfer boundary has a theoretical basis and does not require additional simulation calculations. The formula and the corresponding drag coefficient calculation formula are as follows: .

[0065] in, is the Nusselt number, which represents the ratio of convective heat transfer intensity to thermal conductivity. It is a dimensionless heat transfer coefficient used to measure the heat transfer efficiency between the fluid and the wall; f is the friction coefficient. The Reynolds number characterizes the fluid flow state (laminar or turbulent) and is defined as the ratio of inertial force to viscous force. The calculation formula is Re = ρvd / μ, where ρ is the fluid density, v is the flow velocity, d is the pipe diameter, and μ is the dynamic viscosity. Prandtl number, reflecting the ratio of a fluid's momentum diffusivity to its thermal diffusivity, is a fluid property parameter, calculated using the formula Pr = c. p ×μ / λ, where c p λ is the specific heat capacity, λ is the thermal conductivity, d is the inner diameter of the tube, and I is the tube length.

[0066] Subsequently, the initial shell convective heat transfer coefficient is iteratively adjusted according to a preset adjustment range, i.e., the adjustment range is controlled within ±20% of the initial shell convective heat transfer coefficient, to prevent the initial shell convective heat transfer coefficient from deviating from physical rationality. To ensure that the fine-tuning process does not damage the calibrated internal heat source parameters and contact thermal resistance parameters, and to avoid the heat transfer coefficient deviating from the actual physical operating conditions.

[0067] During the iterative adjustment process, the maximum absolute error between the simulated temperature of the shell and the actual temperature of the shell is obtained in real time. When the maximum absolute error is less than the preset fine-tuning convergence threshold (e.g., 2℃), the iterative adjustment of the initial shell convection heat transfer coefficient is stopped, thereby obtaining the calibrated finite element thermal simulation model.

[0068] Given the low sensitivity of this parameter, secondary adjustments are generally unnecessary, and fine-tuning is only required when there is significant uncertainty in the cooling boundary conditions. After the entire calibration process is completed, the heat source compensation parameters, all iteratively completed contact thermal resistances, and the final fine-tuned shell convective heat transfer coefficient are packaged and saved together to form a thermal simulation parameter set adapted to the actual assembly state of the flywheel and air-cooled conditions. This parameter set can be directly called when performing full-domain simulations without repeated calibration, thereby improving the adaptability of the finite element thermal simulation model to complex assembly tolerances and variable air-cooled environments. At the same time, by solidifying the dedicated parameter set, the high-precision finite element thermal simulation model can be standardized and reused, reducing the repetitive calculation time for subsequent multi-condition finite element thermal simulation model analysis.

[0069] In one embodiment of the present invention, when constructing a lightweight reduced-order model based on a calibrated finite element thermal simulation model, and inputting preset position temperature data as input parameters into the lightweight reduced-order model to output the current estimated temperature of the rotor at the target position, the process includes: performing preset transient thermal field simulation on the calibrated finite element thermal simulation model based on multiple preset operating conditions, collecting time-domain temperature response data of each target monitoring point within the calibrated finite element thermal simulation model, and constructing a multi-condition temperature snapshot sample set; based on the multi-condition temperature snapshot sample set, using multiple preset operating conditions as input variables and the current estimated temperature of the rotor at the target position as the output variable, constructing a lightweight reduced-order model characterizing the mapping relationship between the preset operating conditions and the time-domain temperature response data; when the model deviation of the lightweight reduced-order model relative to the calibrated finite element thermal simulation model is less than or equal to a preset reduction error threshold, determining that the lightweight reduced-order model construction is complete; and inputting the preset position temperature data into the constructed lightweight reduced-order model to output the current estimated temperature of the rotor at the target position.

[0070] In this embodiment, based on multiple sets of preset operating conditions, a preset transient thermal field simulation is performed on the finite element thermal simulation model that has undergone high-precision calibration. Preset operating conditions covering the entire operating range of the flywheel are selected, and calculations are performed for each set of preset operating condition combinations. Temperature response data of each target monitoring point (including easily measurable points and internal hot spots of the rotor) within the finite element thermal simulation model are collected over a complete time interval, thereby constructing a multi-condition temperature snapshot sample set.

