Thermal management controller regulation method and system based on dynamic thermal inertia prediction

Abstract

The present application relates to the technical field of thermal management control, and specifically discloses a thermal management controller regulation and control method and system based on dynamic thermal inertia prediction. The present application constructs a dynamic thermal inertia model through system identification, accurately predicts the future temperature field evolution and thermal accumulation risk of the system. On this basis, a multi-layer nested compensation strategy composed of pre-compensation, main compensation and residual compensation is generated in three dimensions of compensation angle, compensation order and compensation domain, realizing forward-looking and accurate intervention. In order to evaluate the control effect, the artifact intensity entropy value is introduced as a quantitative index, and the model and strategy are corrected in real time through an outer loop optimization mechanism, forming a self-correcting function. When persistent thermal imbalance caused by hardware design defects is detected, the system can further start hardware collaborative regulation, fundamentally optimize local thermal capacity and heat dissipation path, realize cross-level collaborative optimization from control algorithm to physical structure, and ultimately achieve high-precision, low-energy-consumption and self-adaptive system-level thermal management.

Description

Technical Field

[0001] This invention relates to the field of thermal management control technology, and more specifically to a thermal management controller control method and system based on dynamic thermal inertia prediction. Background Technology

[0002] In high-efficiency systems such as electronic devices, high-energy-density battery systems, and aerospace equipment, thermal management accuracy directly determines the system's reliability, lifespan, and performance. Currently widely used thermal management control methods, such as PID-based feedback control, while simple in structure and easy to implement, rely on historical and current temperature data for adjustment, resulting in significant response delays and overshoot issues. Experimental data shows that under power transient scenarios, the response delay of traditional PID control can reach 3–5 seconds, with temperature fluctuations exceeding ±5°C, making it difficult to meet the management requirements of high-dynamic thermal loads.

[0003] To further improve control adaptability, some studies have introduced fuzzy control or neural network control methods. However, these still heavily rely on real-time sensor data and lack the ability to predict the future thermal evolution trend of the system. For example, in lithium-ion battery modules, due to the influence of internal thermal inertia, external temperature sensors often detect hot spots 10–30 seconds too late. This prevents the control system from effectively intervening before thermal runaway, causing local temperature rises to exceed safety thresholds, which in turn leads to system performance degradation or even thermal failure.

[0004] Existing thermal management strategies largely remain at the control algorithm level, lacking dynamic coordination with hardware architecture. When hardware-level thermal imbalances occur due to uneven heat capacity distribution or flawed heat dissipation path design, traditional methods can only compensate by continuously increasing cooling intensity, which is not only energy-intensive but also has limited effectiveness. Real-world data shows that under sustained high load conditions, cooling energy consumption in local hotspot areas of such systems can increase by more than 40%, while the temperature still struggles to remain within the ideal range. Therefore, there is an urgent need for an intelligent thermal management method that integrates thermal behavior prediction, multi-dimensional compensation, and online hardware reconfiguration to achieve system-level thermal optimization and efficient energy utilization. Summary of the Invention

[0005] To address the response delay and hardware thermal imbalance issues in traditional thermal management, this invention provides a thermal management controller control method and system based on dynamic thermal inertia prediction. By constructing a dynamic thermal inertia model to predict the future temperature field evolution of the system, a multi-layered nested strategy including pre-compensation, main compensation, and residual compensation is generated in the three dimensions of compensation angle, compensation order, and compensation domain. The strategy is optimized by introducing artifact intensity entropy value combined with an outer loop optimization mechanism. Simultaneously, hardware collaborative control is initiated when persistent thermal imbalance is detected, ultimately achieving accurate prediction of system thermal behavior, adaptive compensation, and collaborative optimization of hardware-level thermal management.

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

[0007] According to one aspect of this application, a thermal management controller control method based on dynamic thermal inertia prediction is provided, comprising:

[0008] Collect thermal field data during the operation of the thermal management system, and construct a dynamic thermal inertia model based on the thermal field data;

[0009] Based on the dynamic thermal inertia model, the future temperature field evolution and thermal accumulation risk of the system are predicted, and compensation strategies are generated in the three dimensions of compensation angle, compensation order and compensation domain space respectively.

[0010] The compensation strategy for each dimension includes a three-layer nested strategy of pre-compensation, main compensation, and residual compensation.

[0011] Among them, the pre-compensation strategy quickly suppresses the initial disturbance, the main compensation strategy offsets the predicted thermal deviation, and the residual compensation strategy addresses the uncertainty of the dynamic thermal inertia model.

[0012] Collect thermal field data after applying the compensation strategy and calculate the artifact intensity entropy value of the thermal field data. If the artifact intensity entropy value does not reach the preset threshold, it is determined that the compensation strategy has not achieved the expected results. Then, the dynamic thermal inertia model is optimized through the outer loop optimization mechanism, the compensation strategy is regenerated and the artifact intensity entropy value is calculated until the artifact intensity entropy value reaches the preset threshold. If the artifact intensity entropy value reaches the preset threshold, it proves that the compensation strategy is effective and continues to be executed.

[0013] If, after adopting an effective compensation strategy, it is found that a specific area requires a continuous high-intensity similar compensation strategy under various different operating conditions, it is determined that there is a continuous hardware-level thermal imbalance in the specific area. The controller will initiate hardware collaborative regulation to optimize the heat dissipation path and regenerate and verify the compensation strategy based on the thermal field data established after the hardware change.

