A temperature control device and method for a tablet computer

Abstract

The present invention provides a temperature control device and method for a tablet computer, belonging to the technical field of temperature control. The device fuses the collected perception data into a high-dimensional feature vector, inputs the high-dimensional feature vector and the posture of the tablet computer into a thermal field prediction model using a convolutional long short-term memory network to predict the temperature data of each area of the tablet computer, and outputs a predicted thermal field distribution heat map; according to the temperature distribution in the thermal field distribution heat map and in combination with multiple target parameters, a target function is defined, and the target function dynamically allocates the weight of each target parameter according to the usage environment of the tablet computer; a multi-objective optimization algorithm is used to generate a Pareto optimal solution set, and a fuzzy logic decision is used to select the optimal temperature control strategy suitable for the current application scenario from multiple candidate temperature control strategies, and the hardware parameters of the tablet computer are adjusted according to the optimal temperature control strategy. Its accurate thermal field prediction reduces temperature fluctuations and improves the stability of the device under high load conditions.

Description

Technical Field

[0001] The present invention relates to the technical field of temperature control for intelligent terminal devices, and particularly to a temperature control device and method for a tablet computer. Background Art

[0002] There are many technical bottlenecks in the existing tablet computer temperature control systems. Traditional temperature control strategies usually rely on fixed thresholds (for example, when the chip temperature reaches 80 °C, frequency reduction is triggered), which leads to large performance fluctuations and response lags. In addition, traditional solutions usually only collect single-point temperature data and cannot accurately predict the local heat diffusion path, easily resulting in false triggering or temperature control failure. More complicatedly, it is difficult to dynamically adjust the balance among heat dissipation, performance, noise, and energy consumption. For example, the high-speed operation of the fan can effectively cool down, but it increases power consumption and noise. In addition, existing solutions lack adaptation to the user application scenarios and fail to automatically adjust the temperature control strategy for different applications (such as games, video playback, reading).

[0003] In the existing technologies, such as the temperature control method based on the PID algorithm (P represents Proportional, I represents Integral, D represents Differential) and the method of predicting temperature by neural network, although there are breakthroughs, they still fail to solve the problem of multi-variable coupling, and the computational complexity is relatively high, making it difficult to run in real time. Summary of the Invention

[0004] In view of this, the purpose of the embodiments of the present invention is to provide a temperature control device and method for a tablet computer to solve at least one of the above technical problems.

[0005] To achieve the above object, in a first aspect, a temperature control device for a tablet computer is provided. The temperature control device includes:

[0006] A sensing data acquisition unit for acquiring sensing data of the tablet computer;

[0007] A data preprocessing and fusion unit for preprocessing the acquired sensing data and fusing the preprocessed sensing data into a high-dimensional feature vector;

[0008] A thermal field prediction unit for inputting the high-dimensional feature vector and the posture of the tablet computer into a thermal field prediction model using a convolutional long short-term memory network, and predicting the temperature data of each region of the tablet computer through the thermal field prediction model and generating a thermal field distribution heat map;

[0009] An objective function definition unit for defining an objective function according to the temperature distribution in the thermal field distribution heat map and combining multiple objective parameters. The objective function dynamically assigns weight coefficients to each objective parameter according to the usage environment of the tablet computer.

[0010] A temperature control decision-making engine, which is used to generate a Pareto optimal solution set by using a multi-objective optimization algorithm according to the objective function and the weight coefficient. The Pareto optimal solution set contains multiple candidate temperature control strategies; select the optimal temperature control strategy suitable for the current application scenario from the multiple candidate temperature control strategies through fuzzy logic decision-making, and transmit the optimal temperature control strategy to the execution unit;

[0011] An execution unit, which is used to adjust the hardware parameters of the tablet computer according to the optimal temperature control strategy.

[0012] In a second aspect, an embodiment of the present invention provides a temperature control method for a tablet computer. The temperature control method includes:

[0013] Collect the sensing data of the tablet computer;

[0014] Preprocess the collected sensing data, and fuse the preprocessed sensing data into a high-dimensional feature vector;

[0015] Input the high-dimensional feature vector and the posture of the tablet computer into a thermal field prediction model using a convolutional long short-term memory network, and predict the temperature data of each area of the tablet computer through the thermal field prediction model and generate a thermal field distribution heat map;

[0016] Define an objective function according to the temperature distribution in the thermal field distribution heat map and in combination with multiple target parameters. The objective function dynamically assigns the weight coefficient of each target parameter according to the usage environment of the tablet computer;

[0017] Generate a Pareto optimal solution set by using a multi-objective optimization algorithm according to the objective function and the weight coefficient. The Pareto optimal solution set contains multiple candidate temperature control strategies; select the optimal temperature control strategy suitable for the current application scenario from the multiple candidate temperature control strategies through fuzzy logic decision-making, and transmit the optimal temperature control strategy to the execution unit;

[0018] Adjust the hardware parameters of the tablet computer according to the optimal temperature control strategy.

[0019] The above technical solution has the following beneficial effects:

[0020] Through accurate thermal field prediction, the present invention reduces temperature fluctuations, thereby improving the stability of the tablet computer under high load conditions. In high-performance application scenarios, it can reduce the frequency of CPU downclocking, enhance the stability of the frame rate, and improve the temperature control accuracy and overall performance.

[0021] By dynamically adjusting the weights, the present invention achieves a lower overall power consumption and is more efficient in power consumption control compared to traditional temperature control solutions. Moreover, the processing delay of the algorithm is effectively controlled, ensuring the smoothness of the user experience without being affected by the delay of temperature control decisions. Through a complete strategy library, it can automatically adapt to different application scenarios, provide broader support, and ensure the best temperature control effect in various environments.

[0022] The present invention adopts a "prediction, decision-making, execution" closed-loop temperature control architecture, avoiding the limitations of traditional threshold control. It uses a lightweight Conv-LSTM thermal field prediction model to achieve high-precision thermal diffusion modeling on mobile devices for the first time. In addition, the multi-objective Pareto optimal real-time solution algorithm of the present invention effectively solves the multi-variable coupling problems such as temperature, performance, and noise, and realizes scenario-adaptive personalized temperature control through a software-defined mechanism. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] Figure 1 It is a functional block diagram of a temperature control device for a tablet computer according to an embodiment of the present invention;

[0024] Figure 2 It is a functional block diagram of a data preprocessing and fusion unit according to an embodiment of the present invention;

[0025] Figure 3 It is a functional block diagram of a thermal field prediction unit according to an embodiment of the present invention;

[0026] Figure 4 It is a functional block diagram of a sensing data acquisition unit according to an embodiment of the present invention;

[0027] Figure 5 It is a functional block diagram of another temperature control device for a tablet computer according to an embodiment of the present invention;

[0028] Figure 6 It is a functional block diagram of a scenario recognition and adaptive strategy switching unit according to an embodiment of the present invention;

[0029] Figure 7 It is a functional block diagram of a real-time inference and control execution unit according to an embodiment of the present invention;

[0030] Figure 8 It is a functional block diagram of an anomaly diagnosis and model fine-tuning unit according to an embodiment of the present invention;

[0031] Figure 9 It is a functional block diagram of a temperature control execution and feedback adjustment unit according to an embodiment of the present invention;

[0032] Figure 10 It is a flowchart of a temperature control method for a tablet computer according to an embodiment of the present invention;

[0033] Figure 11 is a functional block diagram of a computer-readable storage medium according to an embodiment of the present invention;

[0034] Figure 12 is a functional block diagram of an electronic device according to an embodiment of the present invention. Detailed implementation manners

[0035] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0036] The purpose of the embodiments of the present invention is to solve the defects in the prior art through a software-defined temperature control method based on dynamic thermal field modeling and multi-objective optimization. The specific objectives include: accurately identifying the temperature distribution trend through a real-time thermal field prediction algorithm; designing a lightweight multi-objective decision-making model to dynamically balance temperature, performance, power consumption, and user experience; constructing an adaptive control strategy library based on the user application scenario to achieve personalized temperature control.

[0037] Embodiment 1

[0038] As Figure 1 shown, the embodiments of the present invention provide a temperature control device for a tablet computer. The temperature control device includes:

[0039] A sensing data acquisition unit for acquiring sensing data of the tablet computer. The sensing data includes temperature data of multiple regions of the tablet computer, temperature and humidity data of the surrounding environment, and power consumption during the operation of the tablet computer.

[0040] Specifically, in the tablet computer, temperature data is acquired through a high-precision thermocouple array sensing system. The thermocouple array is distributed in the CPU (Central Processing Unit), GPU (Graphics Processing Unit), battery, screen driver chip, and the holding area of the housing to acquire temperature data of each key part. The temperature data acquisition can be performed in real time at a frequency of 10 Hz to ensure accurate thermal information is obtained. At the same time, the ambient temperature and humidity sensor will synchronously record the temperature and humidity data of the surrounding environment to provide necessary environmental parameters for subsequent temperature prediction and temperature control decisions. In addition, the power consumption monitoring module will also monitor the power consumption of the tablet computer in real time to facilitate the evaluation of the load situation of the tablet computer.

[0041] The data preprocessing and fusion unit is used to preprocess the collected perception data and fuse the preprocessed perception data into a high-dimensional feature vector.

