Tablet computer heat dissipation control system based on intelligent temperature control algorithm

Abstract

The invention belongs to the technical field of heat dissipation control of electronic equipment, and particularly relates to a tablet computer heat dissipation control system based on an intelligent temperature control algorithm, which comprises a data acquisition and processing module, a thermal load prediction module, a self-adaptive control strategy generation module and a heat dissipation execution and feedback learning module, and realizes prediction-decision-execution closed-loop control driven by multi-dimensional data. By introducing the deep learning prediction model based on the LSTM, the temperature change trend of each key hot spot of the tablet computer in a period of time in the future can be predicted in a prospective manner, so that the limitation that passive response can only be performed based on the current temperature in the prior art is eliminated. This predictive capability enables the system to take preventive measures prior to the occurrence of thermal loads, avoiding temperature instantaneous overshoot, and achieving more stable temperature control.

Description

Technical Field

[0001] This invention belongs to the field of heat dissipation control technology for electronic devices, specifically a heat dissipation control system for tablet computers based on intelligent temperature control algorithms. Background Technology

[0002] As a core mobile computing terminal, tablet computers have expanded their applications from entertainment and leisure to business, education, and professional creation, playing a crucial role in the national economy and daily life. Driven by user demands for device performance, functional integration, and slim design, the computing density and operating frequency of core chips such as CPUs and GPUs have continuously increased, leading to a surge in power consumption and significant heat generation. Therefore, efficient and reliable heat dissipation control mechanisms have become a core technological support for ensuring long-term stable operation, maintaining a high-quality user experience, and extending the lifespan of the device.

[0003] To address the heat dissipation issue in tablets, mainstream industry solutions employ a strategy combining hardware heat conduction and software temperature monitoring. On the hardware side, high thermal conductivity materials such as graphene heat dissipation films, vapor chambers, and heat pipes enhance heat conduction, transferring heat from the core heat source to the device casing or cooling module. On the software side, multiple temperature sensors monitor the temperatures of key components such as the CPU, GPU, and battery in real time, managing heat dissipation based on preset thresholds. When the temperature reaches the target, the processor is triggered to reduce frequency to decrease power consumption and suppress heat generation. This solution is intuitive and cost-effective, effectively solving basic overheat protection issues and ensuring operational stability in early mobile devices with low performance requirements and ample design space.

[0004] As technology advances and application scenarios demand higher performance, heat dissipation solutions based on fixed thresholds or simple feedback exhibit inherent limitations. Modern tablets are used in complex scenarios (such as high-intensity gaming and multitasking with heterogeneous loads) and are exposed to fluctuating ambient temperatures. Traditional passive and lagging control methods struggle to balance performance, heat dissipation, and user experience: conservative thresholds lead to premature frequency throttling, sacrificing performance and compromising user experience; higher thresholds cause core overheating, accelerating aging and shortening lifespan, while increased casing temperature affects grip safety. Furthermore, existing technologies lack the ability to predict heat trends, passively responding only to real-time temperatures. This results in sluggish responses to sudden increases in heat load, leading to temperature overshoot and hindering refined, proactive management. This becomes a technological bottleneck in the evolution of devices towards high performance, compact structures, and wide applicability. Therefore, this invention provides a tablet heat dissipation control system based on an intelligent temperature control algorithm. Summary of the Invention

[0005] In order to overcome the shortcomings of the prior art, at least one technical problem raised in the background art is solved.

[0006] The technical solution adopted by this invention to solve its technical problem is as follows: A tablet computer heat dissipation control system based on an intelligent temperature control algorithm, comprising a data acquisition unit, a data processing and fusion unit, a heat load prediction module, an adaptive control strategy generation module, a heat dissipation execution unit, a feedback and learning module, and a safety protection module. The system is installed inside the tablet computer and interacts with the tablet computer's main control unit, power management unit, and operating system via a bus interface to transmit data and control commands.

[0007] Specifically, the data acquisition unit is used to acquire multi-dimensional operational data from both the inside and outside of the tablet computer in real time. The data acquisition unit includes: Multiple temperature sensors, employing high-precision NTC thermistor arrays or digital temperature sensors, have a sensing range of -40℃ to 125℃, a measurement accuracy of ±0.5℃, and a response time of less than 1 second. They communicate with the data processing and fusion unit via I2C or SPI bus protocols. The temperature sensor array is precisely deployed within key heat sources inside the tablet, specifically in locations including, but not limited to, the interior of the central processing unit (CPU) package, the core area of ​​the graphics processing unit (GPU), near the power management integrated circuit (PMIC), the surface of the battery cells inside the battery pack, the edge of the light guide plate of the display backlight module, and typical gripping areas of the tablet's casing. These temperature sensors are responsible for real-time monitoring of temperature changes at these key points and sending the temperature data to the data processing and fusion unit at a preset sampling frequency.

[0008] A power consumption monitoring unit, integrated into the power management integrated circuit (PMIC) of the tablet computer, comprises multiple high-precision current and voltage sensors. This unit monitors the instantaneous power consumption data of key power-consuming components such as the central processing unit, graphics processor, memory, display screen, and wireless communication module in real time, and transmits the power consumption data to the data processing and fusion unit via a PMBus or I2C interface at a preset sampling frequency. The power consumption data includes the actual operating current and voltage of each component.