[0071] The preset operating condition, for example, selects a typical, stringent 1 / 3 duty cycle (180s cycle, 33.3% of total charge / discharge time) to fully cover the flywheel's thermal load limits. This preset operating condition executes a "discharge-stop-charge-stop" reciprocating cycle under limited heat dissipation conditions: first, a deep discharge of 0s~30s occurs, with the flywheel forcibly releasing energy from its maximum speed to the cutoff speed at a constant maximum power; then, a first no-load cooling period of 30s~90s occurs, during which the flywheel maintains its speed in a vacuum environment through inertia; immediately following, forced charging occurs from 90s~120s, with the flywheel switching to electric mode and rapidly increasing its speed from the cutoff value back to the maximum speed at the maximum allowable power; finally, during the second no-load cooling period of 120s~180s, the flywheel remains suspended at the maximum speed until the start of the next cycle. This preset operating condition, by simulating the combined effect of high power density impact and weak heat dissipation conditions, effectively verifies the model's dynamic response characteristics under extreme thermal stress.

[0072] Subsequently, based on this multi-condition temperature snapshot sample set, using multiple sets of preset operating conditions as input variables and the current estimated temperature of the rotor at the target position as the output variable, a lightweight reduced-order model representing the explicit mapping relationship between the preset operating conditions and the time-domain temperature response data is established using numerical fitting. This lightweight reduced-order model adopts static data-driven technology, and after training, it no longer relies on complex finite element solvers and mesh models, but can complete rapid calculations solely based on the pre-constructed mapping relationship.

[0073] Next, the accuracy of the constructed lightweight reduced-order model is verified. Cross-validation tools are used to perform verification using indicators such as coefficient of determination and root mean square error. When the model deviation of the lightweight reduced-order model relative to the calibrated finite element thermal simulation model is less than or equal to the preset reduction error threshold (e.g., 5%), the lightweight reduced-order model is considered to have been successfully constructed.

[0074] Finally, by inputting the preset position temperature data into the constructed lightweight reduced-order model, the current estimated temperature of the rotor at the target position can be output, thereby realizing online temperature estimation of the flywheel rotor at the target position. This effectively breaks through the physical limitations of traditional detection methods, accurately captures local hot spots formed on the rotor surface due to uneven distribution of eddy current losses, and greatly improves the accuracy and real-time performance of temperature detection.

[0075] This lightweight, reduced-order model can be exported as a functional unit file conforming to the FMI (Functional Mock-up Interface) protocol and converted into C / C++ code, which can then be embedded into embedded controllers such as DSPs (Digital Signal Processors), ARMs (Advanced RISC Machines), or FPGAs (Field-Programmable Gate Arrays). Compared to the complete finite element thermal simulation model, its computation time is reduced by more than 99%, and its computational complexity is independent of the original number of nodes. It can stably achieve millisecond-level global temperature estimation, balancing simulation accuracy and operational efficiency.

[0076] In one embodiment of the present invention, when obtaining the test temperature vector of the flywheel under multiple preset operating conditions, the method includes: continuously collecting multi-source temperature data under multiple preset operating conditions, wherein the multi-source temperature data includes stator temperature, bearing temperature, rotor temperature, casing temperature, ambient temperature and fluid medium temperature; preprocessing and time-synchronizing the collected multi-source temperature data to construct the test temperature vector under each preset operating condition.

