Multi-mode adaptive switching control optimization method and system for thermal management of capacitor module

Abstract

The invention provides a multi-mode adaptive switching control optimization method and system for capacitor module thermal management, and relates to the technical field of capacitor module thermal management, and the method comprises the steps: building a state evolution space and a thermal state track through obtaining a temperature field and current data, analyzing and recognizing an energy collection and diffusion region through a curvature tensor and a divergence field, and obtaining a state evolution space and a thermal state track; and establishing a cooling mode equipotential plane boundary, generating a mode switching signal based on a space geometrical relationship, calculating an optimal cooling path, and performing inversion to obtain a cooling regulation sequence. According to the invention, intelligent switching of the cooling modes of the capacitor module is realized, the cooling efficiency is improved, and the service life of the capacitor module is prolonged.

Description

Technical Field

[0001] This invention relates to the field of capacitor module thermal management technology, and in particular to a multi-mode adaptive switching control optimization method and system for capacitor module thermal management. Background Technology

[0002] In the field of thermal management of capacitor modules, existing technologies typically rely on preset fixed cooling strategies or simple switching logic based on a single temperature threshold. The conventional approach is to configure one or more cooling modes for the capacitor module, such as forced air cooling, liquid cooling, or phase change cooling. The system monitors the temperature at key points of the module, and when the temperature exceeds a certain preset threshold, it switches from the current cooling mode to another preset cooling mode, or linearly adjusts the intensity of a single cooling mode. The core of this control logic lies in establishing a direct, static mapping relationship between temperature and cooling action; its decision-making basis is relatively singular, primarily focusing on whether the instantaneous temperature exceeds the limit.

[0003] However, these conventional thermal management methods have significant limitations. Due to the uneven and rapidly changing temperature field distribution within the capacitor module, relying solely on a few measuring points or a single threshold cannot accurately perceive the overall thermal state evolution trend and local heat flow characteristics of the module. This leads to a lag in system response, often resulting in action only after significant thermal problems have accumulated, making it difficult to prevent localized overheating. Furthermore, fixed threshold switching strategies lack adaptability to dynamic changes in module operating conditions. For example, the heat generation rate varies greatly under different current loads; the same cooling strategy may be insufficient in some conditions and excessive in others, causing energy waste and potentially introducing unnecessary thermal stress cycles, affecting the lifespan and reliability of the capacitor module. Summary of the Invention

[0004] This invention provides a multi-mode adaptive switching control optimization method and system for capacitor module thermal management, which can solve the problems in the prior art.

[0005] A first aspect of this invention provides a multi-mode adaptive switching control optimization method for capacitor module thermal management, comprising:

[0006] Acquire temperature field data, current data, and operating parameters for each cooling mode of the capacitor module;

[0007] Based on temperature field data and current data, temperature state variables and heat flux state variables are mapped to construct a state evolution space and plot the thermal state trajectory; by calculating the curvature tensor and divergence field distribution of the thermal state trajectory, energy accumulation regions and energy diffusion regions are identified.

[0008] Based on the energy accumulation area and the energy diffusion area, the equipotential surface boundary corresponding to each cooling mode is extracted, and the spatial geometric relationship between the trajectory point corresponding to the current thermal state and the equipotential surface boundary is calculated. When the spatial geometric relationship meets the preset switching conditions, a mode switching trigger signal is generated.

[0009] The path optimization is initiated by using a mode switching trigger signal. The path integral potential energy and cumulative risk value of the state space path corresponding to each candidate cooling mode are calculated to determine the comprehensive index. The candidate cooling mode with the smallest comprehensive index is selected as the target cooling mode.

[0010] The cooling intensity control sequence is obtained by inverting the state space path corresponding to the target cooling mode, and the cooling execution unit is controlled to adjust the cooling intensity according to the cooling intensity control sequence.

[0011] In one optional embodiment, based on temperature field data and current data, temperature state variables and heat flux state variables are mapped, and a state evolution space is constructed and a thermal state trajectory is plotted, including:

[0012] The temperature field data is divided into multiple sub-regions according to spatial location. The average temperature and temperature variance of each sub-region are calculated. The average temperature is used as the first temperature state component and the temperature variance is used as the second temperature state component. They are combined to form the temperature state variable.

[0013] The current change rate is obtained by time difference analysis of the current data. The thermal power of each sub-region is calculated based on the current change rate and the heat capacity of each sub-region. The total thermal power is obtained by spatial summation. The thermal power distribution non-uniformity is obtained by spatial standard deviation of the thermal power of each sub-region. The total thermal power is used as the first thermal flux state component and the thermal power distribution non-uniformity is used as the second thermal flux state component. They are combined to form the thermal flux state variable.

[0014] A temperature subspace is established based on the first temperature state component and the second temperature state component, and a heat flow subspace is established based on the first heat flow state component and the second heat flow state component. The temperature subspace and the heat flow subspace are combined to form a four-dimensional state evolution space.

[0015] The temperature and heat flux state variables at each moment are mapped to state points in the state evolution space in chronological order, and the state points at adjacent moments are connected in sequence to form a thermal state trajectory.

[0016] In one optional embodiment, identifying energy accumulation and energy diffusion regions by calculating the curvature tensor and divergence field distribution of the thermal state trajectory includes:

[0017] The thermal trajectory is tangent-space projected, and the trajectory motion is decomposed into a thermal potential component along the temperature state direction and an energy flow component along the heat flow state direction. The rate of change of the angle between the thermal potential component and the energy flow component is calculated as the trajectory torsion. The trajectory torsion is combined with the trajectory curvature to construct a curvature tensor.

[0018] The divergence field is obtained by performing divergence calculation on the heat flow state variables in the state evolution space. The divergence field is then decomposed into multi-scale wavelet components to extract the first-scale divergence components and the second-scale divergence components. The corresponding spatial correlation coefficients are calculated, and the divergence components are weighted and fused based on the spatial correlation coefficients to obtain the reconstructed divergence field.

[0019] In the curvature tensor, identify torsional abrupt change points where the torsional degree exceeds a preset torsional threshold. In the reconstructed divergence field, identify divergence reversal points where the divergence value changes from negative to positive or from positive to negative. Calculate the spatiotemporal matching degree between the torsional abrupt change points and the divergence reversal points. Determine the region where the spatiotemporal matching degree exceeds a preset matching threshold as the energy phase transition region.

[0020] Cross-correlation analysis is performed on the temperature state variables and heat flux state variables within the energy phase transition region. The phase lag of the temperature state variables relative to the heat flux state variables is calculated, and the phase lag is compared with a preset lag threshold to determine the energy accumulation region or the energy diffusion region.

[0021] In one optional embodiment, based on the energy accumulation region and the energy diffusion region, the equipotential surface boundary corresponding to each cooling mode is extracted, and the spatial geometric relationship between the trajectory point corresponding to the current thermal state and the equipotential surface boundary is calculated. When the spatial geometric relationship meets the preset switching conditions, a mode switching trigger signal is generated, including:

[0022] In the state evolution space, the thermal state distribution of each cooling mode is sliced ​​into isosurfaces, and the isosurfaces at the boundary between the energy accumulation region and the energy diffusion region are extracted to obtain the transformation equipotential surface of the cooling mode at the accumulation-diffusion transition point. The envelope surface is obtained by fitting multiple transformation equipotential surfaces of each cooling mode under different temperature isosurfaces. The intersection of the envelope surface and the boundary of the state evolution space is taken as the equipotential surface boundary of the cooling mode.

[0023] Calculate the shortest distance from the trajectory point corresponding to the current thermal state to the boundary of the equipotential surface corresponding to each cooling mode, calculate the angle between the direction of movement of the trajectory point and the boundary of the equipotential surface, and combine the shortest distance and the angle to construct a spatial geometric relationship quantity.

[0024] The spatial geometric relationship quantity is compared with the preset relationship threshold. When the spatial geometric relationship quantity is less than the relationship threshold, it is determined that the current thermal state trajectory point has crossed the equipotential surface boundary. The cooling mode corresponding to the equipotential surface boundary with the smallest spatial geometric relationship quantity is identified as the target cooling mode, and a mode switching trigger signal is generated.