[0042] In this embodiment, based on the temperature data of multiple regions of the tablet computer, the humidity data and temperature data of the surrounding environment, and the power consumption during the operation of the tablet computer collected, these data will be preprocessed. The main operations of the preprocessing include denoising, data synchronization, and time alignment. After the data preprocessing, all the perception data will be fused into a high-dimensional feature vector for use by the subsequent thermal field prediction and decision-making engine. The information after data fusion can not only provide the current temperature distribution of the tablet computer, but also comprehensively reflect the overall thermal state of the tablet computer by combining environmental factors and the current power consumption of the tablet computer.

[0043] As Figure 2 shown, in some embodiments, the data preprocessing and fusion unit specifically includes: a data preprocessing subunit, a data fusion subunit, and a data storage and caching subunit.

[0044] The data preprocessing subunit is used to perform denoising, data synchronization, time alignment, normalization, and standardization processing on the collected perception data to obtain the preprocessed perception data.

[0045] Specifically, first, the temperature data of multiple regions of the tablet computer, the temperature data and humidity data of the surrounding environment, and the power consumption data during the operation of the tablet computer collected will be affected by factors such as sensor accuracy, environmental interference, or data transmission, resulting in the appearance of noise. To ensure the accuracy of the data, these data need to be denoised. The denoising methods include: the moving average method, which performs a sliding window average on each group of data to smooth out short-term fluctuations; the median filtering method, which removes outliers or extreme data by taking the median value in the data window; the Kalman filtering method, which performs recursive processing on the data and suppresses noise by combining the prediction model and the actual observation data. Through the denoising process, the noise interference in the data is reduced, making the data cleaner and more accurate, capable of more truly reflecting the actual situation of the monitored object, and improving the quality and reliability of the data. For example, high-frequency noise introduced due to electromagnetic interference and other factors during sensor data collection is removed, making the data curve smoother.

[0046] Secondly, during the data acquisition process, information such as temperature, humidity, and power consumption from different sensors will have different sampling time points. To ensure that all data can be compared and fused at the same moment, data synchronization processing is required. Synchronization methods include: interpolation method and resampling method. The interpolation method refers to interpolating data with different sampling times and unifying them to a common time reference. For the missing data between time points, linear interpolation or spline interpolation can be used for estimation; the resampling method refers to adjusting all data to the same time interval. For example, the data of sensors with a low sampling frequency are resampled to 10Hz to be consistent with other high-frequency data. Through data synchronization, the data of all sensors can be processed within the same time window, ensuring the accuracy of subsequent analysis.

[0047] Furthermore, in addition to synchronizing data from different sources, time alignment also needs to ensure that the time order between data points is exactly matched. The time stamps of data are inconsistent due to processing delays or transmission delays of different sensors. Specific practices include: timestamp matching, attaching accurate timestamps to each data point and sorting the data by timestamps; filling operation, if data at certain moments is missing, interpolation or estimation methods can be used to fill in the missing values so that each time point has complete temperature, humidity, and power consumption data. Through data synchronization and time alignment, the data from different sources and with different sampling frequencies are made consistent and comparable in time. This enables the subsequent fusion and analysis of multi-source data to be carried out within an accurate time framework, avoiding incorrect analysis results caused by time differences. For example, for data from multiple sensors, after processing, it can be ensured that their values at the same time point can correspond, facilitating the comprehensive analysis of the overall state of the device at a certain moment.

[0048] Finally, the data is standardized and normalized. To eliminate the dimensional differences of different data sources, data such as temperature, humidity, and power consumption are standardized or normalized. For example, temperature is in degrees Celsius (°C), power consumption is in watts (W), and humidity is expressed as a percentage (%). After being standardized and normalized, these data can be unified to the same dimension, facilitating subsequent analysis and processing. The processing methods include: Standardization, which subtracts the mean of each feature and divides it by the standard deviation, making the data have zero mean and unit variance; Normalization, which scales the data between 0 and 1, using the min-max normalization method. Through standardization and normalization processing, the dimensional differences between different feature data are eliminated, making the data have the same weight and influence in subsequent analysis and modeling, helping to improve the accuracy and stability of the algorithm, and also facilitating data visualization and comparison. For example, after standardizing or normalizing data of different physical quantities such as temperature, voltage, and current, they can be analyzed and processed on the same scale, and no single feature with a large numerical range will dominate the entire analysis process.

[0049] The data fusion sub-unit is used to fuse the preprocessed perception data to generate a high-dimensional feature vector.

[0050] In this embodiment, after all the preprocessing steps are completed, next, the data from each sensor will be fused to generate a high-dimensional feature vector. The purpose of data fusion is to integrate information from different sources (such as temperature, humidity, power consumption, etc.) into a unified representation for subsequent use by the thermal field prediction and temperature control decision engine. The specific method is as follows: Feature concatenation, which combines data of each dimension such as temperature, humidity, and power consumption through concatenation to form a high-dimensional feature vector. The dimension of each high-dimensional feature vector corresponds to a specific sensor data; Weighted fusion, which assigns weights to each feature data according to different importance levels, and fuses different data into a unified feature vector through weighted summation. For example, different weights can be assigned to temperature and power consumption data according to the accuracy and importance of different sensors.

[0051] The fused high-dimensional feature vector not only contains the temperature distribution information of each area of the tablet computer, but also combines the environmental temperature and humidity and the power consumption data during the operation of the tablet computer, and can comprehensively reflect the overall thermal state of the tablet computer, providing effective input for subsequent thermal field prediction and decision engine.

[0052] The data storage and caching sub-unit is used to store the preprocessed and fused data in a temporary cache or database for subsequent use by the analysis, prediction, and decision engine. This storage process ensures the integrity and accessibility of the data, allowing the subsequent system to quickly access historical data and perform calculations.

[0053] In this embodiment, by preprocessing and fusing the perception data, multi-dimensional data such as temperature data in multiple regions of the tablet computer, temperature and humidity data of the surrounding environment, and power consumption data are denoised, data synchronized, time-aligned, standardized, etc., and finally fused into a high-dimensional feature vector. This process ensures that the subsequent thermal field prediction and temperature control decision-making engine can make temperature control decisions based on accurate and complete input data, thereby effectively improving the temperature control accuracy and performance of the tablet computer.

[0054] A thermal field prediction unit, configured to input the high-dimensional feature vector and the posture of the tablet computer into a thermal field prediction model using a convolutional long short-term memory network, predict the temperature data of each region of the tablet computer, and output a predicted thermal field distribution heat map.

[0055] The thermal field prediction model of this embodiment uses a lightweight convolutional long short-term memory (Convolutional Long Short-Term Memory, abbreviated as Conv-LSTM) model for thermal field prediction. This model takes the temperature sensor data collected in real time, the power consumption of the tablet computer, the ambient temperature and humidity, and the posture of the tablet computer (for example, the posture information of the tablet computer obtained through a gyroscope sensor, which is used to reflect the physical state of the tablet computer in different application scenarios, such as the holding angle, etc.) as inputs to predict the heat diffusion process and can perform real-time calculations on the tablet computer with low latency. The thermal field prediction model can effectively capture the spatial and temporal dependencies, predict the temperature changes in each region of the tablet computer, and then output a predicted thermal field distribution heat map.

[0056] As Figure 3 shown, in some embodiments, the thermal field prediction unit specifically includes: a data input subunit, a first processing subunit, a second processing subunit, an output subunit, a lightweight optimization subunit, and a model evaluation and verification subunit.

[0057] The data input subunit is configured to use the temperature data collected in real time, the power consumption of the tablet computer, the ambient temperature and humidity data, and the posture data of the tablet computer as inputs to the Conv-LSTM model. Before these input data enter the Conv-LSTM model, they need to undergo preprocessing and fusion in the previous steps to ensure that the data formats are consistent and meet the requirements of the model.

[0058] The first processing subunit is used to process the input data through convolution operations to extract the spatial features of the temperature distribution in each region. The first part of the Conv-LSTM model is the convolutional layer, and the role of the convolutional layer is to extract the local features of the temperature distribution from the spatial dimension. This convolutional layer processes the input data through convolution operations to extract the spatial features of the temperature distribution in each region. For example, for temperature sensor data, the convolutional layer will identify and capture the spatial structure of temperature changes, such as the heat distribution differences in the CPU region and the battery region.

[0059] The convolution operation formula is: ； where y is the output of the convolutional layer, W is the convolutional kernel weight, x is the input data, represents the convolution operation, and b is the bias term. Through convolution processing, spatial feature maps can be obtained, and these feature maps will provide effective information for the next LSTM processing.

[0060] The second processing subunit uses the time series information of historical temperature data, the power consumption of the tablet computer, and environmental humidity data to predict the temperature changes in each region of the tablet computer within a preset future time. The second part of the Conv-LSTM is the LSTM layer, and the LSTM layer can effectively capture the dependencies in the time series. The role of the LSTM layer is to use time series information such as past temperature data, power consumption data, and environmental humidity data to predict the temperature changes in each region within a preset future time (such as 30 seconds). LSTM can remember long-term historical information and adjust the weights according to the changes in the time series, so as to accurately predict the future thermal field changes.