[0014] As a further option of the method in this application, the dynamic thermal inertia model is constructed through spatial discretization and time-difference solution, and its state-space equation is: ;in: yes Temperature state vector of all units at any given time; yes The control input vector at time t, yes The control input vector at time t, It is the system state matrix. It is a control input matrix. It is the thermal accumulation effect matrix. It is the length of the time window that takes into account the heat accumulation effect.

[0015] As a further option of the method in this application, the prediction of the future temperature field evolution of the system is achieved by iteratively applying a dynamic thermal inertia model, specifically: ;in, Indicates in Always looking towards the future The predicted temperature state vector at any given time. Indicates in Always looking towards the future The predicted temperature state vector at any given time. Indicates in Always looking towards the future Control the input vector at all times.

[0016] As a further option of the method in this application, the compensation angle dimension is used to find the most effective compensation direction;

[0017] The compensation strategy for the compensation angle dimension is determined by solving a spatial optimization problem. The optimization objective is to minimize the maximum spatial gradient of the temperature field or the thermal stress, i.e.: ;in, For compensation angle, It depends on the angle The compensation operator, It is a spatial gradient operator. It is a node Thermal deviation;

[0018] After solving the optimization problem using gradient descent, the optimal compensation angle is obtained. Based on this, the spatial weight distribution of pre-compensation, main compensation, and residual compensation is generated.

[0019] As a further option of the method in this application, the strengths of pre-compensation, main compensation, and residual compensation in the compensation strategy of the compensation order dimension are determined in the following ways:

[0020] Pre-compensation compensation stage The pre-compensation is proportional to the norm of the current thermal deviation, and its specific form is as follows: ,in, It is pre-compensation of the compensation order dimension. It is a node thermal deviation, For proportional gain;

[0021] The compensation order of the main compensation is obtained by solving the finite-time optimal control problem, and the objective function is: ;in, It's thermal deviation. Indicates in Always looking towards the future Control the input vector at all times. The main compensation sequence is obtained by controlling the input weights and solving the objective function.

[0022] Compensation order of residual compensation It is proportional to the integral of the observation and prediction deviation, i.e. Specific forms of residual compensation ,in At any moment The observed actual thermal deviation, It is a past moment to a moment Predicted thermal deviation, To compensate for the gain of the residual, From the initial moment to the current moment The points.

[0023] As a further option of the method in this application, in the compensation strategy of the compensation domain dimension, the time domains of pre-compensation, main compensation, and residual compensation are respectively defined as:

[0024] Pre-compensation compensation domain For a short time window, that is ,in, It is an integer between 1 and 3. Sampling time;

[0025] The compensation domain of the main compensation Covering the entire prediction time domain ;

[0026] Compensation domain of residual compensation It is a continuous time domain that covers the entire control process.

[0027] As a further option of the method in this application, the calculation of the artifact intensity entropy value includes:

[0028] For each unit The temperature time series was low-pass filtered to obtain a smooth trajectory;

[0029] according to Calculate the artifact intensity for each cell; the larger the artifact intensity value, the stronger the unexpected temperature fluctuation at that cell.

[0030] according to Normalizing artifact intensity yields probability distribution , making :

[0031] according to Calculate the entropy value of artifact intensity ;

[0032] in, For the first The artifact intensity of each node, To assess the length of the time window, For the first Each node at time... The actual observed temperature For the first Each node at time... The desired smooth temperature trajectory, For the first The proportion of artifact intensity at each node, The total number of nodes in the system. It is the natural logarithm. For the first The artifact intensity of each node.

[0033] As a further option of the method in this application, the outer loop optimization mechanism includes using a recursive least squares method with a forgetting factor or a sliding window batch processing algorithm to update the system state matrix, control input matrix, and thermal accumulation effect matrix of the dynamic thermal inertia model online. The optimization objective is to minimize the prediction error of recent data, thereby adaptively correcting the model bias.

[0034] As a further option of the method in this application, the hardware coordinated control includes:

[0035] Microfluidic valves are controlled to guide the deposition of phase change materials in the thermal accumulation region to increase local heat capacity;

[0036] The shape memory alloy guide plate is driven to reconstruct the flow channel to optimize the heat dissipation path;

[0037] After the hardware change, thermal field data is re-acquired and the dynamic thermal inertia model is updated to generate a compensation strategy that matches the new hardware state.

[0038] Another aspect of this application provides a thermal management controller control system based on dynamic thermal inertia prediction, the system comprising:

[0039] The data acquisition module is used to collect thermal field data during the operation of the thermal management system; the model building module is connected to the data acquisition module and is used to build a dynamic thermal inertia model based on the thermal field data.

[0040] The prediction and compensation strategy generation module communicates with the model building module and is used to predict the future temperature field evolution and thermal accumulation risk of the system based on the dynamic thermal inertia model, and generate compensation strategies in three dimensions: compensation angle, compensation order and compensation domain.

[0041] The evaluation and optimization module communicates with the prediction and compensation strategy generation module and the data acquisition module. It is used to collect thermal field data after applying the compensation strategy and calculate the artifact intensity entropy value of the thermal field data. If the artifact intensity entropy value does not reach the preset threshold, it is determined that the compensation strategy has not achieved the expected results. The dynamic thermal inertia model is optimized through the outer loop optimization mechanism, which triggers the prediction and compensation strategy generation module to regenerate the compensation strategy and perform iterative evaluation until the artifact intensity entropy value reaches the preset threshold. If the preset threshold is reached, the compensation strategy is determined to be effective and continues to be executed.