[0025] In one optional embodiment, path optimization is initiated using a mode switching trigger signal. The path integral potential energy and cumulative risk value of the state space path corresponding to each candidate cooling mode are calculated to determine a comprehensive index. The candidate cooling mode with the smallest comprehensive index is then selected as the target cooling mode.

[0026] Receive the mode switching trigger signal, extract the position coordinates and motion velocity vector of the trajectory point corresponding to the current thermal state, construct the path search domain with the current trajectory point as the starting point in the state evolution space, extend the normal of the equipotential surface boundary corresponding to each candidate cooling mode to obtain the equipotential surface gradient field, and perform path tracking along the direction of the equipotential surface gradient field until the steady state point of each candidate cooling mode is reached, thus obtaining the state space path corresponding to each candidate cooling mode.

[0027] Energy density is calculated for sampling points on the state space path. The line integral of the energy density along the path is used as the path integral potential energy. The shortest distance from the sampling point on the state space path to the boundary of the energy accumulation region is calculated. The line integral of the product of the shortest distance and the heat flux gradient magnitude of the sampling point along the path is used as the cumulative risk value.

[0028] The comprehensive index is obtained by multiplying the path integral potential energy by a preset potential energy weighting coefficient and the cumulative risk value by a preset risk weighting coefficient. The comprehensive index of each candidate cooling mode is compared and the candidate cooling mode with the smallest comprehensive index is selected as the target cooling mode.

[0029] In one optional embodiment, the cooling intensity control sequence is obtained by inverting the state space path corresponding to the target cooling mode, and the cooling execution unit is controlled to adjust the cooling intensity according to the cooling intensity control sequence, including:

[0030] Extract the energy density gradient and heat flux density vector of each sampling point on the state space path corresponding to the target cooling mode, construct the energy backtracking path from each sampling point to the heat source of the capacitor module, mark the energy transfer nodes on the energy backtracking path, and solve the source cooling demand corresponding to each sampling point in reverse by using the thermal resistance network at the energy transfer node. Convert the source cooling demand into the cooling intensity value that the cooling execution unit should output at the corresponding time of the sampling point.

[0031] The cooling intensity values ​​corresponding to each sampling point are matched and verified by pattern feature matching. The cooling intensity values ​​are substituted into the state evolution law of the target cooling mode for forward verification. The cooling intensity deviation points that cause the state evolution to deviate from the state space path are identified. The cooling intensity values ​​of the cooling intensity deviation points are iteratively corrected until the state evolution trajectory converges to the state space path. The corrected cooling intensity values ​​are combined according to the time series to form a cooling intensity control sequence.

[0032] The cooling intensity control sequence is sent to the cooling execution unit, which then controls the cooling execution unit to adjust the cooling intensity sequentially according to the cooling intensity value corresponding to each moment in the cooling intensity control sequence.

[0033] A second aspect of this invention provides a multi-mode adaptive switching control optimization system for capacitor module thermal management, comprising:

[0034] The data acquisition unit is used to acquire temperature field data, current data, and operating parameters of each cooling mode of the capacitor module.

[0035] The state construction unit is used to map temperature state variables and heat flux state variables based on temperature field data and current data, construct the state evolution space and draw the thermal state trajectory.

[0036] The region identification unit is used to identify energy accumulation regions and energy diffusion regions by calculating the curvature tensor and divergence field distribution of the thermal state trajectory;

[0037] The boundary extraction unit is used to extract the equipotential surface boundary corresponding to each cooling mode based on the energy accumulation area and the energy diffusion area, calculate the spatial geometric relationship between the trajectory point corresponding to the current thermal state and the equipotential surface boundary, and generate a mode switching trigger signal when the spatial geometric relationship meets the preset switching conditions.

[0038] The mode optimization unit is used to initiate path optimization by using the mode switching trigger signal, calculate the path integral potential energy and cumulative risk value of the state space path corresponding to each candidate cooling mode, determine the comprehensive index, and select the candidate cooling mode with the smallest comprehensive index as the target cooling mode.

[0039] The control execution unit is used to obtain the cooling intensity control sequence by inverting the state space path corresponding to the target cooling mode, and to control the cooling execution unit to adjust the cooling intensity according to the cooling intensity control sequence.

[0040] A third aspect of the present invention provides an electronic device, comprising:

[0041] processor;

[0042] Memory used to store processor-executable instructions;

[0043] The processor is configured to invoke instructions stored in the memory to execute the aforementioned method.

[0044] A fourth aspect of the present invention provides a computer-readable storage medium having stored thereon computer program instructions that, when executed by a processor, implement the aforementioned method.

[0045] In this embodiment of the invention, by fusing temperature field and current data, temperature state variables and heat flow state variables reflecting the essence of thermal dynamics are directly mapped, constructing a multidimensional state evolution space and plotting thermal state trajectories. This allows for a holistic characterization of the spatiotemporal distribution and evolution trend of thermal energy. By calculating the curvature tensor and divergence field distribution of the thermal state trajectory, high-risk areas where energy easily accumulates and safe areas where energy naturally diffuses can be quantitatively distinguished, enabling the early detection of local overheating hazards that are difficult to detect using traditional methods. By extracting the equipotential surface boundaries corresponding to each cooling mode and calculating the spatial geometric relationship between the current thermal state point and these boundaries, automatic generation can be performed based on preset switching conditions. The system generates a trigger signal, making the timing of mode switching more closely match the dynamic needs of the actual thermal state. By calculating the path integral potential energy and cumulative risk value of the state space path corresponding to each candidate mode, and comprehensively evaluating and determining the optimal index, the target cooling mode with the lowest overall cost is selected. This not only considers transient effects but also takes into account the overall energy consumption and system stability of the switching process. Based on the selected target cooling mode and its optimal path, the specific cooling intensity control sequence is derived and the execution unit is driven to adjust sequentially. This system can guide and maintain the thermal state of the capacitor module within a safe and efficient operating range with minimal control cost, significantly improving the energy efficiency and reliability of the thermal management system. Attached Figure Description

[0046] Figure 1 This is a flowchart illustrating the multi-mode adaptive switching control optimization method for capacitor module thermal management according to an embodiment of the present invention.

[0047] Figure 2 Flowchart for identifying energy accumulation and diffusion regions. Detailed Implementation

[0048] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0049] The technical solution of the present invention will be described in detail below with reference to specific embodiments. These specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described again in some embodiments.

[0050] Figure 1 This is a flowchart illustrating the multi-mode adaptive switching control optimization method for capacitor module thermal management according to an embodiment of the present invention. Figure 1 As shown, the method includes:

[0051] Acquire temperature field data, current data, and operating parameters for each cooling mode of the capacitor module;

[0052] Based on temperature field data and current data, temperature state variables and heat flux state variables are mapped to construct a state evolution space and plot the thermal state trajectory; by calculating the curvature tensor and divergence field distribution of the thermal state trajectory, energy accumulation regions and energy diffusion regions are identified.

[0053] Based on the energy accumulation area and the energy diffusion area, the equipotential surface boundary corresponding to each cooling mode is extracted, and the spatial geometric relationship between the trajectory point corresponding to the current thermal state and the equipotential surface boundary is calculated. When the spatial geometric relationship meets the preset switching conditions, a mode switching trigger signal is generated.

[0054] The path optimization is initiated by using a mode switching trigger signal. The path integral potential energy and cumulative risk value of the state space path corresponding to each candidate cooling mode are calculated to determine the comprehensive index. The candidate cooling mode with the smallest comprehensive index is selected as the target cooling mode.

[0055] The cooling intensity control sequence is obtained by inverting the state space path corresponding to the target cooling mode, and the cooling execution unit is controlled to adjust the cooling intensity according to the cooling intensity control sequence.

[0056] In one optional implementation, based on temperature field data and current data, temperature state variables and heat flux state variables are mapped to construct a state evolution space and plot the thermal state trajectory, including:

[0057] The temperature field data is divided into multiple sub-regions according to spatial location. The average temperature and temperature variance of each sub-region are calculated. The average temperature is used as the first temperature state component and the temperature variance is used as the second temperature state component. They are combined to form the temperature state variable.