[0061] The output subunit is used to output the predicted thermal field distribution heat map. After being processed by convolution and LSTM, the Conv-LSTM model will generate a predicted thermal field distribution heat map. This heat map shows the temperature change trend in each region of the tablet computer within the next 30 seconds, and specifically reflects the temperature distribution in different regions (such as the CPU, GPU, screen, shell, etc.). The thermal field distribution heat map can represent different temperature ranges with different colors, which is convenient for the subsequent decision-making engine to identify which regions have higher temperatures and which regions need to strengthen heat dissipation.

[0062] The lightweight optimization subunit is used to perform lightweight processing on the convolutional long short-term memory network.

[0063] Specifically, to enable the Conv-LSTM model to run efficiently on a tablet computer and ensure the inference speed of the model, the model parameters need to be lightweight processed. This includes parameter compression, quantization, and knowledge distillation. Parameter compression refers to reducing the number of model parameters through pruning techniques and removing unimportant parameters; quantization refers to converting floating-point parameters to low-precision representations (e.g., converting 32-bit floating-point numbers to 8-bit integers) to reduce storage and computational overhead; knowledge distillation refers to training a small model to imitate the behavior of a large model to obtain a smaller and more efficient model. Through these optimizations, the size of the Conv-LSTM model will be controlled within 500 KB and can run on the NPU (Neural Processing Unit) of the tablet computer with a latency of less than 5 ms, ensuring real-time performance.

[0064] The model evaluation and verification subunit is used to evaluate and verify the convolutional long short-term memory network.

[0065] In practical applications, to ensure the effectiveness and accuracy of the thermal field prediction model, the Conv-LSTM model is evaluated and verified. This process includes cross-validation and accuracy evaluation. Cross-validation refers to dividing the data into multiple subsets and training and validating the thermal field prediction model multiple times to ensure the generalization ability of the model. Accuracy evaluation refers to evaluating the prediction accuracy of the model by comparing the predicted thermal field distribution with the actually measured temperature data.

[0066] Through the above processing, the Conv-LSTM model in this embodiment can, within a 5-second time window, predict the temperature changes in each area of the tablet computer in the next 30 seconds based on the real-time collected temperature, power consumption, ambient temperature and humidity, and posture data, and generate a thermal field distribution heat map. This model can not only simulate the heat diffusion process with high precision but also run in real time on the tablet computer. The embodiment of the present invention realizes high-precision thermal field modeling on the mobile side and avoids the limitations of traditional methods.

[0067] The objective function definition unit is used to define an objective function according to the temperature distribution in the thermal field distribution heat map and in combination with multiple objective parameters. The objective function dynamically assigns weights to each objective parameter according to the usage environment of the tablet computer; the objective parameters include but are not limited to fan power consumption, CPU downclocking amplitude, and noise generated during the operation of the tablet computer.

[0068] In this embodiment, after obtaining the thermal field prediction result, the system enters the multi-objective optimization decision-making stage. At this time, the system will consider multiple objective parameters, each of which is closely related to the operating conditions of the tablet computer, such as temperature, fan power consumption, CPU downclocking amplitude, sound interference generated during the operation of the tablet computer, etc. For example, reducing the temperature requires higher fan power consumption, while increasing the fan power consumption will lead to increased noise. Therefore, it is necessary to balance these factors by optimizing the objective function. The objective function can be expressed as: ;

[0069] where, is the predicted maximum temperature (°C), that is, the highest temperature reached by the tablet computer in the future period; is the fan power consumption, indicating the power consumed by the fan operation; is the CPU downclocking amplitude (Hz), that is, the new frequency to which the CPU is reduced from the original frequency; Noise is the noise level (dB), indicating the noise generated by the tablet computer in the current operating state. These weight coefficients are dynamically adjusted according to different application scenarios, so as to optimize the temperature control strategy and balance the conflicts between different objectives.

[0070] In the above objective function, the weight coefficients α, β, γ, δ are used to adjust the relative importance of each objective, ensuring that the system can make appropriate optimization decisions according to different application scenarios and requirements. By adjusting the weight coefficients, the performance and temperature can be balanced in different working modes (such as high-performance mode, balanced mode, silent mode) to achieve the optimal temperature control strategy.

[0071] In this embodiment, the priority of temperature control is different under different application scenarios. For example, when performing high-intensity calculations (such as games or video rendering), more attention is paid to the balance between performance and temperature, while in a quiet environment (such as reading mode), more attention is paid to reducing noise and power consumption. Therefore, the weight coefficients (α, β, γ, δ) need to be dynamically adjusted according to the application scenario.

[0072] In high-performance mode, temperature control is more relaxed, but performance requirements are higher. Therefore, the weighting coefficient α (temperature) is lower, while β (fan power consumption) and γ (CPU frequency reduction) are higher to maximize performance. In this embodiment, high-performance mode refers to scenarios that require high tablet performance, such as 3D gaming and video rendering. In this mode, the tablet implements a series of optimization measures to maximize performance, such as 3D gaming and video rendering. For example, the tablet prioritizes more hardware resources, such as the CPU and GPU computing power, to running high-performance applications. For example, when running a 3D game, the system will fully utilize the CPU and GPU to provide smoother gameplay and higher frame rates. Furthermore, to meet high-performance requirements, the device allows for short-term overclocking of hardware such as the CPU. This allows the CPU to run at a speed higher than its default frequency, thereby improving data processing capabilities and enabling applications to more quickly complete complex computing tasks, such as special effects processing in video rendering and scene modeling and rendering in 3D games. Although temperature control is relatively relaxed in high-performance mode, heat dissipation measures are strengthened to ensure device stability under high load. For example, the fan will run at a higher speed to dissipate the heat generated by hardware such as the CPU and GPU as quickly as possible to prevent device performance degradation or failure due to excessive temperature.

[0073] In balanced mode (multitasking, 5G download), a balance needs to be maintained between performance and temperature. The weight coefficients α, β, and γ need to be moderate to ensure a balance between temperature and performance while reducing fan power consumption and noise.

[0074] In silent mode (reading, standby), priority is given to reducing noise and power consumption, the weight coefficient δ (noise) will be higher, and α (temperature) and β (fan power consumption) will be lower.

[0075] The weight coefficients of this embodiment can be dynamically adjusted according to the actual scenario, thereby optimizing the balance of various objectives.

[0076] In an alternative embodiment, the objective function is: ;

[0077] in, is the predicted maximum temperature, which represents the maximum temperature reached in each area of the tablet. is the fan power consumption, which indicates the power consumption required for system cooling. It is the CPU frequency reduction, which indicates the degree to which the CPU frequency is reduced to reduce heat generation. Noise level indicates the noise generated during the operation of the tablet computer. BE (Battery Energy) represents the remaining battery capacity of the tablet computer. The goal is to maximize the battery energy, so a minus sign is used in front to indicate that this goal needs to be maximized. TG (Thermal Gradient) is the thermal gradient, which represents the temperature difference between different areas of the tablet computer. The goal is to minimize the thermal gradient to ensure a uniform temperature distribution on the tablet computer and avoid local overheating. MU (Memory Usage) is the memory usage, which represents the amount of memory used during the operation of the tablet computer. The goal is to minimize the memory usage.

[0078] In the embodiment of the present invention, by dynamically adjusting the weights and adjusting the temperature control strategy in real time according to different application scenarios, the problem of multi-variable coupling in temperature control is solved.

[0079] The temperature control decision-making engine is used to generate a Pareto optimal solution set by using a multi-objective optimization algorithm according to the defined objective function and the dynamically adjusted weight coefficients. The Pareto optimal solution set contains multiple candidate temperature control strategies; the optimal temperature control strategy suitable for the current application scenario is selected from multiple candidate temperature control strategies through fuzzy logic decision-making, and the optimal temperature control strategy is transmitted to the execution unit.

[0080] To obtain the best temperature control strategy, according to the defined objective function and the dynamically adjusted weight coefficients, the system will solve a multi-objective optimization problem. For this purpose, the system uses an improved NSGA-II (Non dominated sorting genetic algorithm -II, the second generation non-dominated sorting genetic algorithm) multi-objective genetic algorithm for solving. Within a calculation time of 10 ms, the algorithm will generate a Pareto optimal solution set, listing multiple control strategies that meet various constraint conditions. Among these strategies, through further fuzzy logic analysis, the final temperature control strategy most suitable for the current application scenario and requirements is selected. Through this multi-objective optimization, it is possible to ensure that while maintaining temperature control, the performance is maximized, the power consumption is reduced, and the noise is reduced. Among them, using the improved NSGA-II (the second generation non-dominated sorting genetic algorithm) multi-objective genetic algorithm for solving specifically includes the following steps:

[0081] Step 5.1: Initialize the population. In the NSGA-II multi-objective genetic algorithm, it is first necessary to initialize the population. Initializing the population means generating multiple candidate solutions, and each candidate solution represents a temperature control strategy. Each solution consists of a set of genes, and the genes represent the various parameters in the control strategy, such as fan speed, CPU frequency, voltage, etc. In the initialization stage, the solutions in the population can be obtained by random generation to ensure the diversity of the population. The initialized population size is set to a certain number, such as 100 candidate solutions.