[0009] An environmental sensor, integrated inside the tablet computer's body, is a digital ambient temperature sensor used to monitor the actual temperature of the environment in which the tablet computer is located. The environmental sensor has a measurement accuracy of ±1.0℃ and communicates with the data processing and fusion unit via the I2C bus protocol.

[0010] A system status monitoring interface is provided, which acquires real-time operating status data of the tablet computer through standard APIs or kernel-level interfaces provided by the operating system. This operating status data includes, but is not limited to: CPU utilization, clock speed and voltage, GPU utilization and clock speed, RAM usage, memory read / write speed, currently running application identifiers, application resource usage, screen brightness settings, battery charging status, battery health status, and charging current. This system status data is pushed to the data processing and fusion unit by the main control unit or the operating system kernel at a preset frequency (e.g., every 200 milliseconds) via shared memory or a message queue mechanism.

[0011] The data processing and fusion unit is electrically connected to the data acquisition unit and is used to receive and process raw data from the data acquisition unit. The data processing and fusion unit is implemented using a dedicated coprocessor embedded in the tablet PC's main control chip or a separate low-power microcontroller unit (MCU). The data processing and fusion unit performs the following processing on the received multi-dimensional raw data: Time synchronization precisely aligns data from different sensors and system interfaces based on timestamps to eliminate time deviations caused by data transmission delays and sampling frequency differences.

[0012] Data filtering and denoising are performed by using Kalman filters or moving average filters to process sensor data, smoothing data fluctuations and removing measurement noise.

[0013] Data normalization and feature extraction normalize data of different dimensions to bring them into a uniform numerical range. Simultaneously, key features relevant to heat load prediction and heat dissipation strategy generation are extracted from the original data. These feature vectors are constructed as fixed-length numerical sequences and used as input to the heat load prediction module.

[0014] The heat load prediction module is electrically connected to the data processing and fusion unit, and is used to predict the temperature change trends of key hotspots on the tablet computer over a future period based on the fused multi-dimensional data. The core of the heat load prediction module is a deep learning model based on a Long Short-Term Memory (LSTM) network. The LSTM model is trained offline using large-scale historical operating data, which covers multi-dimensional data such as temperature, power consumption, and system status of the tablet computer under different ambient temperatures and load types. The input to the LSTM model is the multi-dimensional feature vector output by the data processing and fusion unit, and its output is a sequence of predicted temperature values ​​for each key hotspot over the next N time steps. The LSTM model adopts a multi-layer stacked structure, including an input layer, multiple LSTM layers, a fully connected layer, and an output layer. It is trained using a backpropagation algorithm and the Adam optimizer to minimize the mean square error between the predicted and actual temperatures. After training, the model is deployed to the tablet computer's memory and receives real-time input data during runtime, performs forward inference, and generates prediction results. The prediction results are transmitted to the adaptive control strategy generation module in vector form.

[0015] The adaptive control strategy generation module is electrically connected to the heat load prediction module. Based on the future temperature prediction results and multi-dimensional operating data provided by the heat load prediction module, it dynamically calculates and generates the optimal heat dissipation control strategy. The core of the adaptive control strategy generation module is a model predictive control (MPC) based optimization algorithm. The MPC algorithm operates within a rolling optimization cycle, aiming to maximize the tablet's continuous performance output while meeting device safety and user comfort constraints. The MPC algorithm establishes a multi-objective optimization function that comprehensively considers the following factors: Core temperature constraints ensure that the predicted temperature of core components such as the central processing unit and graphics processing unit does not exceed their preset safe operating upper limit temperature, while keeping them within the temperature range of optimal performance as much as possible.

[0016] The surface temperature of the outer casing is constrained to ensure that the predicted temperature of the tablet casing in the area where the user holds it does not exceed the human comfort threshold.

[0017] To maximize sustained performance output, under the aforementioned temperature constraints, we seek to maximize the average operating frequency and utilization of the CPU / GPU.

[0018] Battery health and battery life: In specific scenarios, priority is given to protecting battery health and extending battery life.

[0019] User experience perception is used to adjust performance weights based on the immediate performance requirements of the currently running application type.

[0020] The control variables of the MPC algorithm include: the frequency and voltage level of the central processing unit (CPU), the frequency and voltage level of the graphics processing unit (GPU), the task scheduling priority at the operating system level, resource limitations of background applications (e.g., CPU utilization limit, network bandwidth limit), battery charging current limit, and display brightness. In each control cycle, the MPC algorithm, based on the future predicted temperature sequence provided by the heat load prediction module and the current system state provided by the data processing and fusion unit, solves the multi-objective optimization function through an iterative optimization algorithm, thereby calculating the optimal combination of control variables for a future control time domain. The optimal combination of control variables is transmitted to the heat dissipation execution unit in the form of control commands.