[0058] The current change rate is obtained by time difference analysis of the current data. The thermal power of each sub-region is calculated based on the current change rate and the heat capacity of each sub-region. The total thermal power is obtained by spatial summation. The thermal power distribution non-uniformity is obtained by spatial standard deviation of the thermal power of each sub-region. The total thermal power is used as the first thermal flux state component and the thermal power distribution non-uniformity is used as the second thermal flux state component. They are combined to form the thermal flux state variable.

[0059] A temperature subspace is established based on the first temperature state component and the second temperature state component, and a heat flow subspace is established based on the first heat flow state component and the second heat flow state component. The temperature subspace and the heat flow subspace are combined to form a four-dimensional state evolution space.

[0060] The temperature and heat flux state variables at each moment are mapped to state points in the state evolution space in chronological order, and the state points at adjacent moments are connected in sequence to form a thermal state trajectory.

[0061] In one specific implementation, detailed temperature field data is acquired by strategically arranging multiple temperature sensors on and inside the capacitor module to form a temperature sensing network. The temperature sensors are thermistor or platinum resistance thermometer type, with a measurement accuracy controlled within 0.1℃ and a sampling frequency set to 1Hz to 10Hz to ensure real-time monitoring of temperature changes in the capacitor module. To accurately characterize the thermal state of the capacitor module, the entire spatial location of the capacitor module is divided into multiple sub-regions. The number of sub-regions is determined based on the size and structural complexity of the capacitor module, typically ranging from 8 to 16. Region division methods include uniform grid division, structure-based functional zoning, and hotspot focusing based on heat flow distribution. Each sub-region has at least three temperature measurement points, forming a dense temperature monitoring network.

[0062] For each sub-region, temperature values ​​from all temperature measuring points within that region are collected in real time. The average temperature is calculated as the first temperature state component, representing the overall temperature level of the region. Taking sub-region A as an example, when the values ​​of the five temperature measuring points within the region are 42.5℃, 43.1℃, 41.9℃, 42.8℃, and 42.3℃, the first temperature state component for that region is calculated to be 42.52℃. Simultaneously, the variance of the temperature measuring point values ​​is calculated as the second temperature state component, reflecting the degree of unevenness in temperature distribution within the region. The temperature state components of all sub-regions are processed, and either a weighted average method or a maximum value method is used to obtain the overall temperature state variable. In the weighted average method, the weighting coefficients are determined based on the importance or volume ratio of the sub-regions, and the sum is 1. The maximum value method directly selects the maximum average temperature and the maximum temperature variance across all sub-regions, suitable for scenarios where the hottest areas of the capacitor module are of interest.

[0063] Current data acquisition is achieved using a high-precision current sensor with a sampling period of 0.5 seconds. The acquired current data is filtered to remove noise. The rate of change of current is calculated by performing a difference operation on the current values ​​at two adjacent moments. Specifically, the current value at the current moment is subtracted from the current value at the previous moment, and then divided by the sampling period to obtain the rate of change of current. For example, if the current values ​​at two adjacent sampling points are 125A and 128A respectively, and the sampling period is 0.5 seconds, then the rate of change of current is 6A / second.

[0064] Based on the thermal characteristics of each sub-region of the capacitor module, the thermal power of each sub-region is calculated. The thermal power calculation considers two parts: current heat loss and energy storage caused by temperature changes. Current heat loss equals the square of the current multiplied by the equivalent resistance, and energy storage caused by temperature changes equals the region's heat capacity multiplied by the rate of temperature change. The equivalent resistance is determined through experimental measurement, and the heat capacity is calculated using the mass and specific heat capacity of the region's materials. For example, sub-region B has an equivalent resistance of 0.05 ohms and a heat capacity of 450 joules / degree Celsius. When the current is 120 amperes and the temperature of this region rises at a rate of 0.2°C / second, the calculated thermal power is 810 watts.

[0065] The total thermal power of the capacitor module is obtained by summing the thermal power of all sub-regions, and is used as the first heat flow state component. When the capacitor module contains 10 sub-regions with thermal powers of 780 W, 810 W, 750 W, 820 W, 790 W, 805 W, 775 W, 795 W, 815 W, and 760 W respectively, the total thermal power is 7900 W. The standard deviation of the thermal power of each sub-region relative to the average thermal power is calculated to reflect the spatial distribution characteristics of heat generation, and is used as the second heat flow state component. The standard deviation of the above thermal power data is 23.3 W, indicating that the thermal power distribution is relatively uniform.

[0066] When constructing the state evolution space, a four-dimensional coordinate system is used to represent the thermal state of the capacitor module. The first and second dimensions correspond to the first and second temperature state components, respectively, forming a temperature subspace; the third and fourth dimensions correspond to the first and second heat flow state components, respectively, forming a heat flow subspace. The combination of these four dimensions constitutes a complete state evolution space, where any point uniquely corresponds to the thermal state of the capacitor module at a given moment.

[0067] Thermal state trajectory plotting is achieved through time-series mapping. Starting from the initial moment, a set of temperature and heat flux state variables are extracted at fixed time intervals and mapped onto the state evolution space to form state points. The sampling time interval is determined based on the rate of change of the capacitor module's thermal state, typically set to 5 to 30 seconds. Adjacent state points are connected in chronological order to form the thermal state trajectory. The connection method uses either straight line segments or cubic spline curves, with the latter providing a smoother trajectory representation.

[0068] In thermal trajectory analysis, trajectory features are extracted, including trajectory shape, rate of state change, and acceleration. The trajectory shape reflects the characteristics of the capacitor module's operating mode, the rate of state change reflects the system's response rate, and acceleration change points indicate changes in the system's dynamic characteristics. The trajectory analysis results are used for the automatic identification and switching control of the capacitor module's thermal management mode. For example, during the charging process of a capacitor module, the temperature state component gradually increases from an initial value of 25.2℃ and 0.15 to 39.8℃ and 0.35, while the heat flux state component increases from an initial value of 2500W and 15W to 6800W and 28W, eventually forming a closed trajectory, indicating that the system has reached thermal equilibrium.

[0069] Abnormal state detection is based on thermal state trajectories. Normal operating range boundaries are defined, and an early warning mechanism is triggered when the trajectory exceeds these boundaries or exhibits abnormal change patterns. Boundary values ​​are determined through historical data analysis and safety standards; for example, the upper limit for temperature is set at 55°C, the upper limit for temperature variance is set at 0.8, the upper limit for total thermal power is set at 12,000 watts, and the upper limit for thermal power distribution non-uniformity is set at 50 watts. The thermal state trajectory analysis method comprehensively monitors the temperature distribution and heat flow status of the capacitor module, achieving accurate characterization of the thermal management status and providing a scientific basis for multi-mode adaptive switching control.

[0070] When establishing the operating condition database, the thermal trajectory of the capacitor module under different operating conditions is collected, including normal charging, fast charging, normal discharging, fast discharging, and static states. At least 50 sets of thermal trajectory data are collected for each operating condition, and trajectory features are extracted to form an operating condition feature database. The current trajectory is matched with standard trajectories in the feature database, and a dynamic time warping algorithm is used to calculate the similarity to identify the current operating condition type. Based on the identification results, a suitable thermal management mode is automatically selected for the current operating condition, optimizing the cooling strategy and improving the safety and lifespan of the capacitor module.

[0071] In one optional implementation, identifying energy accumulation and energy diffusion regions by calculating the curvature tensor and divergence field distribution of the thermal state trajectory includes:

[0072] The thermal trajectory is tangent-space projected, and the trajectory motion is decomposed into a thermal potential component along the temperature state direction and an energy flow component along the heat flow state direction. The rate of change of the angle between the thermal potential component and the energy flow component is calculated as the trajectory torsion. The trajectory torsion is combined with the trajectory curvature to construct a curvature tensor.