[0082] Step 5.2: Calculate fitness and evaluate the objective function. After generating the initial population, the system needs to evaluate the fitness of each candidate solution. The calculation of fitness depends on the previously defined objective function, which comprehensively considers multiple factors such as temperature, fan power consumption, CPU frequency reduction amplitude, and noise. For each candidate solution, its fitness value is evaluated according to this objective function. A solution with a lower fitness value indicates that the solution performs better in multi-objective optimization and can better balance temperature, performance, power consumption, and noise. Among them, the formula for calculating the fitness value is closely related to the objective function. According to the definition of the objective function, the calculation of fitness can be expressed by the following formula: ;

[0083] The lower the fitness value, the better the balance performance of the candidate solution in multiple objectives such as temperature control, power consumption management, performance maintenance, and noise control. Therefore, this solution is more optimal. In multi-objective optimization, the objective function is minimized, that is, by adjusting the weight coefficients and solution parameters, the fitness value is minimized to obtain an ideal temperature control strategy. In this formula, the adjustment of the weight coefficients (α, β, γ, δ) will determine the relative importance between different objectives. According to the requirements of different scenarios, the fitness value reflects the comprehensive performance of the temperature control strategy in meeting different objectives (reducing temperature, reducing noise, reducing power consumption).

[0084] Step 5.3: Non-dominated sorting. The NSGA-II algorithm sorts the individuals in the population through non-dominated sorting. The purpose of non-dominated sorting is to find the dominance relationship according to the performance of individuals in the multi-objective space. An individual A dominates individual B if and only if individual A is not inferior to individual B in all objectives and is superior to individual B in at least one objective. After non-dominated sorting, the individuals in the population are divided into multiple levels (i.e., Pareto fronts). The first level contains the optimal solutions that dominate all other individuals, the second level contains those solutions that are dominated by the individuals in the first level but are no longer dominated by other solutions, and so on. The sorted population provides a basis for the subsequent operations of the algorithm for priority selection.

[0085] Step 5.4: Crowding degree calculation and selection. After completing non-dominated sorting, the NSGA-II algorithm performs crowding degree calculation, which measures the distribution of each solution on the Pareto front. The crowding degree reflects the diversity between solutions. The lower the crowding degree of a solution, the sparser the area where the solution is located and the greater the diversity. Through crowding degree calculation, the system can preferentially select those individuals that are far from other solutions on the Pareto front, which can maintain the diversity of the population and avoid early convergence to a certain local solution. Solutions with a higher crowding degree will be eliminated, while those solutions that occupy a larger area in the multi-objective space will be retained.

[0086] Step 5.5: Crossover and Mutation Operations. After selection, the NSGA-II algorithm will use crossover and mutation operations to generate new candidate solutions. The crossover operation refers to selecting two individuals for crossover (i.e., exchanging some genes) to generate new individuals. The crossover operation helps explore new solution spaces and produce more potential solutions. The mutation operation refers to randomly changing the value of a certain gene to generate a new individual. The mutation operation is beneficial for introducing new features, increasing the diversity of the solution space, and preventing the algorithm from falling into local optima.

[0087] The new solutions after crossover and mutation operations will be merged with the current population, and non-dominated sorting and crowding degree calculation will be performed again to select a new parental population.

[0088] Step 5.6: Generating the Pareto Optimal Solution Set. Through a series of selection, crossover, mutation, and evaluation, the NSGA-II algorithm will finally generate a Pareto optimal solution set. This solution set contains multiple candidate temperature control strategies. Each temperature control strategy makes a balance among different objectives, and no other strategy can simultaneously outperform them in all objectives. These optimal solutions will represent different temperature control strategies and can be used for subsequent decision-making processes.

[0089] Step 5.7: Selecting the Final Strategy by Fuzzy Logic. After generating multiple Pareto optimal solutions, fuzzy logic decision-making is used to select the temperature control strategy most suitable for the current application scenario. Fuzzy logic allows for a smooth transition between multiple optimization solutions according to preset rules and the current usage status (such as user behavior, workload, environmental conditions, etc.) to select the optimal solution. For example, Rule 1, if the current temperature of the tablet is high, preferentially select the strategy with a lower predicted temperature; Rule 2, if the current mode requires low noise, preferentially select the strategy with lower noise; Rule 3, if the battery power is low, preferentially select the strategy with lower power consumption. Through fuzzy logic rules, the system can select the temperature control strategy that best meets the current requirements from multiple candidate solutions.

[0090] Step 5.8: Execution of the Final Temperature Control Strategy. The final temperature control strategy selected by fuzzy logic will be passed to the execution module to adjust the hardware parameters of the tablet, such as fan speed, CPU frequency, voltage, etc., to ensure that the tablet maintains a balance among temperature, power consumption, and noise while meeting user requirements.

[0091] In the embodiments of the present invention, the objective function comprehensively considers multiple factors such as temperature, fan power consumption, CPU frequency reduction amplitude, and noise, and dynamically adjusts the weight coefficients according to different application scenarios. By solving the multi-objective optimization problem, a set of Pareto optimal solutions is generated, and then the optimal temperature control strategy is selected through fuzzy logic, and finally the optimal temperature control decision is realized to balance the conflicts between various objectives and ensure the temperature, performance, and user experience of the tablet computer. By using the Pareto optimal solution set generated by the multi-objective genetic algorithm, the system can select the best solution from multiple temperature control strategies. The application of the decision engine and fuzzy logic ensures real-time performance and accuracy, and at the same time enables the system to select the most suitable strategy according to different requirements in a dynamic environment.

[0092] An execution unit is configured to adjust the hardware parameters of the tablet computer according to the optimal temperature control strategy.

[0093] The temperature control strategy selected through fuzzy logic will be transmitted to the execution module for actual temperature control operations. The execution module will adjust hardware parameters such as fan speed, CPU frequency, and voltage according to the temperature control strategy to ensure that the temperature, noise, and power consumption of the tablet computer are maintained within a suitable range while ensuring performance.

[0094] As Figure 4 shown, in some embodiments, the sensing data acquisition unit specifically includes: a thermocouple array, an ambient temperature and humidity data acquisition sub-unit, a current and voltage monitoring sub-unit, and a data transmission and integration sub-unit.

[0095] The thermocouple array is configured to collect temperature data of multiple regions of the tablet computer in real time, and the multiple regions include any of the processor region, the battery region, the screen driver chip region, and the outer shell grip region.

[0096] First, high-precision thermocouple arrays are arranged at multiple key parts of the tablet computer. These arrays include multiple thermocouple sensors distributed at the following positions: the CPU / GPU region for real-time monitoring of the temperature changes of the core processor and the graphics chip; the battery region to ensure that the battery operates within a safe temperature range; the screen driver chip region to ensure the stable operation of the display system; and the outer shell grip region to sense the temperature of the user's hand contact position and prevent the user experience from being affected by overheating of the fuselage.

[0097] These thermocouple arrays will collect the temperature information of each part in real time to ensure coverage of all key temperature-sensitive regions of the tablet computer.

[0098] In addition, in order to ensure the real-time performance and accuracy of temperature data in this embodiment, the acquisition frequency of temperature data is set to 10 Hz, that is, the data is acquired 10 times per second. This frequency can track temperature changes in sufficient detail without imposing an excessive burden on the system performance. The data of each sensor will be recorded in chronological order to ensure that subsequent analysis and model prediction can be processed based on high-precision time data.

[0099] The ambient temperature and humidity data acquisition subunit is used to synchronously acquire the temperature and humidity data of the surrounding environment of the tablet computer; in this embodiment, in addition to the temperature data of the internal hardware, the changes in ambient temperature and humidity will also affect the temperature of the tablet computer. Therefore, the tablet computer is also equipped with ambient temperature and humidity sensors to synchronously record the temperature and humidity data of the surrounding environment. These sensors can detect the changes in the temperature and humidity of the surrounding environment of the tablet computer and transmit the data to the temperature control system in real time, providing environmental parameters for subsequent temperature prediction and temperature control decision-making. The acquisition frequency of ambient temperature and humidity data also needs to be synchronized with the temperature data to accurately reflect the impact of the environment on the tablet computer temperature control system.

[0100] The power consumption monitoring subunit is used to monitor the power consumption of the tablet computer in real time. In order to comprehensively evaluate the thermal state of the tablet computer, this embodiment also needs to acquire the power consumption data during the operation of the tablet computer. For this purpose, a power consumption monitoring subunit is configured in the tablet computer to monitor the battery voltage, current and the power consumption of other hardware in real time. The power consumption monitoring subunit will record the power consumption data of the CPU, GPU, display screen and other main hardware components and feedback them to the temperature control device in real time. The real-time acquisition of power consumption data is beneficial to evaluating the load condition of the tablet computer, because in the high-load state, the tablet computer will generate more heat and require more heat dissipation measures to maintain the normal operating temperature.

[0101] The data transmission and integration subunit is used to transmit the acquired temperature, ambient temperature and humidity, and power consumption to the central processing unit. The central processing unit integrates the temperature, ambient temperature and humidity, and power consumption into a unified data stream and stores it.

[0102] All the acquired data such as temperature, ambient temperature and humidity, and power consumption will be transmitted to the central processing unit (CPU) through the built-in communication bus or sensor interface. These data will be integrated into a unified data stream for use by the subsequent temperature prediction and temperature control decision-making engine. During the transmission process, to ensure the accuracy and integrity of the data, the system needs to verify the data during the transmission process and handle the problems of transmission delay or packet loss to ensure real-time performance.