[0021] The heat dissipation execution unit is electrically connected to the adaptive control strategy generation module, and is used to receive and execute control commands issued by the adaptive control strategy generation module to achieve fine-grained control of the tablet computer's heat dissipation performance and system power consumption. The heat dissipation execution unit is mainly implemented in the following ways: Dynamic Voltage and Frequency Scaling (DVFS) control for the Central Processing Unit (CPU) and Graphics Processing Unit (GPU). The thermal execution unit directly calls the DVFS interface provided by the System-on-Chips (SoC) firmware layer to adjust the frequencies and voltages of each CPU core and the GPU operating frequency and voltage to optimal settings according to the control instructions. This DVFS control achieves nanosecond-level response, allowing for high-granular adjustment of processor power consumption and thus directly controlling heat generation.

[0022] Task scheduling and resource management. The heat dissipation execution unit dynamically adjusts the priority of currently running processes or threads through the API interface provided by the operating system kernel, and limits the CPU, memory, and I / O resources of non-critical background applications.

[0023] Battery charging management. The heat dissipation unit communicates with the power management integrated circuit (PMIC) to dynamically adjust the battery charging current according to the control commands. When it is predicted that the system temperature is about to rise or is already at a high level, the heat dissipation unit can appropriately reduce the charging current or even suspend charging to reduce heat generation during the charging process, thereby avoiding battery overheating and extending battery life.

[0024] Display brightness adjustment. The heat dissipation unit dynamically adjusts the backlight brightness of the display screen via the operating system or display controller interface. In certain specific scenarios, such as when the temperature of the tablet's casing approaches the user's comfort threshold, the screen brightness can be slightly reduced without being noticeably perceived by the user, thereby reducing the display's power consumption and heat generation.

[0025] The passive cooling structure of the tablet computer, such as a graphene heat dissipation film, a heat spreader, a heat pipe, and high thermal conductivity material filling, works in conjunction with the cooling execution unit to rapidly transfer the heat generated by the core chip from the heat source to the device casing by enhancing heat conduction efficiency, and then dissipate it into the environment through radiation, convection, and other means. The intelligent regulation of the cooling execution unit can effectively reduce the source of heat generation, forming a highly efficient overall cooling solution with the passive cooling structure.

[0026] The feedback and learning module is electrically connected to the heat dissipation execution unit and the data acquisition unit to continuously optimize the performance of the heat load prediction module and the adaptive control strategy generation module. The feedback and learning module receives real-time actual temperature data, power consumption data, and system performance data monitored by the data acquisition unit. The feedback and learning module compares the actual data with the predicted temperature output by the heat load prediction module to calculate the prediction error. The feedback and learning module compares the actual performance output with the target performance output of the adaptive control strategy generation module to evaluate the strategy execution effect. The module periodically updates the parameters of the LSTM model to better adapt it to the thermal behavior patterns of the device under long-term operation and different user behaviors. Furthermore, the module adjusts the weight coefficients and constraints of the multi-objective optimization function in the MPC algorithm based on the optimization effect of the actual feedback, enabling the system to adaptively learn and improve its control strategy to better balance performance, heat dissipation, and user experience. The update process can be performed in the background with low priority or when the device is idle.

[0027] The safety protection module is independent of the intelligent temperature control algorithm, but is electrically connected to the data acquisition unit and communicates directly with the main control chip of the heat dissipation execution unit. The safety protection module has dual over-temperature protection thresholds at both the hardware and software levels. When the temperature value detected by any critical temperature sensor exceeds the preset hardware-level absolute safety limit, the safety protection module will force the heat dissipation execution unit to immediately reduce the CPU / GPU frequency to the minimum safe frequency and trigger an emergency shutdown process through a hardware interrupt mechanism to prevent permanent hardware damage. When the temperature value exceeds the preset software-level secondary safety limit, the safety protection module will trigger a system-level warning and forcibly implement measures such as reducing the charging current, pausing charging, or significantly reducing the CPU / GPU frequency. The safety protection module ensures the absolute safety of the device under extreme conditions, serving as the ultimate guarantee for the intelligent temperature control algorithm.

[0028] The beneficial effects of this invention are as follows: 1. The tablet computer heat dissipation control system based on an intelligent temperature control algorithm described in this invention introduces a deep learning prediction model based on LSTM, which can proactively predict the temperature change trends of key hot spots on the tablet computer over a period of time in the future, thus overcoming the limitation of existing technologies that can only passively respond based on the current temperature. This predictive capability enables the system to take preventive measures before the heat load occurs, avoid instantaneous temperature overshoot, and achieve more stable temperature control.

[0029] 2. The tablet PC heat dissipation control system based on an intelligent temperature control algorithm described in this invention achieves multi-dimensional collaborative optimization. This invention is no longer limited to a single temperature threshold judgment, but comprehensively considers multi-dimensional data such as CPU / GPU load, power consumption, battery status, ambient temperature, application type, and user experience requirements. The adaptive control strategy generation module, based on the MPC algorithm, collaboratively optimizes these multi-dimensional data, enabling it to more accurately seek the optimal balance between performance output, heat dissipation efficiency, device lifespan, and user comfort.

[0030] 3. The tablet PC heat dissipation control system based on an intelligent temperature control algorithm described in this invention enables continuous and precise adjustment of multiple execution parameters, such as CPU / GPU frequency and voltage, task scheduling, charging current, and even screen brightness, rather than simply reducing the frequency. The feedback and learning module allows the system to self-learn and adaptively adjust based on actual operating results, continuously optimizing the prediction model and control strategy, thereby improving the system's long-term stability and robustness, and adapting to the personalized needs of different users and constantly changing usage scenarios.