[0073] The divergence field is obtained by performing divergence calculation on the heat flow state variables in the state evolution space. The divergence field is then decomposed into multi-scale wavelet components to extract the first-scale divergence components and the second-scale divergence components. The corresponding spatial correlation coefficients are calculated, and the divergence components are weighted and fused based on the spatial correlation coefficients to obtain the reconstructed divergence field.

[0074] In the curvature tensor, identify torsional abrupt change points where the torsional degree exceeds a preset torsional threshold. In the reconstructed divergence field, identify divergence reversal points where the divergence value changes from negative to positive or from positive to negative. Calculate the spatiotemporal matching degree between the torsional abrupt change points and the divergence reversal points. Determine the region where the spatiotemporal matching degree exceeds a preset matching threshold as the energy phase transition region.

[0075] Cross-correlation analysis is performed on the temperature state variables and heat flux state variables within the energy phase transition region. The phase lag of the temperature state variables relative to the heat flux state variables is calculated, and the phase lag is compared with a preset lag threshold to determine the energy accumulation region or the energy diffusion region.

[0076] In one specific implementation, the thermal state trajectory is spatially projected, decomposing the trajectory into a thermal potential component along the temperature state direction and an energy flow component along the heat flow state direction. Specifically, the four-dimensional state point is projected onto a temperature subspace and a heat flow subspace, resulting in two two-dimensional projection points. The displacement vector between adjacent projection points represents the thermal potential component and the energy flow component. For example, a state point at a certain moment (38.5℃, 0.25, 5200 W, 18 W) is projected to obtain a temperature subspace point (38.5℃, 0.25) and a heat flow subspace point (5200 W, 18 W).

[0077] The rate of change of the angle between the thermal potential component and the energy flux component is calculated and defined as the trajectory torsion. The angle is calculated using the geometric relationship between displacement vectors. For example, if the angles at three consecutive moments are 25 degrees, 32 degrees, and 45 degrees, with a time interval of 5 seconds, then the torsion at the second moment is 2 degrees / second. Trajectory curvature describes the degree of trajectory bending, and its value is determined using three adjacent state points. For example, if the radius of the fitted curvature circle is 120 units, the curvature value is 0.00833. The trajectory torsion and curvature are combined to construct a curvature tensor, which describes the local geometric properties of the trajectory.

[0078] The divergence field is calculated for the heat flux state variables to describe the diffusion or accumulation trend of heat flux. The central difference method is used to calculate the rate of change of heat flux centered on a grid point. For example, if the heat flux state variables around a certain grid point are 5200 W, 5250 W, 5180 W, 5220 W, 5190 W, and 5260 W, the calculated divergence value is 0.00042, indicating a diffusion trend in heat.

[0079] The divergence field is decomposed using multi-scale wavelet decomposition to extract features at different frequencies. Using Haar or Daubechies wavelets, it is decomposed into low-frequency approximate components and high-frequency detail components, extracting first-scale and second-scale divergence components. The first-scale divergence component corresponds to the high-frequency detail component, reflecting the local rapid changes in the divergence field and capturing transient thermal disturbances in the system. The second-scale divergence component corresponds to the low-frequency approximate component, reflecting the overall trend of the divergence field and capturing the main heat transfer processes in the system. For example, the range of the first-scale component of the divergence field in a certain region is [-0.00095, 0.00085], and the range of the second-scale component is [-0.00055, 0.00048]. The spatial correlation coefficient between the two-scale divergence components is calculated to assess the correlation. The spatial correlation coefficient is calculated using the Pearson correlation coefficient method, quantifying the similarity in spatial distribution between the first-scale and second-scale divergence components. The spatial correlation coefficient has a range of [-1, 1]. A value close to 1 indicates a high positive correlation between the spatial distributions of the two scale components, a value close to -1 indicates a high negative correlation, and a value close to 0 indicates almost no correlation. For example, a correlation coefficient of 0.78 indicates a high correlation. Weighted fusion based on the correlation coefficient yields the reconstructed divergence field.

[0080] Identify abrupt torsional abrupt changes in the curvature tensor where the torsional degree exceeds a preset threshold. The preset torsional threshold is typically 1.8 times the average torsional degree. For example, if the average torsional degree is 1.5 degrees / second, the threshold is set to 2.7 degrees / second. Identify divergence reversal points in the reconstructed divergence field where divergence values ​​transition from positive to negative. For example, a divergence value changing from -0.00025 to 0.00018 at adjacent time points is marked as a reversal point.

[0081] The spatiotemporal matching degree between the torsional mutation point and the divergence reversal point is calculated to assess the correlation. The temporal matching degree is the reciprocal of the time difference, and the spatial matching degree is the reciprocal of the state-space distance. For example, if the mutation point occurs at 35 seconds and the reversal point occurs at 38 seconds, with a spatial distance of 0.25 units, the spatiotemporal matching degree is 0.8. Regions with a matching degree exceeding a preset threshold of 0.7 are identified as energy phase transition regions.

[0082] Cross-correlation analysis was performed on the temperature and heat flux state variables within the energy phase transition region to calculate the phase lag. A sliding window method was used, with a window length of 15 time points. The cross-correlation function was calculated for each window, and the time delay corresponding to the maximum value was determined. For example, the temperature variable lags the heat flux variable by 3.5 seconds, with a phase angle of 63 degrees. The phase lag was compared with a preset threshold of 45 degrees to determine whether the energy accumulation or diffusion region was identified. A lag less than the threshold indicates an energy accumulation region, while a lag greater than the threshold indicates an energy diffusion region.

[0083] Based on the energy distribution, heat flow patterns are determined. When energy accumulation regions dominate, it indicates a heat accumulation phase requiring enhanced cooling; when energy diffusion regions dominate, it indicates a thermal equilibrium phase, allowing for reduced cooling. Precise switching of thermal management modes is achieved by monitoring the evolution of the energy phase transition region. In a practical application, a sudden change in torsion of 3.2 degrees / second was identified during the charging process of a capacitor module, along with a divergence reversal point, with a spatiotemporal matching degree of 0.85, confirming that the energy phase transition region is located in the central region. The phase lag of 38 degrees is less than the threshold of 45 degrees, indicating an energy accumulation region, triggering a forced cooling mode to prevent the risk of thermal runaway.

[0084] like Figure 2 The diagram shown illustrates the flowchart for identifying energy accumulation and diffusion regions.

[0085] In one optional implementation, based on the energy accumulation region and the energy diffusion region, the equipotential surface boundary corresponding to each cooling mode is extracted, and the spatial geometric relationship between the trajectory point corresponding to the current thermal state and the equipotential surface boundary is calculated. When the spatial geometric relationship meets the preset switching conditions, a mode switching trigger signal is generated, including:

[0086] In the state evolution space, the thermal state distribution of each cooling mode is sliced ​​into isosurfaces, and the isosurfaces at the boundary between the energy accumulation region and the energy diffusion region are extracted to obtain the transformation equipotential surface of the cooling mode at the accumulation-diffusion transition point. The envelope surface is obtained by fitting multiple transformation equipotential surfaces of each cooling mode under different temperature isosurfaces. The intersection of the envelope surface and the boundary of the state evolution space is taken as the equipotential surface boundary of the cooling mode.

[0087] Calculate the shortest distance from the trajectory point corresponding to the current thermal state to the boundary of the equipotential surface corresponding to each cooling mode, calculate the angle between the direction of movement of the trajectory point and the boundary of the equipotential surface, and combine the shortest distance and the angle to construct a spatial geometric relationship quantity.

[0088] The spatial geometric relationship quantity is compared with the preset relationship threshold. When the spatial geometric relationship quantity is less than the relationship threshold, it is determined that the current thermal state trajectory point has crossed the equipotential surface boundary. The cooling mode corresponding to the equipotential surface boundary with the smallest spatial geometric relationship quantity is identified as the target cooling mode, and a mode switching trigger signal is generated.

[0089] In one specific implementation, to achieve intelligent switching of cooling modes, an isosurface analysis method based on thermal state distribution is used for mode discrimination. This method performs isosurface slicing analysis on the thermal state evolution space, extracts the transformation equipotential surface at the boundary between energy accumulation and diffusion regions, constructs a spatial geometric relationship evaluation index, and achieves adaptive switching of cooling modes.