[0042] The hardware collaborative control module communicates with the evaluation and optimization module and the prediction and compensation strategy generation module. After adopting an effective compensation strategy, if it is found that a specific area requires a continuous high-intensity similar compensation strategy under various different operating conditions, it is determined that there is a continuous hardware-level thermal imbalance in the specific area, and hardware collaborative control is initiated to optimize the heat dissipation path. At the same time, the thermal field data established after the hardware change is fed back to the model building module to regenerate the compensation strategy and verify it.

[0043] The beneficial effects of this application are as follows:

[0044] Traditional PID control suffers from a 3–5 second response delay and ±5°C temperature fluctuations in power transient scenarios. This invention, however, uses a dynamic thermal inertia model to predict future multi-step temperature fields and combines this with a three-dimensional, multi-layered nested compensation strategy to proactively intervene before thermal disturbances occur. Experiments show that this method can reduce the response delay to less than 1 second and control temperature fluctuations within ±1°C, effectively avoiding heat buildup and local overshoot problems.

[0045] Traditional methods often rely on continuous high-energy-consumption compensation when dealing with hardware-level thermal imbalances, resulting in localized cooling energy consumption increases of over 40% without stabilizing the temperature. This invention uses artifact intensity entropy values ​​to evaluate control effectiveness in real time and combines an outer-loop optimization mechanism to dynamically correct the model and strategy, significantly reducing energy consumption while maintaining control accuracy. Actual operation data shows that the system's overall cooling energy consumption can be reduced by more than 25% under continuous high-load conditions, and temperature field uniformity can be improved by more than 30%.

[0046] Traditional thermal management methods lack dynamic interaction with hardware, making it difficult to fundamentally solve thermal imbalances caused by insufficient heat capacity or defective heat dissipation paths. This invention identifies persistent high-intensity compensation areas and triggers a hardware collaborative control mechanism to achieve online optimization of local heat capacity and heat dissipation capabilities. Test results show that this mechanism can reduce the temperature of hotspot areas by 8–12°C and maintain long-term stability under various operating conditions, fundamentally improving the overall thermal management efficiency and reliability of the system. Attached Figure Description

[0047] Figure 1This is a schematic diagram of the overall process of the thermal management controller control method based on dynamic thermal inertia prediction.

[0048] Figure 2 The flowchart of the S100 step-by-step control method for thermal management controllers based on dynamic thermal inertia prediction is shown below.

[0049] Figure 3 The flowchart of the S200 step-by-step control method for thermal management controllers based on dynamic thermal inertia prediction;

[0050] Figure 4 The flowchart of the S300 thermal management controller control method based on dynamic thermal inertia prediction is shown below.

[0051] Figure 5 The flowchart shows the steps of the S400 thermal management controller control method based on dynamic thermal inertia prediction. Detailed Implementation

[0052] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of this application.

[0053] In modern industrial products such as high-power-density electronic devices, high-end energy storage systems, and aerospace vehicles, the performance of the thermal management system directly determines the system's reliability, lifespan, and efficiency. Traditional thermal management control methods, such as PID-based feedback control, heavily rely on current and historical temperature sensor data. Due to the inherent delay and cumulative effect of heat transfer—thermal inertia—existing methods struggle to proactively intervene before thermal disturbances occur. This leads to risks such as response lag, temperature overshoot, and even localized thermal runaway when facing rapidly changing thermal loads. Existing technologies lack the ability to accurately predict the system's future thermal behavior, fail to systematically consider the dynamic impact of thermal inertia in control strategies, and cannot achieve adaptive synergy between control algorithms and hardware architecture, resulting in limited overall thermal management efficiency.

[0054] This invention accurately describes the thermal delay and cumulative effect of a system by constructing a dynamic thermal inertia model. Based on the model prediction results, a multi-layered nested compensation strategy is generated in the three-dimensional space of compensation angle, compensation order, and compensation domain. Finally, the artifact intensity entropy value is introduced as an evaluation index of the control effect. Combined with outer loop optimization and hardware collaborative control, intelligent, precise, and adaptive thermal management is achieved.

[0055] The core theory is as follows:

[0056] At any point and at any time Thermal state using temperature field Description. Thermal inertia is essentially a comprehensive reflection of the distribution of heat capacity and thermal resistance in a system, resulting in a delay and accumulation of temperature changes relative to changes in heat source power. To quantify this dynamic process, a dynamic thermal inertia model is introduced, the core of which is a partial differential equation containing delay and accumulation terms: ;in: The density of the material; Specific heat capacity of the material; For thermal conductivity tensor; The term represents heat conduction, describing the heat diffusion process caused by the temperature gradient. It is a temperature gradient. It is the heat flux density vector obtained according to Fourier's law. It is a divergence operator, representing the net outflow rate of heat flux density; To generate heat flux density for the internal heat source; This refers to the heat flux density lost by the system to the environment. The heat accumulation coefficient characterizes the extent to which historical heat production affects the current temperature. is the thermal delay time constant, which characterizes the main time scale of thermal response delay. For the historical heat production accumulation term, the calculation from the current moment... Looking back The total heat generated by the heat source within a time window of seconds.

[0057] To implement the dynamic thermal inertia model in the controller, the dynamic thermal inertia model is discretized in space and solved using finite-difference time methods. Specifically, the system is divided into... A finite number of volume elements, each element At any moment The temperature is The discretized dynamic thermal inertia model's state-space equations are expressed as: ;in: yes Temperature state vector of all units at any given time; yes The control input vector at each time step; It is the system state matrix, which describes the dynamics of the temperature field's own evolution and includes information such as heat conduction and convection; It is the control input matrix, which describes the immediate effect of the current control input on the temperature field; It is the length of the time window that considers the heat accumulation effect, and Related, , Sampling time. Model parameters. Using system identification methods, thermal field data collected during operation is utilized. Perform offline or online estimation. It is the thermal cumulative effect matrix, which describes the cumulative effect of historical control inputs on the current temperature field; It is the cumulative sum of historical control inputs, that is, the sum of inputs from the current moment. Backtracking The sum of historical control input vectors at each time step.