[0031] 4. The tablet computer heat dissipation control system based on intelligent temperature control algorithm described in this invention, through forward-looking prediction and multi-dimensional optimization, can maximize the potential performance output capability of the processor while ensuring that the device temperature is within a safe and comfortable range. This avoids the performance waste caused by conservative frequency reduction in traditional solutions and significantly improves the user experience in high-load application scenarios.

[0032] 5. The tablet computer heat dissipation control system based on an intelligent temperature control algorithm described in this invention can effectively prevent long-term overheating of core components through precise temperature control, reduce thermal stress, thereby delaying hardware aging and improving the overall reliability and lifespan of the tablet computer. Simultaneously, intelligent management of battery charging current can also effectively protect battery health. Attached Figure Description

[0033] The invention will now be further described with reference to the accompanying drawings.

[0034] Figure 1This is a structural framework diagram of the tablet computer heat dissipation control system based on the intelligent temperature control algorithm of this invention. Detailed Implementation

[0035] To make the technical means, creative features, objectives and effects of this invention easier to understand, the invention will be further described below in conjunction with specific embodiments.

[0036] As shown in the figure, the tablet computer heat dissipation control system based on an intelligent temperature control algorithm according to an embodiment of the present invention comprises a data acquisition unit, a data processing and fusion unit, a heat load prediction module, an adaptive control strategy generation module, a heat dissipation execution unit, a feedback and learning module, and a safety protection module. The above functional units exchange data and transmit control commands in real time through a high-speed data bus (e.g., PCIe, MIPI, I2C, or SPI) and internal software interface, forming a highly efficient and responsive closed-loop control system.

[0037] Specifically, the data acquisition unit, as the sensing center of the entire system, is responsible for acquiring key thermal and performance parameters of the tablet computer's internal operating status and external environment in real time and with high precision. This unit integrates a series of heterogeneous sensors and system interfaces to ensure the comprehensiveness and accuracy of data input.

[0038] In a preferred embodiment, the data acquisition unit first includes a distributed temperature sensor network. This network consists of no fewer than 12 high-precision NTC thermistors and 4 digital temperature sensors (e.g., Maxim Integrated's DS18B20 or Analog Devices' TMP117 series). The NTC thermistors, with their fast response and cost advantages, are primarily deployed in space-sensitive areas with large temperature gradients, such as micro-gaps along the CPU / GPU core heat dissipation path, the surface of the power management integrated circuit (PMIC), and between multiple cells within a battery pack. The digital temperature sensors, due to their high precision, ease of digital interface integration, and low temperature drift, are strategically placed in critical areas requiring higher measurement accuracy, such as the integrated temperature sensor interface inside the CPU package, the internal diode temperature detection port of the GPU core, near the heatsink of the RAM module, and areas of the tablet casing where the user frequently holds the device (e.g., the center of the left and right side bezels). All temperature sensors are factory calibrated, with a measurement range covering -40℃ to 125℃, typical measurement accuracy within ±0.5℃, and a response time of less than 1 second. They transmit 12-bit to 16-bit raw temperature data (digital values ​​converted by an analog-to-digital converter or directly output) to the data processing and fusion unit every 100 milliseconds via a dedicated I2C or SPI bus (using a 400kHz I2C or 10MHz SPI clock frequency). To further improve the robustness of temperature measurement, redundant sensor configurations are used for temperature measurements in some key areas, and median filtering or weighted averaging is applied to the multiple data streams.

[0039] Furthermore, the data acquisition unit also integrates a highly precise power consumption monitoring unit. This unit is not a single sensor, but rather a combination of multiple dedicated current-voltage sensors integrated into the tablet's power management unit (PMU or PMIC). Specifically, this includes independent current and voltage measurements of the CPU power supply domain, GPU power supply domain, memory power supply domain, display backlight power supply domain, and wireless communication module (Wi-Fi / Bluetooth / 5G Modem) power supply domain of the main control chip (SoC). For example, accurate acquisition of instantaneous power consumption is achieved by connecting high-precision shunt resistors (e.g., resistance as low as 1mΩ, accuracy ±0.1%) in series on critical power paths, coupled with high-precision differential amplifiers and 16-bit ADCs, or by directly using integrated current / voltage monitoring chips (such as the Texas Instruments INA226 series). This power consumption data is transmitted digitally to the data processing and fusion unit at a frequency of once every 50 milliseconds via PMBus or a high-speed I2C interface (e.g., High-Speed ​​Mode with a rate up to 3.4MHz). The power consumption data includes not only instantaneous current and voltage values, but may also include average power consumption values ​​calculated internally by the PMIC and peak power consumption records, providing key input for subsequent thermal load analysis.