[0090] The capacitor module thermal management system includes multiple cooling modes such as natural cooling, forced air cooling, and liquid cooling. Each cooling mode results in different energy evolution characteristics in the thermal state space. The thermal state space is composed of parameters such as the capacitor module's temperature, rate of temperature change, and heat flux density. In the three-dimensional thermal state space, the thermal state evolution trajectory of the capacitor module under different cooling modes is recorded to obtain the thermal state distribution data for each cooling mode.

[0091] For the natural cooling mode, isosurface slices were created at 5°C intervals within the temperature range of 25°C to 85°C, resulting in 12 temperature isosurfaces. On each temperature isosurface, the heat flux density gradient was calculated, and points where the heat flux decreased from high to low were identified. These points constituted the boundaries between energy accumulation and diffusion regions. Connecting the boundaries of the 12 temperature isosurfaces formed a transformation equipotential surface, which characterizes the critical state where energy accumulation transitions to diffusion under the natural cooling mode.

[0092] For the forced air cooling mode, isosurface slices were taken at 5°C intervals within the temperature range of 25°C to 65°C, resulting in 9 temperature isosurfaces. Under forced air cooling conditions, the heat flux density gradient changes more significantly due to the accelerated heat transfer caused by airflow. The energy accumulation and diffusion boundaries on each temperature isosurface were extracted using the same method and connected to form the conversion equipotential surface of the forced air cooling mode.

[0093] For the liquid cooling mode, isosurface slices were created at 5°C intervals within the temperature range of 25°C to 45°C, resulting in five temperature isosurfaces. The liquid cooling mode exhibits high heat transfer efficiency, dramatic changes in heat flux density gradient, and clear boundaries between energy accumulation and diffusion regions. By extracting the boundary lines from each temperature isosurface and connecting them, a transformation equipotential surface for the liquid cooling mode was formed.

[0094] Envelope fitting was performed on the transformed equipotential surfaces obtained for each cooling mode. Specifically, cubic spline interpolation was used to fit a continuous surface to multiple transformed equipotential surfaces for each cooling mode, resulting in the envelope surface. The envelope surface of the natural cooling mode approximates a downward-opening parabola, covering temperatures from 35℃ to 85℃, a temperature change rate from -0.5℃ / min to 0.8℃ / min, and a heat flux density of 50W / m³ in the thermal state space. 2 Up to 200W / m 2 The forced air cooling mode exhibits a relatively steep curved surface, covering a temperature range of 30℃ to 65℃, a temperature change rate of -1.2℃ / min to 1.5℃ / min, and a heat flux density of 150W / m³. 2 Up to 400W / m 2 The liquid cooling mode exhibits an approximately planar envelope, covering a temperature range of 25°C to 45°C, a temperature change rate of -2.5°C / min to 2.0°C / min, and a heat flux density of 350 W / m³. 2Up to 600W / m 2 The area.

[0095] Intersecting each envelope surface with the thermal state evolution space boundary yields the equipotential surface boundaries corresponding to each cooling mode. The thermal state evolution space boundary is determined by the limiting parameters for the safe operation of the capacitor module, including a maximum temperature limit of 85℃, a maximum temperature change rate limit of ±3℃ / min, and a maximum heat flux density limit of 700W / m³. 2 The equipotential surface boundary of the natural cooling mode is elliptical in the temperature-temperature change rate plane, with a major axis of about 50°C and a minor axis of about 1.3°C / min; the equipotential surface boundary of the forced air cooling mode is an irregular closed curve, which extends more in the direction of heat flux density; the equipotential surface boundary of the liquid cooling mode is approximately circular, with a diameter of about 20°C.

[0096] Real-time monitoring of the capacitor module's current thermal state parameters, including temperature, rate of temperature change, and heat flux density, is used to construct thermal state trajectory points. The shortest distance from these trajectory points to the boundaries of the equipotential surfaces of each cooling mode is calculated. Specifically, in the thermal state space, perpendicular lines are drawn from the trajectory points to the boundaries of each equipotential surface; the length of these perpendicular lines represents the shortest distance. This calculation is performed when the capacitor module temperature is 60℃, the rate of temperature change is 0.5℃ / min, and the heat flux density is 150W / m³. 2 At that time, the shortest distance from the trajectory point to the boundary of the equipotential surface in the natural cooling mode was 15.3 units, the shortest distance to the boundary of the equipotential surface in the forced air cooling mode was 8.7 units, and the shortest distance to the boundary of the equipotential surface in the liquid cooling mode was 25.6 units.

[0097] Calculate the angle between the direction of motion of the thermal state trajectory point and the boundaries of each equipotential surface. The direction of motion of the trajectory point is determined by the difference in thermal state parameters between two adjacent sampling times. When the thermal state of the capacitor module changes from (58℃, 0.4℃ / min, 145W / m)... 2 The temperature changes to (60℃, 0.5℃ / min, 150W / m). 2 When the trajectory point moves, the angle between the direction of motion and the boundary of the equipotential surface in the natural cooling mode is 78 degrees, the angle between the direction of motion and the boundary of the equipotential surface in the forced air cooling mode is 35 degrees, and the angle between the direction of motion and the boundary of the equipotential surface in the liquid cooling mode is 115 degrees.

[0098] The shortest distance and the included angle are combined to construct a spatial geometric relation quantity. This quantity is defined as the absolute value of the shortest distance divided by the cosine of the included angle, representing the tendency of the trajectory point to approach the boundary of the equipotential surface. When the included angle is close to 90 degrees, the square of the distance is used as the spatial geometric relation quantity. In the example above, the spatial geometric relation quantity of the trajectory point for the natural cooling mode is 49.7, for the forced air cooling mode it is 10.6, and for the liquid cooling mode it is 66.3.

[0099] The calculated spatial geometric relationship quantity is compared with a preset relationship threshold. The relationship threshold is determined based on the response characteristics of the capacitor module's thermal management system and is typically set to 15. When the spatial geometric relationship quantity is less than the relationship threshold, it is determined that the thermal trajectory point has crossed or is about to cross the equipotential surface boundary. In the example above, the spatial geometric relationship quantity of 10.6 in the forced air cooling mode is less than the threshold of 15, and it is determined that the trajectory point is about to cross the equipotential surface boundary of the forced air cooling mode.

[0100] The cooling mode corresponding to the equipotential surface boundary with the smallest spatial geometric relationship is identified as the target cooling mode. In the example above, the forced air cooling mode has the smallest spatial geometric relationship, therefore forced air cooling is selected as the target cooling mode. A mode switching trigger signal is generated to control the capacitor module thermal management system to switch from the current cooling mode to the target cooling mode.

[0101] The mode switching trigger signal includes information such as a switching timestamp, source mode identifier, target mode identifier, and switching priority. The switching priority is determined based on the difference between the spatial geometric relation quantity and the relation threshold; the larger the difference, the higher the priority. The trigger signal is transmitted to the cooling system execution unit via the control bus to perform the corresponding mode switching operation, realizing multi-mode adaptive switching control for capacitor module thermal management.

[0102] In one optional implementation, path optimization is initiated using a mode switching trigger signal. The path integral potential energy and cumulative risk value of the state space path corresponding to each candidate cooling mode are calculated to determine a comprehensive index. The candidate cooling mode with the smallest comprehensive index is selected as the target cooling mode.

[0103] Receive the mode switching trigger signal, extract the position coordinates and motion velocity vector of the trajectory point corresponding to the current thermal state, construct the path search domain with the current trajectory point as the starting point in the state evolution space, extend the normal of the equipotential surface boundary corresponding to each candidate cooling mode to obtain the equipotential surface gradient field, and perform path tracking along the direction of the equipotential surface gradient field until the steady state point of each candidate cooling mode is reached, thus obtaining the state space path corresponding to each candidate cooling mode.