[0077] In this embodiment, the test temperature vector of the flywheel under multiple preset operating conditions is obtained through a multi-source temperature sensing system customized for harsh conditions such as high vacuum, strong magnetic field, high-speed rotation, and compact space. This multi-source temperature sensing system employs a hybrid sensing network architecture combining "contact point measurement + non-contact surface measurement," enabling comprehensive temperature acquisition of key components such as the stator, bearings, rotor, housing, environment, and fluid medium. For temperature acquisition of stationary components (such as the stator, bearings, and housing), the multi-source temperature sensing system evenly distributes three embedded PT100 platinum resistance thermometers (e.g., buried at a depth of 3mm) circumferentially at the non-drive end of the stator winding, and embeds a K-type thermocouple in each of the upper and lower radial magnetic bearing stator tooth slots. All sensors are filled with thermally conductive silicone grease with a thermal conductivity of not less than 1.5 W / (m·K) to eliminate contact thermal resistance. To cope with strong electromagnetic interference, the lead wires of the contact sensors all adopt double-layer shielded cables of "copper foil + braided mesh" and implement single-end grounding. With the help of twisted pair transmission, differential amplification and 50Hz notch filter circuit, the high-frequency common-mode interference between the frequency converter and the magnetic bearing is effectively suppressed, ensuring a high signal-to-noise ratio for temperature measurement of stationary components.

[0078] For high-speed rotors that cannot be directly contacted, the multi-source temperature sensing system employs a split-type infrared temperature measurement device. The built-in sensor head is compact, fixed within the vacuum chamber, and precisely aligned with the hot spot area on the rotor surface. The lead wire passes through a spare hole in the vacuum pump flange and is hermetically sealed using a vacuum-sealed electrical connector, without damaging the chamber structure or affecting dynamic balance. To resist strong magnetic field interference, the sensor head cable is encased in a metal shielding sleeve and reliably grounded. The communication processing unit is located in an external weak magnetic field zone, exchanging data via a digital communication protocol. Furthermore, the communication unit incorporates peak hold and programmable filtering algorithms, combined with a multi-point layout of up to eight probes, to realistically reproduce the temperature distribution field on the rotor surface, achieving non-contact, continuous, and high-precision monitoring. Simultaneously, auxiliary patch sensors are placed at key locations along the vacuum pump interface, cooling inlet and outlet, and the axial direction of the casing to monitor the fluid medium temperature and casing heat distribution in real time, providing accurate boundary condition inputs for the thermal simulation model.

[0079] After hardware deployment, the multi-source temperature sensing system comprehensively considers variables such as rotational speed, load, cooling medium flow rate, and ambient temperature, selecting multiple typical preset operating conditions for continuous data acquisition. This yields multi-source temperature data including stator temperature, bearing temperature, rotor temperature, casing temperature, ambient temperature, and fluid medium temperature. After acquisition, all multi-source temperature data undergoes high-precision time synchronization alignment, constructing a standardized test temperature vector that strictly corresponds to each preset operating condition. This not only allows for continuous acquisition of full-cycle data without stopping the flywheel or disrupting the vacuum environment, avoiding the natural cooling errors associated with traditional disassembly temperature measurement, but also accurately captures instantaneous temperature rise peaks caused by operating condition fluctuations. This provides high-fidelity, high-temporal-resolution benchmark data for subsequent finite element thermal simulation model calibration and order reduction analysis, thus preventing the continuous propagation of original simulation errors during the finite element thermal simulation model lightweighting process. This achieves high-precision, low-latency real-time estimation of the flywheel's overall temperature.

[0080] In one embodiment of the present invention, when acquiring the temperature data of the flywheel at a preset position during operation, the method includes: acquiring the initial temperature data of the flywheel at a preset position with a preset sampling period; filtering the initial temperature data to obtain an effective temperature measurement result; and using the effective temperature measurement result as the temperature data of the rotor at the preset position during operation.

[0081] In this embodiment, a preset sampling period of 100ms is used to continuously collect raw temperature data at a preset position of the flywheel within a single control cycle as initial temperature data. Secondly, to address the interference issues easily caused by the strong electromagnetic environment of the high-speed rotating rotor, the multi-source temperature sensing system performs programmable filtering on the initial temperature data. Specifically, through built-in peak hold, valley hold, and programmable averaging filtering functions, the mean or median of the multiple initial temperature data is calculated, thereby effectively filtering out measurement errors caused by instantaneous electromagnetic spike noise and obtaining accurate and effective temperature measurement results. Finally, this effective temperature measurement result is used as the preset position temperature data of the rotor during operation. This not only ensures the anti-interference capability of non-contact temperature measurement under complex working conditions but also ensures the authenticity and high fidelity of the temperature data, providing reliable data support for subsequent accurate capture of local hot spots on the rotor surface and calibration of the finite element thermal simulation model.