[0040] In addition, the data acquisition unit embeds an environmental sensor, which is a digital ambient temperature and humidity sensor, such as the Bosch Sensortec BME280 or Sensirion SHT31. This sensor is installed inside the tablet PC, near the cooling vents but out of direct contact with internal heat sources, to monitor the actual temperature and relative humidity of the environment in which the device is located in real time. Its typical measurement accuracy is ±1.0℃ (temperature) and ±2.0%RH (humidity), and the data is transmitted once per second via the I2C bus. Ambient temperature data is crucial for evaluating the overall heat dissipation capacity of the device, adjusting thermal model parameters, and predicting heat dissipation rates.

[0041] Meanwhile, the data acquisition unit interacts deeply with the tablet's main control unit and operating system through a system status monitoring interface. This interface utilizes standard APIs provided by the operating system (such as the Android system's sysfs file system interface, proc file system, dumpsys command, or ActivityManager service) or directly accesses specific registers through kernel-level drivers to obtain real-time, detailed system operating status data. This data includes, but is not limited to: the current utilization, clock speed, and dynamic voltage of each core of the central processing unit (CPU); the utilization and clock speed of the graphics processing unit (GPU); the real-time occupancy of system memory (RAM); the read / write speed and I / O load of internal memory (NAND Flash); the process ID, CPU and GPU resource usage of the currently running application in the foreground; the current brightness setting of the screen backlight module; and the battery's state of charge (SoC), remaining capacity, state of health (SoH), and current charging / discharging current. This system status data is pushed from the main control unit or operating system kernel to the data processing and fusion unit every 200 milliseconds via shared memory mechanisms, message queues, or dedicated DMA channels, ensuring rapid capture of dynamic system changes.

[0042] The data processing and fusion unit is closely connected to the data acquisition unit. Its core function is to receive, preprocess, and integrate raw data from various heterogeneous sensors and system interfaces, transforming it into a unified, high-quality, and informative feature vector. In a typical implementation, this unit uses a low-power, high-performance dedicated coprocessor (e.g., an NPU or DSP unit, or a separate ARM Cortex-M series microcontroller with an operating frequency of up to 200MHz, 512KB SRAM, and 2MB Flash memory) integrated within the tablet PC's main control chip.

[0043] First, the data processing and fusion unit performs strict time synchronization. Due to differences in data sampling frequencies and transmission delays among various sensors and system interfaces, the system employs a strategy combining hardware and software timestamps. Each data packet is marked with a precise nanosecond or microsecond timestamp at the acquisition point and aligned using a global high-precision clock source within the data processing and fusion unit. For jitter introduced by network transmission or system scheduling, linear interpolation or nearest-neighbor interpolation is used for time series alignment to eliminate data deviations in the time dimension and ensure that all data points correspond to the system state at the same moment.

[0044] Secondly, data filtering and denoising are crucial steps. For temperature sensor data, considering its response characteristics and environmental noise, the system applies a Kalman filter. This filter establishes a state-space model for each key hot spot, where state variables include the current temperature and its rate of change. Through a predict-update loop, combining the covariance matrix of the sensor's measurement noise and the system's process noise, the Kalman filter effectively smooths the data, removes high-frequency random noise, and makes reasonable estimates for missing data. For quantities such as power consumption data that may have spikes or instantaneous fluctuations, the system uses an exponentially weighted moving average (EWMA) filter, whose smoothing factor (α) is dynamically adjusted according to the data characteristics to achieve a balance between denoising and response speed. For example, for CPU core temperature, the EWMA α value might be set to 0.1 to achieve a smoother trend; while for instantaneous CPU power consumption, the α value might be set to 0.5 to retain faster response characteristics.

[0045] Next, the data processing and fusion unit normalizes and extracts features from all filtered data. Normalization maps data with different physical dimensions (e.g., temperature in degrees Celsius, power consumption in watts, CPU utilization percentage) to a uniform numerical range (e.g., 0 to 1 or -1 to 1) to prevent uneven influence of different dimensions on the results in subsequent machine learning models. A commonly used method is Min-Max normalization, i.e. ; in and These are the minimum and maximum values ​​of each parameter, derived from historical operational data statistics.

[0046] Simultaneously, the system extracts a series of high-level key features from the normalized data, closely related to heat load prediction and heat dissipation strategy generation. The construction of these features fully considers the physical laws of heat transfer and the chip's operating characteristics. For example: Instantaneous power consumption and average power consumption: The average power consumption of each major component over the past 100ms, and the moving average power consumption over the past 1 second, 5 seconds, and 10 seconds.

[0047] Power consumption change rate: The rate at which the power consumption of each component increases or decreases over a short period of time (e.g., ).

[0048] Temperature gradient: the temperature difference between key hot spots (e.g., the temperature difference between the CPU and the casing), and the rate of temperature change of an individual hot spot.

[0049] CPU / GPU load characteristics: CPU core utilization, average clock speed, maximum clock speed duration, and GPU rendering load percentage and texture fill rate.

[0050] Memory and storage I / O characteristics: memory bandwidth utilization, NAND Flash read / write throughput.

[0051] Application behavior pattern identification: Based on the process ID and resource usage pattern of the current foreground application, the application type (e.g., game, video playback, web browsing, document editing) is identified by looking up a table or a lightweight classifier, and its typical hot behavior patterns are associated with it.

[0052] Battery status characteristics: battery SoC, SOH, charging current and battery temperature.