[0104] Energy density is calculated for sampling points on the state space path. The line integral of the energy density along the path is used as the path integral potential energy. The shortest distance from the sampling point on the state space path to the boundary of the energy accumulation region is calculated. The line integral of the product of the shortest distance and the heat flux gradient magnitude of the sampling point along the path is used as the cumulative risk value.

[0105] The comprehensive index is obtained by multiplying the path integral potential energy by a preset potential energy weighting coefficient and the cumulative risk value by a preset risk weighting coefficient. The comprehensive index of each candidate cooling mode is compared and the candidate cooling mode with the smallest comprehensive index is selected as the target cooling mode.

[0106] In one specific implementation, upon receiving a mode switching trigger signal, the three-dimensional coordinate vector of the trajectory point corresponding to the current thermal state is immediately read from the state evolution space, including temperature state variable components, heat flux state variable components, and time dimension coordinates. Simultaneously, the velocity vector of this trajectory point is extracted. This velocity vector is composed of the temperature change rate and the heat flux change rate, and its direction indicates the evolution trend of the system state. A spherical or ellipsoidal path search domain is established in the state evolution space centered on the current trajectory point. The radius of the search domain is determined based on the statistical characteristics of historical path data to ensure coverage of all possible state transition directions.

[0107] Path planning is performed for multiple preset candidate cooling modes. The equations of the equipotential surface boundary curves corresponding to each candidate cooling mode are extracted, and the normal vector field of the surface is calculated. The equipotential surface boundary is numerically extended along the normal vector direction, with an extension step size set to 0.1 times the state space grid size and an extension depth of 2 to 3 times the equipotential surface thickness, forming an equipotential surface gradient field containing gradient information. In the gradient field, the rate of change of energy density along the normal is the largest, representing the fastest path for the system to evolve to steady state. Starting from the current trajectory point, path tracing is performed along the equipotential surface gradient field direction, using the fourth-order Runge-Kutta method for numerical integration, with the tracing step size adaptively adjusted to ensure path accuracy. Path tracing is considered complete when the velocity vector magnitude of the path point decreases to below 5% of the initial value, or when the path point enters the neighborhood of the steady-state point of the candidate cooling mode. The neighborhood radius of the steady-state point is determined based on the standard deviation of the temperature fluctuation of the cooling mode, ensuring that the system enters the stable operating region.

[0108] After obtaining the state-space paths corresponding to each candidate cooling mode, each path is discretized and sampled. The sampling interval is dynamically adjusted according to the path curvature, with denser sampling in areas of greater curvature to ensure no loss of path detail information. The energy density value is calculated for each sampling point, obtained by a weighted sum of the quadratic term of the temperature state variable and the square root term of the heat flux state variable. The weighting coefficients are calibrated based on the heat capacity and heat dissipation coefficient of the capacitor module. The energy density values ​​of all sampling points are multiplied by the path element length between adjacent sampling points and summed to obtain the path integral potential energy. This potential energy characterizes the total energy dissipated by the system along this path.

[0109] To calculate the cumulative risk value, for each sampling point on the state-space path, the Euclidean distance to the boundaries of all energy accumulation regions is calculated, and the minimum value is selected as the safety margin for that sampling point. The boundaries of energy accumulation regions are identified through curvature tensor eigenvalue analysis; the maximum eigenvalue of the curvature tensor at the boundary exceeds a preset threshold. The magnitude of the heat flux gradient at the sampling point is calculated, reflecting the drastic change in local heat flux density. Multiplying the shortest distance by the magnitude of the heat flux gradient yields the local risk index for that sampling point. Line integration is performed along the path for the local risk indices of all sampling points, considering the path element length during integration, to obtain the cumulative risk value. This value quantifies the degree to which the path approaches the danger zone and the cumulative effect of heat flux fluctuations.

[0110] For each candidate cooling mode, its corresponding path integral potential energy is multiplied by a preset potential energy weighting coefficient, which reflects energy consumption priority and typically ranges from 0.4 to 0.6. The cumulative risk value is then multiplied by a preset risk weighting coefficient, which reflects safety priority and typically ranges from 0.4 to 0.6. The sum of these two coefficients is 1. The combined weighted results are then summed to obtain the comprehensive index of the candidate cooling mode. The comprehensive index values ​​of all candidate cooling modes are compared, and the candidate cooling mode with the lowest comprehensive index is selected as the target cooling mode, thus achieving the path optimization objective.

[0111] In one optional implementation, the cooling intensity control sequence is obtained by inverting the state space path corresponding to the target cooling mode, and the cooling execution unit is controlled to adjust the cooling intensity according to the cooling intensity control sequence, including:

[0112] Extract the energy density gradient and heat flux density vector of each sampling point on the state space path corresponding to the target cooling mode, construct the energy backtracking path from each sampling point to the heat source of the capacitor module, mark the energy transfer nodes on the energy backtracking path, and solve the source cooling demand corresponding to each sampling point in reverse by using the thermal resistance network at the energy transfer node. Convert the source cooling demand into the cooling intensity value that the cooling execution unit should output at the corresponding time of the sampling point.

[0113] The cooling intensity values ​​corresponding to each sampling point are matched and verified by pattern feature matching. The cooling intensity values ​​are substituted into the state evolution law of the target cooling mode for forward verification. The cooling intensity deviation points that cause the state evolution to deviate from the state space path are identified. The cooling intensity values ​​of the cooling intensity deviation points are iteratively corrected until the state evolution trajectory converges to the state space path. The corrected cooling intensity values ​​are combined according to the time series to form a cooling intensity control sequence.

[0114] The cooling intensity control sequence is sent to the cooling execution unit, which then controls the cooling execution unit to adjust the cooling intensity sequentially according to the cooling intensity value corresponding to each moment in the cooling intensity control sequence.

[0115] In one specific implementation, multiple uniformly distributed sampling points are extracted from the state space path. The number of sampling points is determined based on the path complexity, typically 20 to 30 points. Sampling point selection follows an equal time interval principle, with each sampling point corresponding to a time point in the mode switching process. For each sampling point, its energy density gradient and heat flux density vector in the three-dimensional state space are calculated. The energy density gradient is calculated using the central difference method, taking the energy difference in six directions around the sampling point and dividing it by the corresponding spatial distance to obtain the rate of change of energy in space. The heat flux density vector is determined by the ratio of heat flux to area, characterizing the direction and magnitude of heat flow per unit area.

[0116] When constructing the energy backtracking path from the sampling point to the heat source of the capacitor module, a reverse heat flux tracing algorithm is used. This algorithm starts at the sampling point and advances step-by-step in the opposite direction of the heat flux density vector. The step size is set according to the spatial accuracy requirements, typically 0.5 to 1 cm. During the advancement, the heat flux density vector at the current position is recalculated at each step to ensure that the path always follows the opposite direction of heat flux. The backtracking terminates when the tracing reaches the boundary of the heat source region or the heat flux density falls below a preset threshold. The boundary of the heat source region is identified through abrupt temperature gradient changes. The preset threshold is determined based on the system background heat flux level, typically 5% to 10% of the average heat flux density.

[0117] Energy transfer nodes are marked along the energy backtracking path, with the node spacing adaptively adjusted according to the heat flux gradient. In regions with rapid heat flux changes, the node spacing decreases to 0.8 to 1.2 cm; in regions with gradual heat flux changes, the node spacing increases to 1.5 to 2 cm. A detailed thermal resistance network model is established at each transfer node, including thermal resistance elements corresponding to the three heat transfer modes: conduction, convection, and radiation. Conductive thermal resistance is calculated based on the material's thermal conductivity and geometric dimensions, convective thermal resistance is calculated based on the convective heat transfer coefficient and contact area, and radiative thermal resistance is determined based on the Stefan-Boltzmann law and the radiation apparent factor. The thermal resistance network adopts a three-dimensional mesh structure, with each node establishing thermal resistance connections with its surrounding adjacent nodes to form a complete heat transfer network.