[0082] According to the flywheel rotor temperature estimation method of the present invention, test temperature vectors of the flywheel under multiple preset operating conditions are obtained, and the rotor temperature data of the flywheel at preset positions during operation are obtained. Then, based on the test temperature vectors, offline parameter calibration is performed on the finite element thermal simulation model to obtain the calibrated finite element thermal simulation model. This not only corrects the inherent deviation of the finite element thermal simulation model itself, but also overcomes the defect of traditional models that cannot adapt to actual operating conditions due to the use of fixed parameter settings. Thus, it avoids the continuous transmission of the original simulation error in the process of reducing the order of the finite element thermal simulation model from the source, and improves the reproduction accuracy of the finite element thermal simulation model of the complex thermal behavior of the flywheel. Based on this, a lightweight reduced-order model is constructed using the calibrated finite element thermal simulation model. Preset location temperature data is input into the lightweight reduced-order model, which outputs the estimated current temperature of the rotor at the target location. This effectively overcomes the physical limitations of traditional detection methods, accurately capturing local hotspots formed on the rotor surface due to uneven eddy current losses, improving the accuracy and real-time performance of temperature detection. Furthermore, the computation time after training the lightweight reduced-order model is significantly reduced, fundamentally overcoming the limitation of easy accuracy decay in traditional open-loop prediction modes. This achieves high-precision, low-latency real-time estimation of the flywheel's overall temperature. This enables the flywheel to shift from reactive, post-event response to proactive early warning, effectively preventing serious failures such as thermal runaway and improving the safety and reliability of flywheel operation.

[0083] A further embodiment of the present invention discloses a flywheel rotor temperature estimation system.

[0084] like Figure 2 The diagram shown is a structural block diagram of the rotor temperature estimation system for a flywheel according to an embodiment of the present invention.

[0085] like Figure 2 As shown, the rotor temperature estimation system 100 of the flywheel includes: an acquisition module 110, a calibration module 120, and an estimation module 130.

[0086] The acquisition module 110 is used to acquire the test temperature vector of the flywheel under multiple preset operating conditions and acquire the preset position temperature data of the flywheel rotor during operation; the calibration module 120 is used to perform offline parameter calibration of the finite element thermal simulation model according to the test temperature vector to obtain the calibrated finite element thermal simulation model; the estimation module 130 is used to construct a lightweight reduced-order model based on the calibrated finite element thermal simulation model, and input the preset position temperature data as input parameters into the lightweight reduced-order model, and output the current estimated temperature of the rotor at the target position.

[0087] In one embodiment of the present invention, when the calibration module 120 performs offline parameter calibration on the finite element thermal simulation model according to the test temperature vector to obtain the calibrated finite element thermal simulation model, it is used to: obtain the initial simulation results of the finite element thermal simulation model and perform point-by-point comparison and verification between the initial simulation results and the test temperature vector; perform partitioned evaluation of the point-by-point comparison and verification according to the preset component regions, and determine the partitioning error in each preset component region according to the preset algorithm; perform multi-condition calibration of the heat source compensation coefficient and / or iterative adjustment of the contact thermal resistance used to characterize the heat transfer path; after completing the multi-condition calibration of the heat source compensation coefficient and / or iterative adjustment of the contact thermal resistance used to characterize the heat transfer path, determine whether the partitioning error meets the preset convergence condition; if so, adjust the convective heat transfer coefficient of the flywheel shell to obtain the calibrated finite element thermal simulation model; wherein, the preset convergence condition includes: the partitioning error is less than the preset convergence threshold.