[0053] Environmental factors: ambient temperature and humidity. These extracted features are organized into a fixed-length feature vector (e.g., containing approximately 128 numerical features) and output to the heat load prediction module at a preset frequency (e.g., once every 200 milliseconds) as input to its deep learning model.

[0054] The heat load prediction module is closely connected to the data processing and fusion unit. Its core function is to accurately predict the temperature change trends of key hot spots on the tablet computer over a future period based on time series analysis. At the heart of this module is a deep learning model based on a Long Short-Term Memory (LSTM) network. LSTM models are highly favored for their powerful ability to process and predict time series data; they can capture long-term dependencies in sequence data, which is crucial for predicting complex dynamic thermal behavior patterns.

[0055] During the offline phase, the LSTM model is trained using massive amounts of historical operational data. This historical data was collected under strictly controlled experimental conditions, covering the tablet's performance in various typical usage scenarios and environmental conditions. For example, the ambient temperature range covered by the data collection ranged from 5°C to 40°C; the load types included, but were not limited to: running large 3D games such as Genshin Impact for 1 hour continuously, playing 4K HDR video for 2 hours continuously, conducting a video conference for 30 minutes continuously, performing large CAD rendering tasks, and long-term standby. For each scenario, the system simultaneously recorded real-time temperature, CPU / GPU power consumption, system utilization, battery status, and other multi-dimensional data for each key hotspot, forming a dataset containing millions of time steps.

[0056] In a preferred embodiment, the specific architecture of the LSTM model includes: an input layer that receives an N-dimensional feature vector from a data processing and fusion unit; followed by three stacked LSTM layers, each containing 256 hidden units and employing the ReLU activation function; a Dropout layer (e.g., with a Dropout rate of 0.3) added between each LSTM layer to prevent overfitting; finally, a fully connected layer that maps the output of the LSTM layers to the final prediction result; and an output layer that outputs the predicted temperature sequence of key hotspots (such as CPU cores, GPU cores, PMIC, battery pack interior, and casing grip area) over the next N time steps (e.g., N=30, corresponding to the next 60 seconds, with each step being 2 seconds). The entire model is trained using the backpropagation algorithm and the Adam optimizer (with an initial learning rate of 0.001), aiming to minimize the mean squared error (MSE) between the predicted and actual temperatures. The training process is performed on a high-performance GPU server cluster, taking approximately 72 hours, and the final model's mean absolute error (MAE) on the validation set is less than 0.8°C.

[0057] After training, the parameters (weights and biases) of the LSTM model are compressed and deployed to the tablet's internal memory, and loaded into system memory at runtime for inference computation by the coprocessor. During runtime, the model receives real-time feature vectors from the data processing and fusion unit as input and performs forward inference within nanoseconds to generate a predicted temperature sequence. This predicted sequence is transmitted to the adaptive control strategy generation module in vector form (e.g., an array of 30 predicted temperature values, each corresponding to a point in time two seconds in the future). This forward-looking predictive capability is key to the intelligent heat dissipation management of this invention, enabling the system to anticipate potential thermal overload and take early intervention measures, avoiding performance lag and instantaneous temperature overshoot caused by passive responses.

[0058] The adaptive control strategy generation module is electrically connected to the heat load prediction module. Its core task is to dynamically calculate and generate the optimal heat dissipation control strategy based on the predicted future heat load trend and the current system operating state. This module is implemented based on an advanced model predictive control (MPC) algorithm, which can achieve preset performance targets in complex, multi-variable, and multi-constraint systems by optimizing the control input over a future period.

[0059] In a preferred embodiment, the MPC algorithm operates within a rolling optimization cycle (e.g., executed once every 2 seconds). Its core lies in establishing a multi-objective optimization function (CostFunction) that comprehensively considers multiple dimensions such as device safety, user comfort, continuous performance output, battery health and battery life, and user experience perception. This optimization function can be expressed as: .

[0060] and These are the predicted core temperature and the outer casing temperature, respectively.

[0061] and These are preset core safety upper limits (e.g., CPU 90°C, GPU 85°C) and user comfort case temperature upper limits (e.g., 45°C).

[0062] It quantifies performance metrics such as average CPU / GPU frequency and throughput within the future prediction time window.

[0063] Penalties are imposed based on the impact of predicted values ​​such as battery temperature and charging current on battery health.

[0064] Penalize user experience based on application type and predicted performance loss.

[0065] It is a penalty for smoothing control actions to avoid frequent and drastic policy adjustments.

[0066] , , , , , These are weighting coefficients that are dynamically adjusted based on the current system state and user preferences. For example, when the battery level is below 20%, The weight of [the system / mechanism] will increase significantly, prioritizing battery protection. This is especially important when running demanding games. The weight will increase, prioritizing performance.

[0067] The control variables (manipulated variables) of the MPC algorithm include: CPU frequency and voltage levels: For example, the available DVFS (Dynamic Voltage and Frequency Scaling) state set contains several predefined frequency-voltage pairs, ranging from the lowest power consumption state to the highest performance state.

[0068] GPU frequency and voltage levels: Similar to CPUs, GPUs also have a set of DVFS states.

[0069] Operating system-level task scheduling priorities: For example, adjusting the priority of foreground application processes to the highest level while lowering the priority of non-critical background tasks.