[0058] Based on the discretized dynamic thermal inertia model state-space equations, future... Temperature field prediction step: ;in, Indicates in Always looking towards the future The predicted temperature state vector at any given time. Indicates in Always looking towards the future Control the input vector at all times.

[0059] Predicted future temperature field and expected temperature field The deviation is defined as thermal deviation. Risk of thermal buildup Then by identifying thermal deviation Exceeding the safety threshold The region is used to determine this.

[0060] The compensation strategy is generated in three dimensions:

[0061] Compensation Angle This represents the directionality of the compensation effect, that is, in the spatial distribution of the temperature field, in which direction or region the compensation energy should be preferentially applied to most effectively counteract thermal deviations. This is determined by solving an optimization problem with the objective of minimizing the maximum temperature gradient or thermal stress in space.

[0062] Compensation order This represents the strength or magnitude of the compensating effect, i.e., how much compensating energy needs to be applied.

[0063] Compensation domain The time domain representing the compensation effect refers to how long the compensation action needs to last, or how long in the future the compensation plan will be carried out.

[0064] In each dimension, the compensation strategy consists of a three-layer nested structure:

[0065] Pre-compensation Aimed at quickly suppressing initial disturbances and pre-compensating The output is related to the current thermal deviation or its derivative, resulting in a fast response speed. ;in To pre-compensate gain, For compensation angle Spatial weighting function.

[0066] Main compensation The aim is to offset the predicted thermal deviation at the core, calculated based on the principle of model predictive control, namely: ;in These are the weights of the control input vector, used to balance control effectiveness and energy consumption.

[0067] Residual compensation To address the uncertainties of dynamic thermal inertia models, robust or adaptive control methods are employed, namely: ;in At any moment The observed actual thermal deviation, It is a past moment to a moment Predicted thermal deviation, To compensate for the gain of the residual, From the initial moment to the current moment The points.

[0068] Overall compensation strategy The combination of the three: .

[0069] To evaluate the effectiveness of the compensation strategy, an artifact intensity entropy value is defined. Artifacts refer to unintended and harmful temperature fluctuations or spatial inhomogeneities introduced by compensation strategies. Artifact intensity. At each node The above is defined as the deviation between the actual temperature trajectory and the desired smooth trajectory, i.e.: ;in, For the first The artifact intensity of each node, To assess the length of the time window, For the first Each node at time... The actual observed temperature. For the first Each node at time... The desired smooth temperature trajectory.

[0070] Artifact intensity entropy value of the entire temperature field The calculation is as follows: ;in, For the first The proportion of artifact intensity at each node, The total number of nodes in the system. It is the natural logarithm. For the first The artifact intensity of each node. Artifact intensity entropy value. This indicates that the higher the disorder of the temperature field, the stronger the unexpected effects introduced by the compensation strategy, and the worse the control effect.

[0071] The specific embodiments of the present invention will be described in detail below.

[0072] Example 1:

[0073] Please see Figure 1 This illustrates a thermal management controller control method based on dynamic thermal inertia prediction provided by an embodiment of the present invention, the method comprising:

[0074] S100: By collecting thermal field data, a dynamic thermal inertia model is constructed that can describe the thermal delay and cumulative effects of the system.

[0075] S200: Based on this model, predict future temperature fields and risks, and generate multi-layered nested strategies consisting of pre-compensation, main compensation and residual compensation in the three dimensions of compensation angle, order and domain.

[0076] S300: Quantifies the control effect by calculating the artifact intensity entropy value, and initiates outer loop optimization based on this criterion to dynamically correct the model and strategy to achieve self-correction.

[0077] S400: When a persistent thermal imbalance caused by hardware design limitations is identified, it automatically triggers an adaptive reconfiguration at the hardware level and updates it in conjunction with the control strategy to achieve cross-level collaborative optimization.

[0078] The specific plan is as follows:

[0079] In a thermal management controller control method based on dynamic thermal inertia prediction, S100 provides a model foundation for the entire prediction and control process by acquiring system operation data and constructing a dynamic thermal inertia model. The dynamic thermal inertia model captures the delay and energy accumulation effects in the heat transfer process by integrating real-time thermal field data and system identification technology.

[0080] Please refer to Figure 2 The diagram illustrates a flowchart of an exemplary thermal management controller control method S100 based on dynamic thermal inertia prediction, which includes:

[0081] S110: Deploy a sensor network and synchronously collect multi-source thermal field data, and preprocess the multi-source thermal field data.

[0082] Specifically, sensors are installed in key parts of the thermal management system to collect system control thermal field data simultaneously.

[0083] In one possible implementation, the data acquisition system employs a distributed architecture to perform synchronous acquisition at a fixed sampling period.

[0084] Specifically, thermal field data preprocessing involves performing preprocessing operations such as filtering and noise reduction, outlier removal, and data alignment on the thermal field data, and extracting feature quantities for system identification.

[0085] S120: Perform system identification based on the preprocessed thermal field data using a dynamic thermal inertia model.

[0086] Specifically, using the thermal field data obtained from S120 processing, a system identification algorithm is employed to estimate the parameters of the dynamic thermal inertia model. ,in: It is the system state matrix. It is a control input matrix. It is the thermal accumulation effect matrix.