[0088] In one embodiment of the present invention, when the calibration module 120 obtains the initial simulation results of the finite element thermal simulation model, it is used to: assign corresponding electromagnetic parameters and thermal property parameters to each functional component in the finite element thermal simulation model; determine the loss of each functional component based on the electromagnetic parameters; use the loss of each functional component as the heat source input of the finite element thermal simulation model; and solve the problem according to the preset thermal field algorithm based on the thermal property parameters, the heat source input and the preset boundary conditions of the finite element thermal simulation model to obtain the initial simulation results of the finite element thermal simulation model.

[0089] In one embodiment of the present invention, when performing multi-condition calibration of the heat source compensation coefficient, the calibration module 120 is used to: obtain the rotor speed of the flywheel, the charging and discharging power of the flywheel, and the rotor operating temperature; and determine the heat source compensation coefficient based on the partition error, rotor speed, charging and discharging power, and rotor operating temperature, through a preset fitting algorithm and according to a preset differential operating condition calibration strategy, wherein the heat source compensation coefficient includes a reference coefficient, a speed influence factor, and a temperature influence factor.

[0090] In one embodiment of the present invention, when the calibration module 120 performs iterative adjustment of the contact thermal resistance for characterizing the heat transfer path, it is configured to: sequentially perform iterative calibration on the contact thermal resistance of each target contact interface according to a preset heat transfer path sequence; during the iterative calibration process for the current target contact interface, obtain the simulation temperature of the current target contact interface output by the finite element thermal simulation model, and the actual temperature of the current target contact interface; determine the temperature deviation between the actual temperature and the simulation temperature; use the temperature deviation as the iterative objective function, solve it in reverse using a preset algorithm, and correct the contact pressure of the current target contact interface so as to update the contact thermal resistance of the current target contact interface through the change of contact pressure; when the temperature deviation is less than a preset temperature difference threshold, determine that the contact thermal resistance calibration of the current target contact interface is completed, and jump to the next target contact interface to continue performing iterative calibration.

[0091] In one embodiment of the present invention, when the calibration module 120 adjusts the convective heat transfer coefficient of the flywheel shell to obtain a calibrated finite element thermal simulation model, it is used to: obtain the shell simulation temperature output by the finite element thermal simulation model and obtain the actual shell temperature; based on the shell simulation temperature and the actual shell temperature, determine whether the temperature difference change trend between the shell simulation temperature and the actual shell temperature is in the same direction; if so, determine the initial shell convective heat transfer coefficient based on a preset heat transfer formula, and iteratively adjust the initial shell convective heat transfer coefficient according to a preset adjustment range; when iteratively adjusting the initial shell convective heat transfer coefficient according to the preset adjustment range, obtain the maximum absolute error between the shell simulation temperature and the actual shell temperature; when the maximum absolute error is less than a preset fine-tuning convergence threshold, stop the iterative adjustment of the initial shell convective heat transfer coefficient to obtain the calibrated finite element thermal simulation model.

[0092] In one embodiment of the present invention, when the estimation module 130 constructs a lightweight reduced-order model based on the calibrated finite element thermal simulation model, and inputs preset position temperature data as input parameters into the lightweight reduced-order model, and outputs the current estimated temperature of the rotor at the target position, the module is configured to: perform preset transient thermal field simulation on the calibrated finite element thermal simulation model based on multiple preset operating conditions, collect time-domain temperature response data of each target monitoring point within the calibrated finite element thermal simulation model, and construct a multi-condition temperature snapshot sample set; based on the multi-condition temperature snapshot sample set, construct a lightweight reduced-order model characterizing the mapping relationship between the preset operating conditions and the time-domain temperature response data, using multiple preset operating conditions as input variables and the current estimated temperature of the rotor at the target position as the output variable; determine that the lightweight reduced-order model is constructed when the model deviation of the lightweight reduced-order model relative to the calibrated finite element thermal simulation model is less than or equal to a preset reduction error threshold; and input the preset position temperature data into the constructed lightweight reduced-order model to output the current estimated temperature of the rotor at the target position.