[0118] The inverse solution of the thermal resistance network employs the principle of node temperature balance, starting from the temperature at the sampling point and progressing node by node towards the heat source. An energy balance equation is established for each node, including heat exchange terms between that node and all its neighboring nodes. The equations are solved using a matrix iteration method, establishing a relationship matrix between the node temperature vector and the heat flux vector. During the iterative calculation, an initial temperature distribution is assumed, and the heat flux at each node is calculated based on the thermal resistance network. The node temperature is updated according to the heat flux, and this process is repeated until the temperature change is less than the convergence threshold. This method ultimately yields the amount of heat that needs to be removed from the heat source, i.e., the cooling requirement at the source.

[0119] When converting the cooling demand at the source end into the cooling intensity value of the cooling execution unit, the characteristic curve and operating efficiency of the cooling system are considered. The characteristic curve describes the nonlinear relationship between cooling power and the output parameters of the execution unit. The source end cooling demand is mapped to the corresponding cooling intensity value through table lookup interpolation. For air-cooled systems, the cooling intensity value is expressed as fan speed or airflow; for liquid-cooled systems, it is expressed as coolant flow rate or pump pressure. The conversion process also needs to consider the influence of ambient temperature. The cooling efficiency is adjusted with changes in ambient temperature to ensure accurate cooling intensity values ​​under different environmental conditions.

[0120] The calculated cooling intensity values ​​are validated using pattern feature matching to verify whether these values ​​can achieve the expected state evolution effect. The validation process employs forward simulation based on the state evolution law of the target cooling mode, which is extracted from historical operating data using a system identification method. The forward simulation starts from the initial state point and sequentially applies the corresponding cooling intensity values ​​at each sampling point according to the time sequence, calculating the evolution trajectory of the system state. The simulation step size is consistent with the control cycle, typically 3 to 5 seconds. The simulated state evolution trajectory is compared with the expected state space path, and the state deviation at each time point is calculated.

[0121] State deviation calculation employs weighted Euclidean distance, assigning different weights to the three state variables—temperature, heat flux gradient, and heat flux power—to reflect the degree of influence of each variable on system stability. Sampling points where state deviation exceeds a preset threshold are identified and marked as cooling intensity deviation points. The deviation point judgment threshold is set according to the system control accuracy requirements, typically 10% to 20% of the unit distance in the state space. For identified deviation points, the cooling intensity value is iteratively corrected.

[0122] The correction process employs a gradient optimization method, aiming to minimize the state deviation. It calculates the sensitivity of the cooling intensity value to the state deviation (partial derivative) and adjusts the cooling intensity value in the opposite direction of the sensitivity. Sensitivity calculation uses the finite difference method, calculating the change in state deviation for small increases and decreases in the current cooling intensity value, and determining the sensitivity through the difference ratio. The adjustment step size adaptively varies according to the sensitivity; a small step size is used when the sensitivity is high, and a large step size is used when the sensitivity is low, avoiding oscillations or slow convergence during the optimization process. After each adjustment, a forward simulation is performed to verify the new state deviation. If the deviation decreases, the adjustment is accepted; otherwise, the step size is reduced, and the process is restarted. The iterative process continues until the state deviation falls below a threshold or the maximum number of iterations is reached.

[0123] After correcting all deviations, a global verification is performed on the corrected set of cooling intensity values. Global verification involves continuously simulating the entire state evolution process to evaluate the overall fit between the state evolution trajectory and the target path. Evaluation metrics include average deviation, maximum deviation, and trajectory smoothness. Average deviation reflects overall tracking accuracy, maximum deviation characterizes control stability under extreme conditions, and trajectory smoothness evaluates the smoothness of state changes. Once the global verification is successful, the corrected cooling intensity values ​​are organized chronologically to form a complete cooling intensity control sequence.

[0124] The cooling intensity control sequence is sent to the cooling execution unit via a communication interface. Upon receiving the sequence, the execution unit executes the corresponding cooling intensity adjustments according to the timestamp sequence. For air-cooled systems, this involves adjusting the fan speed and airflow direction; for liquid-cooled systems, it involves adjusting the pump speed, flow valve opening, and coolant temperature. During execution, a temperature sensor continuously monitors the capacitor module temperature changes, forming a closed-loop control to ensure that the actual temperature change conforms to the expected path. When a temperature deviation exceeds the safe range, an emergency adjustment mechanism is triggered, temporarily deviating from the preset sequence to prioritize system safety. The system returns to the preset sequence once the temperature returns to the normal range.

[0125] The multi-mode adaptive switching control optimization system for capacitor module thermal management in this embodiment of the invention includes:

[0126] The data acquisition unit is used to acquire temperature field data, current data, and operating parameters of each cooling mode of the capacitor module.

[0127] The state construction unit is used to map temperature state variables and heat flux state variables based on temperature field data and current data, construct the state evolution space and draw the thermal state trajectory.

[0128] The region identification unit is used to identify energy accumulation regions and energy diffusion regions by calculating the curvature tensor and divergence field distribution of the thermal state trajectory;

[0129] The boundary extraction unit is used to extract the equipotential surface boundary corresponding to each cooling mode based on the energy accumulation area and the energy diffusion area, calculate the spatial geometric relationship between the trajectory point corresponding to the current thermal state and the equipotential surface boundary, and generate a mode switching trigger signal when the spatial geometric relationship meets the preset switching conditions.

[0130] The mode optimization unit is used to initiate path optimization by using the mode switching trigger signal, calculate the path integral potential energy and cumulative risk value of the state space path corresponding to each candidate cooling mode, determine the comprehensive index, and select the candidate cooling mode with the smallest comprehensive index as the target cooling mode.

[0131] The control execution unit is used to obtain the cooling intensity control sequence by inverting the state space path corresponding to the target cooling mode, and to control the cooling execution unit to adjust the cooling intensity according to the cooling intensity control sequence.

[0132] A third aspect of the present invention provides an electronic device, comprising:

[0133] processor;

[0134] Memory used to store processor-executable instructions;

[0135] The processor is configured to invoke instructions stored in the memory to execute the aforementioned method.

[0136] A fourth aspect of the present invention provides a computer-readable storage medium having stored thereon computer program instructions that, when executed by a processor, implement the aforementioned method.

[0137] This invention can be a method, apparatus, system, and / or computer program product. The computer program product may include a computer-readable storage medium having computer-readable program instructions loaded thereon for performing various aspects of the invention.

[0138] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A multi-mode adaptive switching control optimization method for thermal management of capacitor modules, characterized in that, include: Acquire temperature field data, current data, and operating parameters for each cooling mode of the capacitor module; Based on temperature field data and current data, temperature state variables and heat flow state variables are mapped to construct a state evolution space and plot the thermal state trajectory. By calculating the curvature tensor and divergence field distribution of the thermal trajectory, energy accumulation and energy diffusion regions can be identified. Based on the energy accumulation area and the energy diffusion area, the equipotential surface boundary corresponding to each cooling mode is extracted, and the spatial geometric relationship between the trajectory point corresponding to the current thermal state and the equipotential surface boundary is calculated. When the spatial geometric relationship meets the preset switching conditions, a mode switching trigger signal is generated. The path optimization is initiated by using a mode switching trigger signal. The path integral potential energy and cumulative risk value of the state space path corresponding to each candidate cooling mode are calculated to determine the comprehensive index. The candidate cooling mode with the smallest comprehensive index is selected as the target cooling mode. The cooling intensity control sequence is obtained by inverting the state space path corresponding to the target cooling mode, and the cooling execution unit is controlled to adjust the cooling intensity according to the cooling intensity control sequence.

2. The method according to claim 1, characterized in that, Based on temperature field data and current data, temperature state variables and heat flux state variables are mapped, and a state evolution space is constructed and a thermal state trajectory is plotted, including: The temperature field data is divided into multiple sub-regions according to spatial location. The average temperature and temperature variance of each sub-region are calculated. The average temperature is used as the first temperature state component and the temperature variance is used as the second temperature state component. They are combined to form the temperature state variable. The current change rate is obtained by time difference analysis of the current data. The thermal power of each sub-region is calculated based on the current change rate and the heat capacity of each sub-region. The total thermal power is obtained by spatial summation. The thermal power distribution non-uniformity is obtained by spatial standard deviation of the thermal power of each sub-region. The total thermal power is used as the first thermal flux state component and the thermal power distribution non-uniformity is used as the second thermal flux state component. They are combined to form the thermal flux state variable. A temperature subspace is established based on the first temperature state component and the second temperature state component, and a heat flow subspace is established based on the first heat flow state component and the second heat flow state component. The temperature subspace and the heat flow subspace are combined to form a four-dimensional state evolution space. The temperature and heat flux state variables at each moment are mapped to state points in the state evolution space in chronological order, and the state points at adjacent moments are connected in sequence to form a thermal state trajectory.