[0087] In one possible implementation, the system identification algorithm employs a subspace identification method or a prediction error minimization method, using the thermal field data obtained from S120 to determine the parameters. , , .

[0088] S130: Construct a dynamic thermal inertia model.

[0089] Specifically, a dynamic thermal inertia model describing the system's thermal response delay and cumulative effect is constructed based on thermal field data.

[0090] In one possible implementation, the dynamic thermal inertia model is as follows: ;in: yes Temperature state vector of all units at any given time; yes The control input vector at each time step; It is the length of the time window that takes into account the heat accumulation effect. It is the cumulative sum of historical control inputs.

[0091] The dynamic thermal inertia model was finally validated. After the validation was passed, the model parameters were... , , , The model is loaded into the controller's memory, completing the model initialization. The initialized model is then integrated into the controller and can be used for real-time prediction and control.

[0092] In one possible implementation, model validation includes: evaluating the model's generalization ability on a test dataset different from the training and validation datasets; and checking the model's stability to ensure the system state matrix... The eigenvalues ​​are all within the unit circle, ensuring that the prediction does not diverge.

[0093] In a thermal management controller control method based on dynamic thermal inertia prediction, S200 uses the dynamic thermal inertia model established by S100 as the core to predict future thermal behavior and generate a three-dimensional multi-layer nested compensation strategy.

[0094] Please refer to Figure 3 The diagram illustrates a flowchart of an exemplary thermal management controller control method S200 based on dynamic thermal inertia prediction, the contents of which include:

[0095] S210: Performs multi-step look-ahead temperature field prediction based on a dynamic thermal inertia model.

[0096] Specifically, utilizing the current moment Temperature state Based on the established dynamic thermal inertia model and historical control input sequences, future predictions are made in a rolling manner. Temperature field at a time step The prediction process is achieved by iteratively applying the state-space equations of a discretized dynamic thermal inertia model: in, Indicates in Always looking towards the future The control input vector is controlled at each step. During the initial prediction, if the future control input is unknown, it can be assumed to maintain its current value or initialized according to a certain strategy.

[0097] In one possible implementation, the prediction time domain is... The main thermal response time of the coverage system.

[0098] S220: Calculate thermal deviation and assess thermal accumulation risk based on predicted temperature field.

[0099] Specifically, it will predict the temperature field. With respect to the system's desired safe temperature field Compare and calculate thermal deviation It also identifies areas with a risk of heat buildup. Heat buildup risk arises from the accumulation of heat due to thermal inertia, which can cause local temperatures to exceed safe limits.

[0100] In one possible implementation, the risk of thermal buildup. The method of determination is as follows: ,in It is a node The safe temperature threshold. That is to say, in There is always a risk of heat buildup. The size reflects the severity of thermal buildup.

[0101] S230: Generate a space-oriented compensation strategy in the compensation angle dimension.

[0102] Specifically, the compensation angle Determining the compensation angle is a spatial optimization problem, namely, finding the most effective direction of compensation to minimize thermal deviation and thermal stress. The compensation angle represents the directionality of the compensation action, determining which direction or region the compensation energy should be preferentially applied to to most effectively counteract thermal deviation.

[0103] In one possible implementation, the compensation angle This is determined by solving the following optimization problem: ,in, It depends on the angle The compensation operator, It is a spatial gradient operator. The optimization objective is to minimize the maximum spatial temperature gradient introduced by the compensation action, thereby avoiding the generation of new local hot spots or thermal stress concentrations.

[0104] After solving the optimization problem using gradient descent, the optimal compensation angle is obtained. Based on this, a pre-compensation is generated. Main compensation and residual compensation Weight distribution in space.

[0105] S240: A compensation strategy for generating intensity modulation in the compensation order dimension.

[0106] Specifically, compensation order This determines the strength of each compensation component. Compensation order. The amplitude of the compensation energy is modulated based on the magnitude and characteristics of the thermal deviation.

[0107] In one possible implementation:

[0108] Pre-compensation compensation stage It is proportional to the norm of the current thermal deviation, that is To ensure rapid response, the specific form of pre-compensation is as follows: ,in This is the proportional gain.

[0109] The compensation stage of the main compensation This is obtained by solving a finite-time optimal control problem, the objective of which is to minimize the thermal deviation and control energy consumption in the prediction time domain. The problem is formulated as follows: ;in These are the control input weights, used to balance control effectiveness and energy consumption. Solving this optimization problem yields the main compensation sequence. .

[0110] Compensation order of residual compensation It is proportional to the integral of the observation and prediction deviation, i.e. This is used to progressively correct model errors. Specific forms of residual compensation. ,in This is the integral gain.

[0111] S250: Generate compensation strategies for time planning in the compensation domain dimension.

[0112] Specifically, the compensation domain The temporal characteristics of the compensation effect are defined. Compensation domain. The time domain representing the compensation effect refers to how long the compensation action needs to last, or how long in the future the compensation plan will be carried out.

[0113] In one possible implementation:

[0114] Pre-compensation compensation domain Shorter duration, typically lasting only a few control cycles, designed for rapid braking and suppression of initial rapid temperature changes. Pre-compensated compensation domain. Defined as ,in, It is an integer between 1 and 3.

[0115] The compensation domain of the main compensation Covering the entire prediction time domain Long-term optimization planning is carried out. The compensation domain for pre-compensation. for .

[0116] Compensation domain of residual compensation It is continuous; as long as unmodeled dynamics exist in the system, the compensation domain of residual compensation continues to operate. The compensation domain of residual compensation is... It covers the entire control process.