[0070] Resource limits for background applications: Set limits on CPU usage (e.g., no more than 5%), memory usage, or network bandwidth for background applications.

[0071] Battery charging current limit: A continuously adjustable range from the maximum charging current (e.g., 3A) to the stop charging (0A).

[0072] Display brightness: linear adjustment from 0% to 100%.

[0073] The MPC algorithm receives the predicted temperature sequence from the thermal load prediction module and the current system state (including current CPU / GPU load, battery SoC, ambient temperature, etc.) from the data processing and fusion unit within each control cycle. Then, it uses an efficient iterative optimization algorithm, such as Sequential Quadratic Programming (SQP) or the Interior-Point Method, to solve the multi-objective optimization function to calculate the optimal combination of control variables for a future control time domain (consistent with the prediction time domain length, e.g., 60 seconds). This optimization process typically runs on the dedicated DSP or NPU core of the main control chip, ensuring calculation is completed within 100ms, and transmitting the optimal combination of control variables to the thermal execution unit in the form of refined control instructions. This MPC-based strategy generation method enables the system to make decisions from a globally optimal perspective, rather than being limited to local greedy strategies, thereby achieving a dynamic and intelligent balance between performance, temperature, and user experience.

[0074] The heat dissipation execution unit is electrically connected to the adaptive control strategy generation module. Its responsibility is to receive and accurately execute the control commands issued by the adaptive control strategy generation module, thereby achieving fine-grained and dynamic control of the tablet's heat dissipation performance and system power consumption. This unit interacts with the tablet's underlying system through various hardware interfaces and software APIs.

[0075] First, the thermal execution unit dynamically adjusts the operating frequency and voltage of the central processing unit (CPU) and graphics processing unit (GPU) by directly calling the DVFS (Dynamic Voltage and Frequency Scaling) interface provided by the tablet's main control unit (SoC) firmware layer. This interface allows the system to precisely set the operating state of each CPU core and GPU with extremely high granularity (e.g., frequency step size of 10MHz, voltage step size of 5mV) and extremely fast response speed (nanosecond level). For example, when the strategy module determines that the future thermal load will increase, the execution unit can fine-tune the CPU's highest performance core frequency from 2.8GHz to 2.7GHz without affecting the user's perception, while correspondingly reducing its operating voltage, thereby significantly reducing heat generation. Conversely, when the thermal load allows and performance requirements increase, the frequency can be quickly increased to respond to performance requests. This fine-grained DVFS control is the core mechanism of active thermal management, directly controlling heat generation at its source.

[0076] Secondly, the heat dissipation execution unit dynamically intervenes in task scheduling and resource management through API interfaces provided by the operating system (OS) kernel. For example, in Android or Linux systems, the process priority of critical foreground applications can be dynamically increased through the sched_setattr() system call or by modifying cgroup parameters, ensuring that they receive sufficient CPU time slices and computing resources. Simultaneously, for non-critical background applications, the system can utilize the ActivityManager service or cgroup mechanism to strictly limit their CPU utilization, memory usage limits, and I / O throughput. For example, when the system predicts that the temperature is about to exceed the comfort zone, it can forcibly limit the CPU utilization of background synchronization services to below 2%, or suspend unnecessary background data downloads, thereby freeing up computing resources and heat capacity, and allocating more energy to the applications currently being used by the user.

[0077] Furthermore, the thermal execution unit communicates with the power management integrated circuit (PMIC) to dynamically adjust the battery charging current. The PMIC typically provides a series of register interfaces, allowing an external controller to finely set the maximum value of the charging current. For example, when the system predicts that the battery temperature or overall device temperature is about to rise, the execution unit can, according to instructions from the strategy module, reduce the charging current from the default 3A to 1.5A, or even temporarily stop charging in extreme cases, to minimize the heat generated by the battery during charging. This not only helps prevent battery overheating and extend battery life but also provides more heat capacity for other components.

[0078] Furthermore, the heat dissipation unit dynamically adjusts the backlight brightness of the display screen via the display controller interface provided by the operating system. The backlight module of the display screen is one of the significant power sources of a tablet computer. In certain specific scenarios, such as when the temperature of the grip area of ​​the tablet's casing approaches the user's comfort threshold (e.g., 42°C), the system can slightly reduce the screen brightness from 80% to 70% without being noticeably perceived by the user. This fine-tuning effectively reduces the power consumption and heat generation of the display screen while ensuring the user's visual experience, providing additional margin for overall heat dissipation.

[0079] It is worth emphasizing that the passive cooling structure of the tablet (including a high thermal conductivity graphene heat dissipation film, a vapor chamber, an embedded heat pipe, and high thermal conductivity and conductive silicone grease filling) and the intelligent regulation of the cooling execution unit are not independent of each other, but rather work collaboratively and complement each other. The passive cooling structure is responsible for rapidly and efficiently conducting the heat generated by the core chip from the heat source to the device casing, and dissipating it into the environment through radiation, convection, and other means. Meanwhile, the intelligent regulation of the cooling execution unit effectively reduces the instantaneous heat load demand on the passive cooling structure by controlling the heat generation rate at the source, ensuring that it always operates within its optimal efficiency range. This combination of active and passive cooling strategies forms a highly efficient and robust overall cooling solution.