3. The method according to claim 1, characterized in that, By calculating the curvature tensor and divergence field distribution of the thermal trajectory, energy accumulation and energy diffusion regions are identified, including: The thermal trajectory is tangent-space projected, and the trajectory motion is decomposed into a thermal potential component along the temperature state direction and an energy flow component along the heat flow state direction. The rate of change of the angle between the thermal potential component and the energy flow component is calculated as the trajectory torsion. The trajectory torsion is combined with the trajectory curvature to construct a curvature tensor. The divergence field is obtained by performing divergence calculation on the heat flow state variables in the state evolution space. The divergence field is then decomposed into multi-scale wavelet components to extract the first-scale divergence components and the second-scale divergence components. The corresponding spatial correlation coefficients are calculated, and the divergence components are weighted and fused based on the spatial correlation coefficients to obtain the reconstructed divergence field. In the curvature tensor, identify torsional abrupt change points where the torsional degree exceeds a preset torsional threshold. In the reconstructed divergence field, identify divergence reversal points where the divergence value changes from negative to positive or from positive to negative. Calculate the spatiotemporal matching degree between the torsional abrupt change points and the divergence reversal points. Determine the region where the spatiotemporal matching degree exceeds a preset matching threshold as the energy phase transition region. Cross-correlation analysis is performed on the temperature state variables and heat flux state variables within the energy phase transition region. The phase lag of the temperature state variables relative to the heat flux state variables is calculated, and the phase lag is compared with a preset lag threshold to determine the energy accumulation region or the energy diffusion region.

4. The method according to claim 1, characterized in that, Based on the energy accumulation region and the energy diffusion region, the equipotential surface boundary corresponding to each cooling mode is extracted. The spatial geometric relationship between the trajectory point corresponding to the current thermal state and the equipotential surface boundary is calculated. When the spatial geometric relationship meets the preset switching conditions, a mode switching trigger signal is generated, including: In the state evolution space, the thermal state distribution of each cooling mode is sliced ​​into isosurfaces, and the isosurfaces at the boundary between the energy accumulation region and the energy diffusion region are extracted to obtain the transformation equipotential surface of the cooling mode at the accumulation-diffusion transition point. The envelope surface is obtained by fitting multiple transformation equipotential surfaces of each cooling mode under different temperature isosurfaces. The intersection of the envelope surface and the boundary of the state evolution space is taken as the equipotential surface boundary of the cooling mode. Calculate the shortest distance from the trajectory point corresponding to the current thermal state to the boundary of the equipotential surface corresponding to each cooling mode, calculate the angle between the direction of movement of the trajectory point and the boundary of the equipotential surface, and combine the shortest distance and the angle to construct a spatial geometric relationship quantity. The spatial geometric relationship quantity is compared with the preset relationship threshold. When the spatial geometric relationship quantity is less than the relationship threshold, it is determined that the current thermal state trajectory point has crossed the equipotential surface boundary. The cooling mode corresponding to the equipotential surface boundary with the smallest spatial geometric relationship quantity is identified as the target cooling mode, and a mode switching trigger signal is generated.

5. The method according to claim 1, characterized in that, The optimization of the cooling path is initiated by using a mode switching trigger signal. The path integral potential energy and cumulative risk value of the state space path corresponding to each candidate cooling mode are calculated to determine the comprehensive index. The candidate cooling mode with the smallest comprehensive index is selected as the target cooling mode, including: Receive the mode switching trigger signal, extract the position coordinates and motion velocity vector of the trajectory point corresponding to the current thermal state, construct the path search domain with the current trajectory point as the starting point in the state evolution space, extend the normal of the equipotential surface boundary corresponding to each candidate cooling mode to obtain the equipotential surface gradient field, and perform path tracking along the direction of the equipotential surface gradient field until the steady state point of each candidate cooling mode is reached, thus obtaining the state space path corresponding to each candidate cooling mode. Energy density is calculated for sampling points on the state space path. The line integral of the energy density along the path is used as the path integral potential energy. The shortest distance from the sampling point on the state space path to the boundary of the energy accumulation region is calculated. The line integral of the product of the shortest distance and the heat flux gradient magnitude of the sampling point along the path is used as the cumulative risk value. The comprehensive index is obtained by multiplying the path integral potential energy by a preset potential energy weighting coefficient and the cumulative risk value by a preset risk weighting coefficient. The comprehensive index of each candidate cooling mode is compared, and the candidate cooling mode with the smallest comprehensive index is selected as the target cooling mode.

6. The method according to claim 1, characterized in that, The cooling intensity control sequence is obtained by inverting the state space path corresponding to the target cooling mode. The cooling execution unit adjusts the cooling intensity according to the cooling intensity control sequence, including: Extract the energy density gradient and heat flux density vector of each sampling point on the state space path corresponding to the target cooling mode, construct the energy backtracking path from each sampling point to the heat source of the capacitor module, mark the energy transfer nodes on the energy backtracking path, and solve the source cooling demand corresponding to each sampling point in reverse by using the thermal resistance network at the energy transfer node. Convert the source cooling demand into the cooling intensity value that the cooling execution unit should output at the corresponding time of the sampling point. The cooling intensity values ​​corresponding to each sampling point are matched and verified by pattern feature matching. The cooling intensity values ​​are substituted into the state evolution law of the target cooling mode for forward verification. The cooling intensity deviation points that cause the state evolution to deviate from the state space path are identified. The cooling intensity values ​​of the cooling intensity deviation points are iteratively corrected until the state evolution trajectory converges to the state space path. The corrected cooling intensity values ​​are combined according to the time series to form a cooling intensity control sequence. The cooling intensity control sequence is sent to the cooling execution unit, which then controls the cooling execution unit to adjust the cooling intensity sequentially according to the cooling intensity value corresponding to each moment in the cooling intensity control sequence.

7. A multi-mode adaptive switching control optimization system for capacitor module thermal management, used to implement the method of any one of claims 1-6, characterized in that, include: The data acquisition unit is used to acquire temperature field data, current data, and operating parameters of each cooling mode of the capacitor module. The state construction unit is used to map temperature state variables and heat flux state variables based on temperature field data and current data, construct the state evolution space and draw the thermal state trajectory. The region identification unit is used to identify energy accumulation regions and energy diffusion regions by calculating the curvature tensor and divergence field distribution of the thermal state trajectory; The boundary extraction unit is used to extract the equipotential surface boundary corresponding to each cooling mode based on the energy accumulation area and the energy diffusion area, calculate the spatial geometric relationship between the trajectory point corresponding to the current thermal state and the equipotential surface boundary, and generate a mode switching trigger signal when the spatial geometric relationship meets the preset switching conditions. The mode optimization unit is used to initiate path optimization by using the mode switching trigger signal, calculate the path integral potential energy and cumulative risk value of the state space path corresponding to each candidate cooling mode, determine the comprehensive index, and select the candidate cooling mode with the smallest comprehensive index as the target cooling mode. The control execution unit is used to obtain the cooling intensity control sequence by inverting the state space path corresponding to the target cooling mode, and to control the cooling execution unit to adjust the cooling intensity according to the cooling intensity control sequence.

8. An electronic device, characterized in that, include: processor; Memory used to store processor-executable instructions; The processor is configured to invoke instructions stored in the memory to execute the method according to any one of claims 1 to 6.

9. A computer-readable storage medium having computer program instructions stored thereon, characterized in that, When the computer program instructions are executed by the processor, they implement the method described in any one of claims 1 to 6.

Images