[0093] In one embodiment of the present invention, when the acquisition module 110 acquires the test temperature vector of the flywheel under multiple preset operating conditions, it is used to: continuously collect multi-source temperature data under multiple preset operating conditions, wherein the multi-source temperature data includes stator temperature, bearing temperature, rotor temperature, shell temperature, ambient temperature and fluid medium temperature; preprocess and time synchronize the collected multi-source temperature data to construct the test temperature vector under each preset operating condition.

[0094] In one embodiment of the present invention, when the acquisition module 110 acquires the temperature data of the flywheel at a preset position during operation, it is used to: collect the initial temperature data of the flywheel at a preset position with a preset sampling period; filter the initial temperature data to obtain an effective temperature measurement result; and use the effective temperature measurement result as the temperature data of the rotor at the preset position during operation.

[0095] According to the flywheel rotor temperature estimation system 100 of the present invention, the acquisition module 110 acquires the test temperature vector of the flywheel under multiple preset operating conditions and acquires the temperature data of the flywheel rotor at preset positions during operation. Subsequently, the calibration module 120 performs offline parameter calibration on the finite element thermal simulation model based on the test temperature vector to obtain the calibrated finite element thermal simulation model. This not only corrects the inherent deviation of the finite element thermal simulation model itself, but also overcomes the defect of traditional models that cannot adapt to actual operating conditions due to the use of fixed parameter settings. Thus, it avoids the continuous transmission of the original simulation error in the process of reducing the order of the finite element thermal simulation model from the source, and improves the reproduction accuracy of the finite element thermal simulation model of the complex thermal behavior of the flywheel. Based on this, the estimation module 130 constructs a lightweight reduced-order model based on the calibrated finite element thermal simulation model. It inputs preset location temperature data as parameters into the lightweight reduced-order model and outputs the estimated current temperature of the rotor at the target location. This effectively overcomes the physical limitations of traditional detection methods, accurately capturing local hotspots formed on the rotor surface due to uneven eddy current losses, improving the accuracy and real-time performance of temperature detection. Furthermore, the computation time after training the lightweight reduced-order model is significantly reduced, fundamentally overcoming the limitation of easy accuracy decay in traditional open-loop prediction modes. This achieves high-precision, low-latency real-time estimation of the flywheel's overall temperature. This enables the flywheel to shift from reactive, post-event response to proactive early warning, effectively preventing serious failures such as thermal runaway and improving the safety and reliability of flywheel operation.

[0096] In the description of this specification, references to terms such as "one embodiment," "some embodiments," "illustrative embodiment," "example," "specific example," or "some examples," etc., refer to specific features, structures, materials, or characteristics described in connection with that embodiment or example, which are included in at least one embodiment or example of the present invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example.

[0097] Although embodiments of the invention have been shown and described, those skilled in the art will understand that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the claims and their equivalents.