[0103] In addition, to ensure data integrity and meet the requirements of subsequent analysis, the collected temperature, ambient temperature and humidity, and power consumption data will be stored in a dedicated storage module. The data storage module needs to be able to efficiently and securely store a large amount of real-time data, and also have fast access capabilities for historical data retrieval or analysis when needed.

[0104] The tablet computer of this embodiment can collect temperature data, ambient temperature and humidity information, and system power consumption data of its various key parts in real time. Through precise data collection and processing, it ensures that the temperature control system can perform subsequent thermal field prediction, decision-making generation, and temperature control execution based on real and accurate data. These data provide the necessary input for the entire temperature control system and ensure the efficient operation of the system.

[0105] As Figure 5 shown, in some embodiments, the temperature control device may further include: a scenario recognition and adaptive policy switching unit, configured to identify the working mode of the tablet computer by analyzing the application process list, touch event frequency, and screen brightness change rate of the tablet computer, and automatically switch the temperature control policy according to the preset policy library according to the working mode of the tablet computer.

[0106] As Figure 6 shown, in some embodiments, the scenario recognition and adaptive policy switching unit specifically includes: an information collection subunit, a touch event frequency monitoring subunit, a screen brightness change rate detection subunit, a temperature control policy dynamic adjustment subunit, and a real-time monitoring and mode switching subunit.

[0107] The information collection subunit is configured to collect information about currently running applications by monitoring the application process list.

[0108] This embodiment first collects information about currently running applications by monitoring the application process list. The loads of these applications will affect the performance requirements of the tablet computer. For example, when the user runs high-intensity games or video rendering, the resource consumption of the CPU and GPU is relatively high, and the system will identify it as the high-performance mode; when running light-load applications such as daily office software or browsers, the system will identify it as the balanced mode. By identifying different application processes, the system can dynamically judge the current workload and provide a basis for subsequent selection of temperature control policies.

[0109] The touch event frequency monitoring subunit is configured to judge the interaction intensity of the tablet computer by monitoring the frequency of touch events, and judge the working mode of the tablet computer according to the interaction intensity. The working modes include high-performance mode, silent mode, and balanced mode.

[0110] In this embodiment, the interaction intensity of the device is judged by monitoring the frequency of touch events. If the touch event frequency is high, it indicates that the user is actively operating the tablet computer, such as engaging in high-intensity activities like playing games or video editing. In this case, the system will consider the tablet computer to be in the high-performance mode and needs to enhance its processing power; on the contrary, low-frequency touch events indicate that the tablet computer is in a relatively static state, such as reading or standby, and at this time the tablet computer will be recognized as the silent mode. In this step, the touch frequency, as an important factor, can accurately identify the user's usage requirements.

[0111] The screen brightness change rate detection subunit is used to detect the screen brightness change rate of the tablet computer and further confirm the working mode of the tablet computer according to the screen brightness change rate.

[0112] The screen brightness change rate is another indicator to help the system judge the application scenario. In high-load scenarios (such as games, video playback, etc.), the screen brightness will change relatively frequently, while in low-load scenarios (such as static reading, standby, etc.), the screen brightness change is small. By monitoring the change of the screen brightness, it is possible to further confirm whether it is in the high-performance mode or the silent mode. For example, when the brightness changes rapidly, it is determined to be a high-load scenario, and then the temperature control strategy is adjusted to the high-performance mode; if the brightness change is small, it is determined that the tablet computer is in a low-load state and enters the silent mode.

[0113] The application scenario recognition subunit is used to identify the current application scenario according to the application program information, the frequency of touch events, and the screen brightness change rate.

[0114] After collecting and analyzing the application process, the touch event frequency, and the screen brightness change rate, these data are integrated to identify the current application scenario. For example, in high-load applications (such as 3D games or video rendering), the system will automatically recognize it as the high-performance mode, and at this time more heat dissipation and performance improvement are required; when multitasking or browsing the web, the system enters the balanced mode and dynamically adjusts the fan speed and CPU frequency to balance the temperature and performance; while in low-load applications such as reading and standby, the system will recognize it as the silent mode, turn off the fan and control the temperature by reducing the frequency.

[0115] The temperature control strategy dynamic adjustment subunit is used to automatically select and switch the appropriate temperature control strategy from the preset temperature control strategy library according to the current application scenario.

[0116] In this embodiment, the temperature control strategy is automatically selected and switched from a preset strategy library according to the recognized scenario. In the high-performance mode, the fan is preferentially started to run at high speed, and the CPU is allowed to overclock for a short time to ensure sufficient processing power under high load; in the balanced mode, the voltage and frequency of the CPU / GPU, as well as the fan speed, are dynamically adjusted to balance the relationship between temperature and performance; in the silent mode, the fan is turned off, and the CPU frequency is reduced to reduce power consumption and noise, thereby improving user comfort.

[0117] The real-time monitoring and mode switching subunit is used to continuously monitor the working state of the tablet computer, and adjust the working mode of the tablet computer in real time according to the working state, and dynamically adjust the temperature control strategy of the tablet computer according to the working mode.

[0118] During the operation of the tablet computer, continuously monitor the working state of the tablet computer, and continuously adjust the temperature control strategy according to the working state (such as CPU load, temperature change, touch frequency, etc.). If the user switches from a high-performance application to a low-load application, it will automatically switch to the silent mode or the balanced mode; on the contrary, when the load increases, it will promptly switch to the high-performance mode to enhance heat dissipation and performance. Through this dynamic adjustment, the best temperature control effect can be maintained.

[0119] Specifically, the external environmental temperature, humidity and other conditions of the tablet computer will change. Even if the temperature control strategy has been adjusted in a specific mode, if the environmental temperature suddenly rises, the heat dissipation will be affected. At this time, it is necessary to adjust the strategy in real time according to the new temperature change to maintain the best temperature control effect. For example, in a high-temperature environment in summer, the temperature and performance balance can be maintained well in the balanced mode originally. However, as the environmental temperature rises, it is necessary to further adjust the fan speed or reduce the CPU frequency to prevent overheating.

[0120] Although the mode switching and temperature control strategy adjustment have been carried out according to the application switching situation, the workload of the device is constantly changing during actual operation. Taking the balanced mode as an example, when running some light office software, the system will dynamically adjust the voltage, frequency and fan speed according to real-time data such as CPU load. Because even in a low-load office scenario, different operations such as opening multiple documents or performing complex graphic rendering will cause different CPU loads, and it is necessary to continuously adjust the strategy to adapt to these subtle changes to achieve the best balance between temperature and performance.

[0121] In addition, the ways and scenarios in which users use the device are diverse and uncertain. Through continuous monitoring and adjustment, it can be ensured that no matter how the user uses the device, whether it is playing games for a long time, watching videos, or occasionally performing simple browsing operations, the device can provide stable performance and a comfortable user experience in different situations. For example, when the user temporarily opens a large software during use, the system can promptly detect the load change and switch to the high-performance mode, enhancing heat dissipation and performance, avoiding throttling caused by lag or overheating, and enabling the user to obtain a smooth experience.

[0122] Moreover, the hardware performance and lifespan of the tablet computer are closely related to temperature. Continuous monitoring and adjustment of the temperature control strategy can prevent the device from being in an overheated or overcooled state for a long time, thereby protecting the hardware and extending the service life of the device. For example, it can prevent the CPU from experiencing electron migration due to long-term high temperature, which affects its performance and stability. At the same time, a reasonable temperature control strategy can also enable the device to maintain good performance under different loads and give full play to the potential of the hardware.

[0123] By identifying different application scenarios (such as high-performance mode, balanced mode, silent mode) in the embodiments of the present invention, the temperature control strategy can be automatically switched according to the change of the application scenario. This scenario-adaptive mechanism realizes personalized and flexible adjustment in the temperature control of the tablet computer, adjusts the temperature control strategy according to the actual usage requirements and environment, and improves the user experience.

[0124] As Figure 5 shown, in some embodiments, the temperature control device may further include: a real-time inference and control execution unit, which is used to adjust the temperature control strategy according to the predicted thermal field distribution heat map, record the user's behavior habits, learn the usage preferences at different time periods, and pre-load the temperature control strategy suitable for the user's current needs according to the behavior habits and usage preferences.

[0125] In this embodiment, the calculation tasks of the thermal field prediction model and the decision engine will be assigned to the CPU (Central Processing Unit) and NPU (Neural network Processing Unit) of the device. This heterogeneous computing can accelerate the temperature control decision-making process and ensure real-time performance. With the support of the inference engine, the system can not only adjust the temperature control strategy according to the predicted temperature trend, but also record the user's behavior habits and learn the usage preferences at different time periods. In this way, the system can pre-load the temperature control strategy most suitable for the user's current needs and achieve personalized temperature control.

[0126] As Figure 7 shown, in some embodiments, the real-time inference and control execution unit specifically includes: a calculation task allocation sub-unit, an inference engine, a recording and learning sub-unit, a pre-loading sub-unit for personalized temperature control strategies, and a real-time feedback and strategy adjustment sub-unit.

[0127] The computing task allocation subunit is used to allocate the computing tasks of the thermal field prediction model and the decision engine to the processing units of the device, including the CPU and the NPU.