[0117] S260: Synthesizes a three-dimensional compensation strategy and outputs it to the actuator.

[0118] Specifically, the pre-compensation, main compensation, and residual compensation calculated in the three dimensions are vectorized to obtain the final overall control command. The synthesis formula is: This master control command It is then distributed to the corresponding actuators to achieve precise control of the thermal management system.

[0119] In a thermal management controller control method based on dynamic thermal inertia prediction, S300 introduces artifact intensity entropy as a new evaluation index and establishes an outer loop optimization mechanism to achieve self-correction of the compensation strategy and adaptive updating of the model.

[0120] Please refer to Figure 4The diagram illustrates a flowchart of an exemplary thermal management controller control method S300 based on dynamic thermal inertia prediction, the contents of which include:

[0121] S310: Collects real-time thermal field data after applying the compensation strategy.

[0122] Specifically, after implementing the compensation strategy generated by S200, the actual observed temperature of the system is continuously collected through the sensor network. .

[0123] S320: Calculate the entropy value of artifact intensity in the actual temperature field.

[0124] Specifically, the artifact intensity within the current time window is calculated. and its artifact intensity entropy value Artifact intensity entropy value The spatial disorder of unexpected temperature fluctuations introduced by the compensation strategy was quantified.

[0125] In one possible implementation, the calculation steps are as follows:

[0126] For each unit The temperature time series was low-pass filtered to obtain a smooth trajectory.

[0127] according to Calculate the artifact intensity for each cell. Artifact intensity The larger the value, the stronger the unexpected temperature fluctuation at that unit.

[0128] according to Normalized artifact intensity Obtain the probability distribution , making :

[0129] according to Calculate the entropy value of artifact intensity Artifact intensity entropy value The larger the value, the higher the disorder of the temperature field, the stronger the unexpected effects introduced by the compensation strategy, and the worse the control effect.

[0130] S330: Determine the effectiveness of the compensation strategy based on the artifact intensity entropy value.

[0131] Specifically, the calculated artifact intensity entropy value With preset threshold Compare. Threshold The settings are based on the system's requirements for temperature uniformity and stability.

[0132] In one possible implementation, the artifact intensity entropy value With threshold The logic behind the judgment is as follows:

[0133] like If the compensation strategy is deemed effective, the control flow jumps to S340 to continue executing the current strategy.

[0134] like If the compensation strategy fails to meet expectations, the control flow jumps to S350 to initiate outer loop optimization.

[0135] S340: Continuously execute and monitor the current compensation strategy.

[0136] Specifically, once the compensation strategy is deemed effective, the controller will continuously output control commands corresponding to the effective compensation strategy and continue to monitor the artifact intensity entropy value. Ensure that it is always below the threshold. Meanwhile, the controller continuously monitors the system's operating status and thermal load changes, preparing for possible strategy adjustments.

[0137] S350: The outer loop optimization mechanism is activated to optimize the dynamic thermal inertia model.

[0138] Specifically, when the artifact intensity entropy value If the value exceeds the limit, it indicates that there is a significant deviation between the current model and the actual system dynamics. In this case, the dynamic thermal inertia model is optimized and updated through the outer loop optimization mechanism.

[0139] In one possible implementation, the outer-loop optimization mechanism for updating the dynamic thermal inertia model includes:

[0140] Collect complete operational data over a recent period.

[0141] The recursive least squares method with a forgetting factor or the sliding window batch processing algorithm is used to process the parameters of the dynamic thermal inertia model. , , Perform online updates.

[0142] The optimization objective is to minimize the prediction error of recent data.

[0143] After the model is updated, the process returns to S200, where a compensation strategy is regenerated based on the updated model, and the artifact intensity entropy is calculated again. Evaluate.

[0144] In a thermal management controller control method based on dynamic thermal inertia prediction, the S400 detects persistent thermal imbalance modes and initiates hardware collaborative control, realizing cross-level collaborative optimization from control algorithm to physical structure, fundamentally solving thermal problems caused by hardware design defects.

[0145] Please refer to Figure 5 It shows a flowchart of an exemplary thermal management controller control method S400 based on dynamic thermal inertia prediction, the contents of which include:

[0146] S410: Monitor and identify areas of continuous high-intensity compensation.

[0147] Specifically, after the compensation strategy has been running effectively for a period of time, historical data of the compensation strategy is analyzed to identify specific spatial regions that require continuous high-intensity compensation. Continuous application of high-intensity compensation often indicates an imbalance in the underlying hardware thermal design.

[0148] In one possible implementation, the method for identifying specific spatial regions by continuously applying high-intensity compensation includes:

[0149] The compensation order received by each spatial unit under various typical operating conditions is statistically analyzed. The compensation order is directly obtained from the output of the main compensation strategy.

[0150] Set a high-intensity compensation threshold.

[0151] If the average compensation intensity of units within a certain region is greater than the compensation threshold under more than three different operating conditions, then it is determined that there is a persistent hardware-level thermal imbalance in that region.

[0152] S420: Determines hardware-level thermal imbalance and generates hardware control instructions.

[0153] Specifically, once a region of sustained high-intensity compensation is identified, the controller determines that there is a thermal imbalance in that region due to hardware design limitations such as insufficient thermal capacity or excessive thermal resistance, and generates corresponding hardware control commands.

[0154] S430: Performs hardware changes and collects new thermal field data.

[0155] Specifically, the actuator receives instructions and adjusts the hardware state. After the hardware change is complete, the system needs to run in a stable state for a period of time and collect new thermal field data. Due to the hardware change, the system's dynamic thermal inertia model parameters... , , The situation has changed, so thermal field data is needed to capture this change.