Claims

1. A method for estimating the rotor temperature of a flywheel, characterized in that, include: The test temperature vector of the flywheel under multiple preset operating conditions is obtained, and the temperature data of the flywheel rotor at preset positions during operation is obtained. Based on the test temperature vector, the finite element thermal simulation model is calibrated offline to obtain the calibrated finite element thermal simulation model. A lightweight reduced-order model is constructed based on the calibrated finite element thermal simulation model, and the preset position temperature data is input into the lightweight reduced-order model as input parameters to output the current estimated temperature of the rotor at the target position. The process of performing offline parameter calibration on the finite element thermal simulation model based on the test temperature vector to obtain the calibrated finite element thermal simulation model includes: Obtain the initial simulation results of the finite element thermal simulation model, and compare and verify the initial simulation results with the test temperature vector point by point; The point-by-point comparison verification is evaluated by partitioning according to the preset component area, and the partitioning error in each preset component area is determined according to the preset algorithm. Perform multi-condition calibration of the heat source compensation coefficient and / or iterative adjustment of the contact thermal resistance used to characterize the heat transfer path; After completing the multi-condition calibration of the heat source compensation coefficient and / or the iterative adjustment of the contact thermal resistance used to characterize the heat transfer path, it is determined whether the partitioning error meets the preset convergence condition. If so, the convective heat transfer coefficient of the flywheel shell is adjusted to obtain the calibrated finite element thermal simulation model; The preset convergence condition includes: the partitioning error is less than a preset convergence threshold; When constructing a lightweight reduced-order model based on the calibrated finite element thermal simulation model, and inputting the preset position temperature data as input parameters into the lightweight reduced-order model, the output of the current estimated temperature of the rotor at the target position includes: Based on multiple sets of preset operating conditions, preset transient thermal field simulation is performed on the calibrated finite element thermal simulation model, and time-domain temperature response data of each target monitoring point in the calibrated finite element thermal simulation model are collected to construct a multi-condition temperature snapshot sample set. Based on the multi-condition temperature snapshot sample set, with multiple sets of the preset operating conditions as input variables and the current estimated temperature of the rotor at the target position as the output variable, a lightweight reduced-order model is constructed to characterize the mapping relationship between the preset operating conditions and the time-domain temperature response data. When the model deviation of the lightweight reduced-order model relative to the calibrated finite element thermal simulation model is less than or equal to the preset reduction error threshold, the lightweight reduced-order model is determined to be constructed successfully. The preset position temperature data is input into the constructed lightweight reduced-order model to output the current estimated temperature of the rotor at the target position.

2. The method for estimating the rotor temperature of a flywheel according to claim 1, characterized in that, When obtaining the initial simulation results of the finite element thermal simulation model, the following are included: Assign corresponding electromagnetic and thermal property parameters to each functional component in the finite element thermal simulation model; The losses of each functional component are determined based on the electromagnetic parameters. The losses of each of the aforementioned functional components are used as the heat source input for the finite element thermal simulation model; Based on the thermal property parameters, the heat source input, and the preset boundary conditions of the finite element thermal simulation model, the initial simulation results of the finite element thermal simulation model are obtained by solving according to the preset thermal field algorithm.

3. The method for estimating the rotor temperature of a flywheel according to claim 1, characterized in that, When performing multi-condition calibration of the heat source compensation coefficient, the following are included: The rotor speed of the flywheel, the charging and discharging power of the flywheel, and the rotor operating temperature are obtained. Based on the partition error, the rotor speed, the charging and discharging power, and the rotor operating temperature, the heat source compensation coefficient is determined by a preset fitting algorithm and according to a preset differential operating condition calibration strategy. The heat source compensation coefficient includes a reference coefficient, a speed influence factor, and a temperature influence factor.

4. The method for estimating the rotor temperature of a flywheel according to claim 1, characterized in that, When performing iterative adjustments to the contact thermal resistance used to characterize the heat transfer path, the following is included: According to the preset heat transfer path sequence, the contact thermal resistance of each target contact interface is iteratively calibrated in turn. During the iterative calibration process for the current target contact interface, the simulated temperature of the current target contact interface output by the finite element thermal simulation model and the actual temperature of the current target contact interface are obtained. Determine the temperature deviation between the actual temperature and the simulated temperature; The temperature deviation is used as the iterative objective function. The contact pressure of the current target contact interface is solved and corrected by a preset algorithm in reverse, so as to update the contact thermal resistance of the current target contact interface by the change of contact pressure. When the temperature deviation is less than the preset temperature difference threshold, it is determined that the contact thermal resistance calibration of the current target contact interface is complete, and the process jumps to the next target contact interface to continue the iterative calibration.

5. The method for estimating the rotor temperature of a flywheel according to claim 1, characterized in that, When adjusting the convective heat transfer coefficient of the flywheel's outer shell to obtain the calibrated finite element thermal simulation model, the following steps are included: Obtain the simulated shell temperature output by the finite element thermal simulation model, and obtain the actual shell temperature; Based on the simulated temperature of the outer shell and the actual temperature of the outer shell, determine whether the temperature difference between the simulated temperature of the outer shell and the actual temperature of the outer shell changes in the same direction; If so, the initial shell convection heat transfer coefficient is determined based on the preset heat transfer formula, and the initial shell convection heat transfer coefficient is iteratively adjusted according to the preset adjustment range; When iteratively adjusting the initial shell convection heat transfer coefficient according to the preset adjustment range, the maximum absolute error between the simulated shell temperature and the actual shell temperature is obtained. When the maximum absolute error is less than the preset fine-tuning convergence threshold, the iterative adjustment of the initial shell convection heat transfer coefficient is stopped, and the calibrated finite element thermal simulation model is obtained.