[0128] In the real-time inference stage, calculations need to be performed on the thermal field prediction model and the decision engine. Since these two tasks involve complex calculations, the system adopts a heterogeneous computing method to allocate these computing tasks to the CPU and NPU of the tablet computer. The NPU (Neural Network Processing Unit) is good at handling a large number of parallel calculations and is suitable for executing the calculations of the thermal field prediction and deep learning models, which can accelerate the temperature prediction process. The CPU is responsible for handling the control logic and other computing tasks in the decision engine to ensure that the system can make temperature control decisions comprehensively and quickly. This task allocation optimizes the computing resources, improves the response speed of the system, and ensures that real-time inference can be completed within 10 ms.

[0129] The inference engine is used to perform inference calculations based on the temperature data, power consumption, temperature and humidity data collected in real time to predict the future thermal field distribution.

[0130] The inference engine is used to perform inference calculations based on the input such as the temperature data, power consumption, and environmental information collected in real time to predict the future thermal field distribution. Through the inference engine, the prediction results of the thermal field distribution can be updated in real time, and then the temperature control strategy can be adjusted. For example, if the predicted temperature tends to be too high, the fan will be started in advance or the CPU frequency will be reduced to prevent the tablet computer from overheating. The inference engine supports low-latency operation to ensure that the temperature control decision can respond quickly during the actual use of the tablet computer and avoid adverse effects of temperature fluctuations on the user experience.

[0131] The recording and learning subunit records the user's behavior habits and learns the usage preferences at different time periods.

[0132] To achieve personalized temperature control, it is necessary to record the user's behavior habits and learn the usage preferences at different time periods. By analyzing historical usage data, such as the user's operation habits, application scenarios, and temperature control strategy preferences, a user model is established. On this basis, the user's temperature control requirements can be inferred, and the most suitable temperature control strategy can be pre-loaded in different application scenarios. For example, at night, the user is more inclined to use the silent mode with low noise, while during the day or when playing games, high-performance mode is more needed. In this way, the most suitable temperature control strategy can be intelligently selected according to the user's habits and preferences to improve the user experience.

[0133] The pre-loading subunit of the personalized temperature control strategy is used to pre-load the personalized temperature control strategy according to the learning of the user's behavior habits.

[0134] Through the learning of the user's behavior habits, this embodiment can preload personalized temperature control strategies for quick application when the user starts using the tablet computer. By analyzing the usage patterns at different time periods, it can predict the user's needs and preload the temperature control strategies suitable for the current usage scenario in advance. For example, when the system detects that the user enters a high-load application (such as a game or video editing), it can automatically enable the high-performance mode; while in a quiet environment, it will switch to the silent mode, turn off the fan and reduce the CPU frequency. The implementation of the preloading strategy not only improves the temperature control efficiency but also reduces the latency caused by strategy switching, making the temperature control response more sensitive and personalized.

[0135] In addition, in some embodiments, it may further include a real-time feedback and strategy adjustment subunit for real-time monitoring of the temperature change and user behavior of the tablet computer. When implementing the temperature control strategy, it real-time monitors the temperature change and user behavior of the tablet computer to ensure the effectiveness of the temperature control strategy. If it is real-time monitored that the temperature exceeds the set range or an abnormality occurs, it will automatically adjust the temperature control strategy. For example, if the temperature rises too fast, it will temporarily increase the fan speed or adjust the CPU frequency to prevent the device from overheating. In addition, it will continuously adjust and optimize the temperature control strategy according to the user's feedback and usage pattern to ensure that it can always adapt to the user's needs and the state change of the tablet computer.

[0136] Through real-time inference and control execution, the embodiment of the present invention can achieve fast temperature control decision-making and personalized optimization according to the user's behavior habits. Through heterogeneous computing, the thermal field prediction and decision-making tasks are assigned to the CPU and NPU, ensuring the high efficiency and real-time performance of the system; at the same time, by learning the user's behavior and preloading personalized temperature control strategies, the user experience is further improved. The real-time feedback mechanism enables the temperature control strategy to flexibly respond to the changes in the device state, ensuring that the temperature control effect of the tablet computer is always in the best state in various application scenarios.

[0137] As Figure 5 shown, in some embodiments, the temperature control device may further include: an abnormality diagnosis and model fine-tuning unit for continuously monitoring the actual temperature of each area of the tablet computer, comparing the actual temperature with the predicted temperature, and when the difference between the monitored actual temperature and the predicted temperature exceeds the set threshold, dynamically adjusting the thermal field prediction model according to the usage environment and load change of the tablet computer, and real-time adjusting the temperature control strategy according to the adjusted thermal field prediction model.

[0138] In this embodiment, scene recognition is performed according to the current application scenario of the tablet computer. By analyzing the application process list, touch event frequency, and screen brightness change rate, the usage status of the tablet computer can be recognized. For example, when high-intensity games or video rendering are being performed, it is recognized as the high-performance mode, while when reading or in standby, it is recognized as the silent mode. After the scene is recognized, the temperature control strategy is automatically switched according to the preset policy library. Specifically, in the high-performance mode, the high-speed operation of the fan is preferentially enabled, and the CPU is allowed to overclock for a short time to improve performance. In the balanced mode, the voltage and frequency of the CPU or GPU, as well as the fan speed, are dynamically adjusted to ensure the balance between temperature and performance. In the silent mode, temperature control is achieved only by downclocking, and at the same time, the fan is turned off to reduce power consumption and noise.

[0139] The abnormal temperature self-diagnosis function in this embodiment can continuously detect the difference between the predicted temperature and the actual temperature, and trigger the online fine-tuning function of the thermal field prediction model. This enables the thermal field prediction model to be adjusted according to environmental changes and usage status, improving the temperature control accuracy and system robustness.

[0140] As Figure 8 shown, in some embodiments, the abnormal diagnosis and model fine-tuning unit specifically includes: a temperature detection subunit, an abnormal diagnosis subunit, an online fine-tuning subunit, a verification and evaluation subunit, and a real-time adjustment temperature control strategy subunit.

[0141] The temperature detection subunit is used to continuously monitor the actual temperature data of the tablet computer, and compare the monitored actual temperature data with the predicted temperature data. When the difference between the monitored actual temperature data and the predicted temperature data exceeds the set temperature error threshold, the abnormal diagnosis subunit is triggered.

[0142] This embodiment continuously monitors the data of each temperature sensor of the tablet computer, and compares the actual temperature with the predicted temperature. When the difference between the actual temperature and the predicted temperature exceeds the set threshold (such as 3°C), the system will trigger the abnormal temperature detection mechanism. This mechanism can ensure that the temperature control system can perceive the deviation of temperature prediction in real time, and avoid overheating or temperature control failure of the tablet computer. The goal of abnormal detection is to quickly respond to the emerging temperature control problems by accurately comparing the predicted and actual temperatures, and ensure the reliability of the temperature control system.

[0143] The abnormal diagnosis subunit is used to evaluate the accuracy of the current thermal field prediction model according to the difference between the monitored actual temperature data and the predicted temperature data, and analyze the reason for exceeding the set temperature error threshold.

[0144] In this embodiment, once the system detects an abnormal temperature, it will activate the abnormal diagnosis mechanism. This mechanism first evaluates the current accuracy of the temperature prediction model and analyzes the reasons for the temperature prediction error. It checks the status of the temperature sensors to confirm whether there are sensor failures or data anomalies, and at the same time analyzes the impact of the current environment and load conditions on the temperature. This process helps to locate the root cause of the temperature control failure and provides a basis for subsequent model fine-tuning to ensure adaptability under various environmental changes.

[0145] An online fine-tuning subunit, configured to adjust the thermal field prediction model according to the reasons, so that the thermal field prediction can accurately reflect the changing trends of the current environment and load, and obtain a fine-tuned thermal field prediction model.

[0146] When the system identifies a significant deviation between the predicted temperature and the actual temperature, the system will automatically activate the online fine-tuning mechanism. The online fine-tuning mechanism adjusts the parameters of the thermal field prediction model so that the model can more accurately reflect the changes in the current environment and load. This process includes optimizing model weights, short-term incremental training, and data update. Through this fine-tuning, the system can improve the accuracy of temperature prediction, ensure that the temperature control strategy can accurately respond to real-time temperature changes, and avoid overheating or temperature control failure. Among them, optimizing model weights means that when the online fine-tuning mechanism is activated, the weights of the thermal field prediction model will be optimized first. Specifically, based on the error between the actual temperature and the predicted temperature, the gradient descent method or other optimization algorithms are used to adjust the weights of the model to make the prediction results more accurate. The optimization of the weights adjusts the influence of each parameter in the model on the prediction results, ensuring that the model can better adapt to the current load conditions and environmental conditions. This process is the core step of fine-tuning, which can help the model reduce errors and improve prediction accuracy. Short-term incremental training means that after optimizing the weights, short-term incremental training is carried out. The training in this stage uses new temperature data, especially the data obtained from the current actual environment and load, to further adjust the parameters of the model. Different from traditional full-scale training, incremental training only uses new data for update, avoiding the computational cost of completely retraining the model. The purpose of incremental training is to enable the model to quickly adapt to new temperature patterns while avoiding overfitting to the existing data, thus maintaining the generalization ability of the model. Data update means that during the fine-tuning process, the training data will also be updated in real time. This means that the system will update its dataset according to new temperature data, environmental variables, and load conditions, and incorporate these new data into the model training process. Data update ensures that the model can always make predictions based on the latest environmental changes, thereby improving the model's prediction ability for future temperature changes. By continuously receiving and processing new data, the tablet computer can timely adjust the temperature control strategy and avoid temperature control failure caused by outdated data.