[0156] In one possible implementation, the new thermal field data and the control input data during hardware changes are fed back to the system identification process of S100 as a new dataset to update the dynamic thermal inertia model.

[0157] Based on the updated dynamic thermal inertia model, the S200 to S300 processes are re-executed to generate optimized compensation strategies that match the new hardware state, and their effectiveness is verified. Since the hardware bottleneck has been addressed, the newly generated compensation strategies are generally less stringent and more effective, resulting in improved overall system thermal management performance.

[0158] Thus, this invention completes the entire route from software control to hardware reconstruction and back to software control optimization, realizing true intelligence and adaptability of the thermal management system.

[0159] Example 2:

[0160] A thermal management controller control system based on dynamic thermal inertia prediction includes:

[0161] The data acquisition module is used to collect thermal field data during the operation of the thermal management system; the model building module is connected to the data acquisition module and is used to build a dynamic thermal inertia model based on the thermal field data.

[0162] The prediction and compensation strategy generation module communicates with the model building module and is used to predict the future temperature field evolution and thermal accumulation risk of the system based on the dynamic thermal inertia model, and generate compensation strategies in three dimensions: compensation angle, compensation order and compensation domain.

[0163] The evaluation and optimization module communicates with the prediction and compensation strategy generation module and the data acquisition module. It is used to collect thermal field data after applying the compensation strategy and calculate the artifact intensity entropy value of the thermal field data. If the artifact intensity entropy value does not reach the preset threshold, it is determined that the compensation strategy has not achieved the expected results. The dynamic thermal inertia model is optimized through the outer loop optimization mechanism, which triggers the prediction and compensation strategy generation module to regenerate the compensation strategy and perform iterative evaluation until the artifact intensity entropy value reaches the preset threshold. If the preset threshold is reached, the compensation strategy is determined to be effective and continues to be executed.

[0164] The hardware collaborative control module communicates with the evaluation and optimization module and the prediction and compensation strategy generation module. After adopting an effective compensation strategy, if it is found that a specific area requires a continuous high-intensity similar compensation strategy under various different operating conditions, it is determined that there is a continuous hardware-level thermal imbalance in the specific area, and hardware collaborative control is initiated to optimize the heat dissipation path. At the same time, the thermal field data established after the hardware change is fed back to the model building module to regenerate the compensation strategy and verify it.

[0165] Those skilled in the art will understand that the embodiments of this application are provided as methods, systems, or computer program products. Therefore, this application takes the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application takes the form of a computer program product implemented on one or more computer storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer program code. The solutions in the embodiments of this application are implemented using various computer languages, exemplified by the object-oriented programming language Java and the interpreted scripting language JavaScript.

[0166] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, are implemented by computer program instructions. These computer program instructions are provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0167] These computer program instructions are also stored in a computer read-memory memory (CROM) that can direct a computer or other programmed data processing device to operate in a specific manner, such that the instructions stored in the CROM produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0168] These computer program instructions are also loaded onto a computer or other programmed data processing device, causing a series of operational steps to be performed on the computer or other programmed device to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmed device for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0169] Although preferred embodiments of this application have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments as well as all changes and modifications falling within the scope of this application.

[0170] Obviously, those skilled in the art can make various modifications and variations to this application without departing from the spirit and scope of this application. Therefore, if such modifications and variations fall within the scope of the claims of this application and their equivalents, this application also intends to include such modifications and variations.

Claims

1. A thermal management controller control method based on dynamic thermal inertia prediction, characterized in that, include: Collect thermal field data during the operation of the thermal management system, and construct a dynamic thermal inertia model based on the thermal field data; Based on the dynamic thermal inertia model, the future temperature field evolution and thermal accumulation risk of the system are predicted, and compensation strategies are generated in the three dimensions of compensation angle, compensation order and compensation domain space respectively. The compensation strategy for each dimension includes a three-layer nested strategy of pre-compensation, main compensation, and residual compensation. Among them, the pre-compensation strategy quickly suppresses the initial disturbance, the main compensation strategy offsets the predicted thermal deviation, and the residual compensation strategy addresses the uncertainty of the dynamic thermal inertia model. Collect thermal field data after applying the compensation strategy and calculate the artifact intensity entropy value of the thermal field data. If the artifact intensity entropy value does not reach the preset threshold, it is determined that the compensation strategy has not achieved the expected results. Then, the dynamic thermal inertia model is optimized through the outer loop optimization mechanism, the compensation strategy is regenerated and the artifact intensity entropy value is calculated until the artifact intensity entropy value reaches the preset threshold. If the artifact intensity entropy value reaches the preset threshold, it proves that the compensation strategy is effective and continues to be executed. If, after adopting an effective compensation strategy, it is found that a specific area requires a continuous high-intensity similar compensation strategy under various different operating conditions, it is determined that there is a continuous hardware-level thermal imbalance in the specific area. The controller will initiate hardware collaborative regulation to optimize the heat dissipation path and regenerate and verify the compensation strategy based on the thermal field data established after the hardware change. In the compensation strategy of the compensation stage, the strengths of pre-compensation, main compensation, and residual compensation are determined in the following ways: Pre-compensation compensation stage The pre-compensation is proportional to the norm of the current thermal deviation, and its specific form is as follows: ,in, It is pre-compensation of the compensation order dimension. It is a node thermal deviation, For proportional gain; The compensation order of the main compensation is obtained by solving the finite-time optimal control problem, and the objective function is: ;in, It's thermal deviation. Indicates in Always looking towards the future Control the input vector at all times. The main compensation sequence is obtained by controlling the input weights and solving the objective function. Compensation order of residual compensation It is proportional to the integral of the observation and prediction deviation, i.e. Specific forms of residual compensation ,in At any moment The observed actual thermal deviation, It is a past moment to a moment Predicted thermal deviation, To compensate for the gain of the residual, From the initial moment to the current moment The integral; The calculation of the artifact intensity entropy value includes: For each unit The temperature time series was low-pass filtered to obtain a smooth trajectory; according to Calculate the artifact intensity for each cell; the larger the artifact intensity value, the stronger the unexpected temperature fluctuation at that cell. according to Normalizing artifact intensity yields probability distribution , making : according to Calculate the entropy value of artifact intensity ; in, For the first The artifact intensity of each node, To assess the length of the time window, For the first Each node at time... The actual observed temperature For the first Each node at time... The desired smooth temperature trajectory, For the first The proportion of artifact intensity at each node, The total number of nodes in the system. It is the natural logarithm. For the first The artifact intensity of each node.