6. The method for estimating the rotor temperature of a flywheel according to claim 1, characterized in that, When obtaining the test temperature vector of the flywheel under multiple preset operating conditions, the following steps are included: Under multiple preset operating conditions, multi-source temperature data are continuously collected, including stator temperature, bearing temperature, rotor temperature, housing temperature, ambient temperature, and fluid medium temperature. The collected multi-source temperature data are preprocessed and time-synchronized to construct the test temperature vector under each preset operating condition.

7. The method for estimating the rotor temperature of a flywheel according to claim 1, characterized in that, When acquiring the temperature data of the flywheel at a preset position during operation, the following steps are included: The initial temperature data of the flywheel at a preset position is collected at a preset sampling period; The initial temperature data is filtered to obtain effective temperature measurement results; The effective temperature measurement results are used as the preset position temperature data of the rotor during operation.

8. A flywheel rotor temperature estimation system, characterized in that, include: The acquisition module is used to acquire the test temperature vector of the flywheel under multiple preset operating conditions, and to acquire the temperature data of the flywheel rotor at preset positions during operation. The calibration module is used to perform offline parameter calibration on the finite element thermal simulation model according to the test temperature vector, so as to obtain the calibrated finite element thermal simulation model. The estimation module is used to construct a lightweight reduced-order model based on the calibrated finite element thermal simulation model, and input the preset position temperature data as input parameters into the lightweight reduced-order model, and output the current estimated temperature of the rotor at the target position; Specifically, when performing offline parameter calibration on the finite element thermal simulation model based on the test temperature vector to obtain a calibrated finite element thermal simulation model, the calibration module is used for: Obtain the initial simulation results of the finite element thermal simulation model, and compare and verify the initial simulation results with the test temperature vector point by point; The point-by-point comparison verification is evaluated by partitioning according to the preset component area, and the partitioning error in each preset component area is determined according to the preset algorithm. Perform multi-condition calibration of the heat source compensation coefficient and / or iterative adjustment of the contact thermal resistance used to characterize the heat transfer path; After completing the multi-condition calibration of the heat source compensation coefficient and / or the iterative adjustment of the contact thermal resistance used to characterize the heat transfer path, it is determined whether the partitioning error meets the preset convergence condition. If so, the convective heat transfer coefficient of the flywheel shell is adjusted to obtain the calibrated finite element thermal simulation model; The preset convergence condition includes: the partitioning error is less than a preset convergence threshold; When constructing a lightweight reduced-order model based on the calibrated finite element thermal simulation model, and inputting the preset position temperature data as input parameters into the lightweight reduced-order model, and outputting the current estimated temperature of the rotor at the target position, the estimation module is used for: Based on multiple sets of preset operating conditions, preset transient thermal field simulation is performed on the calibrated finite element thermal simulation model, and time-domain temperature response data of each target monitoring point in the calibrated finite element thermal simulation model are collected to construct a multi-condition temperature snapshot sample set. Based on the multi-condition temperature snapshot sample set, with multiple sets of the preset operating conditions as input variables and the current estimated temperature of the rotor at the target position as the output variable, a lightweight reduced-order model is constructed to characterize the mapping relationship between the preset operating conditions and the time-domain temperature response data. When the model deviation of the lightweight reduced-order model relative to the calibrated finite element thermal simulation model is less than or equal to the preset reduction error threshold, the lightweight reduced-order model is determined to be constructed successfully. The preset position temperature data is input into the constructed lightweight reduced-order model to output the current estimated temperature of the rotor at the target position.

Images