[0147] A verification and evaluation subunit is used to verify whether the fine-tuned thermal field prediction model can accurately predict temperature and provide the correct temperature control strategy by using actual temperature data.

[0148] After completing the model fine-tuning, the new model is verified and evaluated to ensure its effectiveness. By using new temperature data, especially data containing abnormal temperature changes, it is verified whether the fine-tuned model can accurately predict temperature and provide the correct temperature control decision. The evaluation process includes comparing with the non-fine-tuned model, analyzing whether the error after fine-tuning is significantly improved, ensuring that the model can operate stably in the long term, and avoiding overfitting or other unnecessary errors.

[0149] A real-time temperature control strategy adjustment subunit is used to adjust the temperature control strategy in real time according to the thermal field prediction model verified by the verification and evaluation subunit.

[0150] The fine-tuned model will be used to adjust the temperature control strategy in real time. When the predicted temperature is more accurate, temperature fluctuations can be identified earlier and measures can be taken. For example, when it is predicted that the temperature is too high, a fan with a higher rotation speed will be enabled in advance or the CPU frequency will be lowered to prevent the tablet computer from overheating. When the temperature is low, the fan power consumption will be reduced and the noise will be reduced, thereby saving energy and improving the user experience. Through this real-time adjustment, the system can make a more accurate and rapid response when the temperature changes.

[0151] In this embodiment, by continuously monitoring and adjusting the temperature prediction model, it is ensured that the temperature control strategy always maintains high precision. Through the online fine-tuning mechanism, it can quickly adapt to different usage environments and load changes, avoiding temperature runaway or overheating of the tablet computer. This mechanism enhances the adaptability of the temperature control system, enabling it to operate stably in different situations and improving the overall user experience.

[0152] As Figure 5 described, in some embodiments, the temperature control device may further include: a temperature control execution and feedback adjustment unit for performing hardware control tasks according to the optimal temperature control strategy to adjust the temperature of the tablet computer in real time; wherein, the hardware control tasks include but are not limited to controlling the rotation speed of the fan, adjusting the frequency of the processor, and adjusting the screen brightness.

[0153] In this embodiment, once the optimized temperature control strategy is determined, the system will execute relevant hardware control tasks. For example, by controlling the rotation speed of the fan, adjusting the frequency of the CPU or GPU, or adjusting the screen brightness, etc., the temperature of the tablet computer is adjusted in real time. At this time, the temperature change is continuously monitored, and the strategy is dynamically adjusted according to the real-time feedback to ensure that the temperature of the tablet computer always remains within a safe range. Through this closed-loop control, a stable temperature control effect can be maintained under various usage conditions, ensuring that the user experience is not affected.

[0154] AsFigure 9 As shown, in some embodiments, the temperature control execution and feedback adjustment unit specifically includes: an execution optimization temperature control strategy subunit, a real-time temperature change monitoring subunit, a dynamic temperature control strategy adjustment subunit, a closed-loop control and feedback optimization subunit, and a user experience monitoring and adjustment subunit.

[0155] The execution optimization temperature control strategy subunit is used to implement the optimal temperature control strategy through hardware control tasks.

[0156] In this embodiment, once the optimized temperature control strategy is determined, it will be implemented through hardware control tasks. These control tasks include adjusting the fan speed, adjusting the frequency of the CPU or GPU, adjusting the screen brightness, etc. By these means, measures can be taken quickly to affect the temperature of the tablet computer. For example, the fan speed can be increased as needed to enhance the heat dissipation capacity, or the fan speed can be reduced to reduce power consumption and noise; the frequency of the CPU or GPU can be dynamically adjusted according to the temperature change, thereby controlling the thermal output of the system and reducing the risk of overheating; the adjustment of the screen brightness is also part of the temperature control. Especially in the low-load case, the screen brightness is reduced to reduce power consumption.

[0157] The real-time temperature change monitoring subunit is used to continuously monitor the temperature change trend of the tablet computer.

[0158] In this embodiment, the temperature change of the tablet computer will be continuously monitored to ensure the effectiveness of the temperature control strategy. This includes the collection and analysis of real-time data provided by the temperature sensor. The temperature change data will be transmitted to the central processing unit (CPU) for processing, and the current temperature status will be fed back in real time. For example, the CPU temperature (T CPU ), GPU temperature (T GPU ), battery temperature (T Battery ) and other data will be used to evaluate the overall thermal state of the tablet computer and determine whether the temperature control strategy needs to be adjusted. The system will continuously evaluate whether it is necessary to enhance or reduce the heat dissipation measures according to these data.

[0159] The dynamic temperature control strategy adjustment subunit is used to dynamically adjust the temperature control strategy of the tablet computer according to the temperature change trend of the tablet computer.

[0160] In this embodiment, according to the real-time monitored temperature data, the system will dynamically adjust the temperature control strategy. For example, if the system detects that the temperature of a certain area (such as the CPU or GPU) is too high (exceeding the safety threshold), it will temporarily increase the fan speed or reduce the CPU frequency to quickly reduce the temperature; if the temperature is relatively low, the fan speed will be reduced to reduce power consumption and improve the battery life of the tablet computer. The adjusted temperature control strategy will take effect immediately according to the feedback data to ensure that the temperature of the tablet computer is always kept within the safe range and avoid overheating or performance limitation.

[0161] The closed-loop control and feedback optimization sub-unit is used to evaluate the temperature control strategy according to the current temperature data of the tablet computer after each adjustment of the temperature control strategy, obtain the temperature control effect of the temperature control strategy, and fine-tune the temperature control strategy according to the temperature control effect.

[0162] After implementing the temperature control strategy, the system will enter the closed-loop control stage, continuously feedback and adjust. After each strategy adjustment, the temperature control system will re-evaluate the current temperature and the effect of the strategy. If the temperature fails to reach the expected target, the strategy will be fine-tuned according to the feedback. This closed-loop control mechanism ensures that the temperature control strategy can flexibly respond under various usage conditions. For example, when the user switches from a high-intensity application to a low-intensity application, the temperature control system will correspondingly adjust the fan speed and CPU frequency to achieve the best temperature control effect. During this process, the temperature control strategy will also be continuously optimized according to the long-term operation data and environmental changes of the tablet computer, further improving the stability and accuracy of the temperature control system.

[0163] In addition, in some embodiments, it further includes a user experience monitoring and adjustment sub-unit, which is used to continuously monitor the user experience and ensure that the impact of the temperature control strategy on user operations is minimized. In this embodiment, the user experience will be continuously monitored to ensure that the impact of the temperature control system on user operations is minimized. For example, indicators such as the noise level, response speed, and battery life of the device are monitored to ensure that the temperature control process will not significantly affect the user experience. If excessive noise or performance degradation (such as lags caused by frequent downclocking) is detected, it will adjust the strategy accordingly to seek the best balance between performance and temperature control. Through these feedback mechanisms, not only can the stable operation of the tablet computer be maintained, but also the user experience can be optimized to ensure that users can obtain a smooth usage experience under different load conditions.

[0164] The temperature control execution and feedback adjustment in this embodiment is a continuous process, which dynamically adjusts the temperature control strategy according to the actual temperature data monitored in real time to ensure that the tablet computer can maintain a safe temperature in various usage environments. Through closed-loop control, the temperature control strategy can be optimized according to real-time feedback, and while ensuring the temperature control effect, the performance can be maximally improved, the power consumption and noise can be reduced, so as to ensure that the user experience is not affected.

[0165] Embodiment 2

[0166] As Figure 10 shown, the present invention also provides a temperature control method for a tablet computer, and the temperature control method includes the following steps:

[0167] S11: Collect the perception data of the tablet computer, where the perception data includes the temperature data of multiple regions of the tablet computer, the temperature and humidity data of the surrounding environment, and the power consumption data during the operation of the tablet computer;

[0168] S12: Preprocess the collected sensing data and fuse the preprocessed sensing data into a high-dimensional feature vector;

[0169] S13: Input the high-dimensional feature vector and the posture of the tablet computer into a thermal field prediction model using a convolutional long short-term memory network, and predict the temperature data of each area of the tablet computer through the thermal field prediction model and generate a thermal field distribution heat map;

[0170] S14: Define an objective function according to the temperature distribution in the thermal field distribution heat map and in combination with multiple target parameters, and the objective function dynamically assigns weight coefficients to each target parameter according to the usage environment of the tablet computer; the target parameters include fan power consumption, the down-frequency amplitude of the central processing unit, and the noise generated during the operation of the tablet computer;

[0171] S15: Generate a Pareto optimal solution set using a multi-objective optimization algorithm according to the objective function and the dynamically adjusted weight coefficients, and the Pareto optimal solution set contains multiple candidate temperature control strategies; select the optimal temperature control strategy suitable for the current application scenario from the multiple candidate temperature control strategies through fuzzy logic decision-making, and transmit the optimal temperature control strategy to the execution unit;

[0172] S16: Adjust the hardware parameters of the tablet computer according to the optimal temperature control strategy.

[0173] For specific implementation details, please refer to Figures 1 to 9 the temperature control device shown.