2. The thermal management controller control method based on dynamic thermal inertia prediction according to claim 1, characterized in that, The dynamic thermal inertia model is constructed through spatial discretization and time-difference solution, and its state-space equations are as follows: ;in: yes Temperature state vector of all units at any given time; yes The control input vector at time t, yes The control input vector at time t, It is the system state matrix. It is a control input matrix. It is the thermal accumulation effect matrix. It is the length of the time window that takes into account the heat accumulation effect.

3. The thermal management controller control method based on dynamic thermal inertia prediction according to claim 2, characterized in that, The prediction system for the future temperature field evolution is achieved by iteratively applying a dynamic thermal inertia model, specifically: ;in, Indicates in Always looking towards the future The predicted temperature state vector at any given time. Indicates in Always looking towards the future The predicted temperature state vector at any given time. Indicates in Always looking towards the future Control the input vector at all times.

4. The thermal management controller control method based on dynamic thermal inertia prediction according to claim 1, characterized in that, The compensation angle is used to find the most effective direction of compensation. The compensation strategy for the compensation angle dimension is determined by solving a spatial optimization problem. The optimization objective is to minimize the maximum spatial gradient of the temperature field or the thermal stress, i.e.: ;in, For compensation angle, It depends on the angle The compensation operator, It is a spatial gradient operator. It is a node Thermal deviation; After solving the optimization problem using gradient descent, the optimal compensation angle is obtained. Based on this, the spatial weight distribution of pre-compensation, main compensation, and residual compensation is generated.

5. The thermal management controller control method based on dynamic thermal inertia prediction according to claim 1, characterized in that, In the compensation strategy of the compensation domain, the time domains of pre-compensation, main compensation, and residual compensation are defined as follows: Pre-compensation compensation domain For a short time window, that is ,in, It is an integer between 1 and 3. Sampling time; The compensation domain of the main compensation Covering the entire prediction time domain ; Compensation domain of residual compensation It is a continuous time domain that covers the entire control process.

6. The thermal management controller control method based on dynamic thermal inertia prediction according to claim 1, characterized in that, The outer loop optimization mechanism includes using a recursive least squares method with a forgetting factor or a sliding window batch processing algorithm to update the system state matrix, control input matrix, and thermal accumulation effect matrix of the dynamic thermal inertia model online. The optimization objective is to minimize the prediction error of recent data, thereby adaptively correcting the model bias.

7. The thermal management controller control method based on dynamic thermal inertia prediction according to claim 1, characterized in that, The hardware coordinated control includes: Microfluidic valves are controlled to guide the deposition of phase change materials in the thermal accumulation region to increase local heat capacity; The shape memory alloy guide plate is driven to reconstruct the flow channel to optimize the heat dissipation path; After the hardware change, thermal field data is re-acquired and the dynamic thermal inertia model is updated to generate a compensation strategy that matches the new hardware state.

8. A system for a thermal management controller control method based on dynamic thermal inertia prediction as described in any one of claims 1-7, characterized in that, The system includes: The data acquisition module is used to collect thermal field data during the operation of the thermal management system; The model building module communicates with the data acquisition module and is used to build a dynamic thermal inertia model based on thermal field data. The prediction and compensation strategy generation module communicates with the model building module and is used to predict the future temperature field evolution and thermal accumulation risk of the system based on the dynamic thermal inertia model, and generate compensation strategies in three dimensions: compensation angle, compensation order and compensation domain. The evaluation and optimization module communicates with the prediction and compensation strategy generation module and the data acquisition module. It is used to collect thermal field data after applying the compensation strategy and calculate the artifact intensity entropy value of the thermal field data. If the artifact intensity entropy value does not reach the preset threshold, it is determined that the compensation strategy has not achieved the expected results. The dynamic thermal inertia model is optimized through the outer loop optimization mechanism, which triggers the prediction and compensation strategy generation module to regenerate the compensation strategy and perform iterative evaluation until the artifact intensity entropy value reaches the preset threshold. If the preset threshold is reached, the compensation strategy is determined to be effective and continues to be executed. The hardware collaborative control module communicates with the evaluation and optimization module and the prediction and compensation strategy generation module. After adopting an effective compensation strategy, if it is found that a specific area requires a continuous high-intensity similar compensation strategy under various different operating conditions, it is determined that there is a continuous hardware-level thermal imbalance in the specific area, and hardware collaborative control is initiated to optimize the heat dissipation path. At the same time, the thermal field data established after the hardware change is fed back to the model building module to regenerate the compensation strategy and verify it.

Images