[0174] Embodiment III

[0175] As Figure 11 shown, an embodiment of the present invention also provides a computer-readable storage medium, and program codes for executing the steps in the method embodiments are stored in the computer-readable storage medium. When the program codes are executed by a processor, the temperature control method of any one of the above-mentioned tablet computers is implemented. Figures 1 to 9 If the integrated module / unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on such an understanding, all or part of the processes in the method embodiments of the present invention can also be completed by instructing relevant hardware through a computer program. The computer program can be stored in a computer-readable storage medium, and when the computer program is executed by a processor, the steps of the method embodiments can be implemented.

[0176] Embodiment IV

[0177] Figure 12It is a schematic structural diagram of an electronic device according to an embodiment of the present invention. An embodiment of the present invention also provides an electronic device. Please refer to Figure 12 , at the hardware level, the electronic device includes a processor, and optionally also includes an internal bus, a network interface, and a memory. Among them, the memory includes internal memory, such as high-speed random access memory (Random-Access Memory, RAM), and also includes non-volatile memory, such as at least one disk memory, etc. Of course, the electronic device also includes other hardware required for other services. The processor, network interface, and memory can be interconnected through the internal bus. The bus can be divided into an address bus, a data bus, a control bus, etc. The memory is used to store a program, and the program can include program code, and the program code includes computer operation instructions. The memory can include internal memory and non-volatile memory, and provide instructions and data to the processor. The processor reads the corresponding computer program from the non-volatile memory into the internal memory and then runs it, and executes Figure 10 a temperature control method for a tablet computer shown. Each embodiment in this specification is described in a related manner. For the same and similar parts among the embodiments, reference can be made to each other. The key points of each embodiment are the differences from other embodiments.

[0178] The above is only a preferred embodiment of the present invention, and is not intended to limit the protection scope of the present invention. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present invention are all included in the protection scope of the present invention.

Claims

1. A temperature control device for a tablet computer, characterized in that The temperature control device includes: A sensing data acquisition unit, which is used to collect temperature data of multiple regions in real time at a preset frequency through a thermocouple array distributed in the CPU, GPU, battery, screen driver chip, and the housing grip area, and synchronously collect ambient temperature and humidity data and power consumption data; A data preprocessing and fusion unit, which is used to perform denoising, data synchronization, time alignment, and normalization processing on the temperature data, the ambient temperature and humidity data, and the power consumption data, and fuse the preprocessed data into a high-dimensional feature vector containing spatial heat distribution features and time series features; A thermal field prediction unit, which is used to input the high-dimensional feature vector and the posture of the tablet computer into a thermal field prediction model using a convolutional long short-term memory network, and perform lightweight processing on the thermal field prediction model. The lightweight processing includes parameter compression, quantization, and knowledge distillation; predict the temperature data of each region of the tablet computer through the thermal field prediction model and generate a thermal field distribution heat map; An objective function definition unit, which is used to define an objective function according to the temperature distribution in the thermal field distribution heat map and in combination with multiple objective parameters. The objective function dynamically assigns weight coefficients to each objective parameter according to the usage environment of the tablet computer; A temperature control decision engine, which is used to generate a Pareto optimal solution set using a multi-objective optimization algorithm according to the objective function and the weight coefficients. The Pareto optimal solution set contains multiple candidate temperature control strategies; select the optimal temperature control strategy suitable for the current application scenario from the multiple candidate temperature control strategies through fuzzy logic decision-making, and transmit the optimal temperature control strategy to the execution unit; An execution unit, which is used to adjust the hardware parameters of the tablet computer according to the optimal temperature control strategy; The temperature control device further includes: An anomaly diagnosis and model fine-tuning unit, which is used to continuously monitor the actual temperature data of each region of the tablet computer, and compare the actual temperature data with the predicted temperature data. When the difference between the actual temperature data and the predicted temperature data exceeds a set threshold, dynamically adjust the thermal field prediction model according to the usage environment and load changes of the tablet computer, and adjust the temperature control strategy in real time according to the adjusted thermal field prediction model.

2. The temperature control device of the tablet computer according to claim 1, characterized in that The sensing data acquisition unit specifically includes: A thermocouple array, which is used to collect temperature data of the multiple regions in real time; An ambient temperature and humidity data acquisition sub-unit, which is used to synchronously collect ambient temperature data and ambient humidity data of the surrounding environment of the tablet computer; A power consumption monitoring sub-unit, which is used to monitor the power consumption data of the tablet computer in real time.

3. The temperature control device of the tablet computer according to claim 1, wherein The temperature control device further includes: A scenario recognition and adaptive strategy switching unit, which is used to determine the working mode of the tablet computer by analyzing the application process list, the frequency of touch events, and the screen brightness change rate of the tablet computer, and automatically switch the temperature control strategy according to the working mode of the tablet computer according to a preset temperature control strategy library.

4. The temperature control device of the tablet computer according to claim 3, characterized in that, The scenario recognition and adaptive strategy switching unit specifically includes: An information collection sub-unit, which is used to collect information about the currently running applications by monitoring the application process list; A touch event frequency monitoring subunit, which is used to judge the interaction intensity of the tablet computer by monitoring the frequency of touch events, and judge the working mode of the tablet computer according to the interaction intensity; A screen brightness change rate detection subunit, which is used to detect the screen brightness change rate of the tablet computer, and further confirm the working mode of the tablet computer according to the screen brightness change rate; An application scenario recognition subunit, which is used to recognize the current application scenario according to the application program information, the frequency of touch events, and the screen brightness change rate; A temperature control strategy dynamic adjustment subunit, which is used to automatically select and switch to a suitable temperature control strategy from a preset temperature control strategy library according to the current application scenario.

5. The temperature control device of the tablet computer according to claim 1, characterized in that The temperature control device further includes: A real-time inference and control execution unit, which is used to adjust the temperature control strategy according to the predicted thermal field distribution heat map, record the user's behavior habits and learn the usage preferences of different time periods, and preload the temperature control strategy suitable for the user's current needs according to the behavior habits and the usage preferences.

6. The temperature control device of the tablet computer according to claim 1, characterized in that, The anomaly diagnosis and model fine-tuning unit specifically includes: A temperature detection subunit, which is used to continuously monitor the actual temperature data of each area of the tablet computer, compare the monitored actual temperature data with the predicted temperature data, and trigger the anomaly diagnosis subunit when the difference between the monitored actual temperature data and the predicted temperature data exceeds the set temperature error threshold; An anomaly diagnosis subunit, which is used to evaluate the accuracy of the current thermal field prediction model according to the difference between the monitored actual temperature data and the predicted temperature data, and analyze the reason for exceeding the set temperature error threshold; An online fine-tuning subunit, which is used to adjust the thermal field prediction model according to the reason, so that the thermal field prediction can accurately reflect the change trend of the current environment and load, and obtain a fine-tuned thermal field prediction model; A verification and evaluation subunit, which is used to verify whether the fine-tuned thermal field prediction model can accurately predict the temperature and provide the correct temperature control strategy by using the actual temperature data; A real-time temperature control strategy adjustment subunit, which is used to adjust the temperature control strategy in real time according to the thermal field prediction model verified by the verification and evaluation subunit.

7. The temperature control device of the tablet computer according to claim 1, characterized in that, The temperature control device further includes: A temperature control execution and feedback adjustment unit, which is used to execute hardware control tasks according to the optimal temperature control strategy to adjust the temperature of the tablet computer in real time.

8. The temperature control device of the tablet computer according to claim 7, characterized in that, The temperature control execution and feedback adjustment unit specifically includes: An execution optimization temperature control strategy subunit, which is used to implement the optimal temperature control strategy through hardware control tasks; A real-time temperature change monitoring subunit, which is used to continuously monitor the temperature data of the tablet computer; A dynamic temperature control strategy adjustment subunit, which is used to dynamically adjust the temperature control strategy of the tablet computer according to the temperature data of the tablet computer; A closed-loop control and feedback optimization subunit, which is used to evaluate the temperature control strategy according to the current temperature data of the tablet computer after each adjustment of the temperature control strategy, obtain the temperature control effect of the temperature control strategy, and fine-tune the temperature control strategy according to the temperature control effect.

9. A temperature control method for a tablet computer, characterized in that, The temperature control method is based on the temperature control device of the tablet computer described in any one of claims 1-8, and the temperature control method includes: Collecting the sensed data of the tablet computer; Preprocessing the collected sensed data and fusing the preprocessed sensed data into a high-dimensional feature vector; Inputting the high-dimensional feature vector and the posture of the tablet computer into a thermal field prediction model using a convolutional long short-term memory network, and predicting the temperature data of each area of the tablet computer through the thermal field prediction model and generating a thermal field distribution heat map; Defining an objective function according to the temperature distribution in the thermal field distribution heat map and in combination with multiple target parameters, and the objective function dynamically assigns a weight coefficient to each target parameter according to the usage environment of the tablet computer; Generating a Pareto optimal solution set using a multi-objective optimization algorithm according to the objective function and the weight coefficient, where the Pareto optimal solution set contains multiple candidate temperature control strategies; selecting the optimal temperature control strategy suitable for the current application scenario from the multiple candidate temperature control strategies through fuzzy logic decision-making, and transmitting the optimal temperature control strategy to the execution unit; Adjusting the hardware parameters of the tablet computer according to the optimal temperature control strategy.

Images