[0080] The feedback and learning module is electrically connected to the heat dissipation execution unit and the data acquisition unit. Its key function is to continuously monitor the system's operating performance and, based on this monitoring, adaptively optimize the performance of the heat load prediction module and the adaptive control strategy generation module. This module is the core component for realizing the system's intelligence and self-learning capabilities.

[0081] Specifically, the feedback and learning module receives real-time actual temperature data, power consumption data, and system performance data (e.g., actual average CPU / GPU frequency, actual application frame rate, etc.) monitored by the data acquisition unit. It then precisely compares this actual data with the predicted temperature output by the heat load prediction module and calculates the prediction error (e.g., mean squared error (MSE) or mean absolute error (MAE)). Simultaneously, the module also compares the actual system performance output (e.g., the actual average frame rate or throughput of a specific application under the control strategy) with the target performance output set by the adaptive control strategy generation module during the optimization process, thereby evaluating the effectiveness and performance achievement of the current strategy.

[0082] In a preferred embodiment, the feedback and learning module periodically (e.g., every 24 hours or every 10 hours of cumulative operation) fine-tunes or incrementally updates some parameters of the LSTM model using the latest, timestamped, and filtered real-time running data. This online learning mechanism allows the LSTM model to better adapt to performance degradation that may occur during long-term device operation, decreased thermal conductivity due to aging of thermal interface materials, and changes in thermal behavior patterns brought about by user behavior patterns and the introduction of new applications. The update process is typically run in the background with low priority, or when the device is idle or charging, to avoid impacting the user experience. The update algorithm can employ stochastic gradient descent (SGD) or the Adam optimizer, and perform mini-batch training on the latest dataset to ensure that the model can capture new thermal behavior patterns in a timely manner.

[0083] Furthermore, the feedback and learning module dynamically adjusts the weight coefficients and constraints of the multi-objective optimization function in the MPC algorithm based on the optimization effect of actual feedback. For example, if the system continuously monitors that the temperature of the outer shell's grip area frequently reaches the upper limit of user comfort, even if the core component temperature is still within a safe range, the feedback and learning module will automatically increase the weight coefficient of w_T_skin, thereby prompting the MPC algorithm to prioritize the control of the outer shell surface temperature in subsequent decisions. Conversely, if the system performs poorly in a specific application scenario but has sufficient temperature margin, the weight of w_T_core may be appropriately reduced, and the weight of w_P increased, to allow the system to further improve performance within a safe range. This adaptive adjustment mechanism enables the system to continuously learn and improve its control strategy based on the actual operating environment, user preferences, and device aging status, thereby ensuring an optimal balance between performance, heat dissipation, and user experience in different scenarios, significantly improving the long-term stability and robustness of the system.

[0084] The safety protection module, as the core protection layer of the system, is designed independently of the intelligent temperature control algorithm, but works closely with it to ensure the absolute safety of the equipment and users under any extreme operating conditions. This module is directly electrically connected to the data acquisition unit and can communicate directly with the main control chip (SoC) of the heat dissipation execution unit through a hardware interrupt mechanism.

[0085] In a preferred embodiment, the safety protection module is equipped with dual over-temperature protection thresholds at both the hardware and software levels. At the hardware level, the system incorporates an independent hardware thermal switch or dedicated thermal protection circuit. When any critical temperature sensor (e.g., CPU core, battery internals, PMIC) detects a temperature exceeding the preset hardware-level absolute safety limit (e.g., CPU core temperature reaches 95°C, GPU core temperature reaches 90°C, battery pack internal temperature reaches 60°C), the hardware protection circuit immediately triggers a non-maskable interrupt (NMI) signal to the SoC. Upon receiving the NMI, the SoC will forcibly execute a pre-programmed emergency shutdown procedure, or reduce the CPU / GPU frequency to the minimum safe frequency (e.g., CPU to 400MHz, GPU to 100MHz), and cut off power to some high-power modules to prevent permanent hardware damage or safety incidents (such as battery thermal runaway). This process does not rely on any software response from the operating system or intelligent control algorithms and is the highest priority and most reliable protection mechanism.

[0086] At the software level, the security protection module includes an independent daemon running in the operating system kernel. This daemon continuously monitors the temperature data stream from the data acquisition unit. When the monitored temperature exceeds a preset software-level secondary safety limit (e.g., CPU core temperature reaches 90°C, battery temperature reaches 55°C, and the temperature of the casing's grip area reaches 48°C), the daemon immediately triggers a system-level warning and enforces a series of preset, stringent degradation measures via the kernel API. These measures include, but are not limited to: forcibly reducing the charging current to a minimum safe value (e.g., 0.5A) or pausing charging; significantly reducing the CPU and GPU operating frequencies to a low-to-medium performance state (e.g., CPU reduced to 1.2GHz, GPU reduced to 300MHz); closing non-critical background application processes; and displaying an over-temperature warning on the user interface to prompt the user to take action. These software-level protection measures aim to provide a buffer layer, preventing the system from entering an emergency shutdown state as much as possible through gentler but stronger intervention before hardware-level protection is triggered, thereby maintaining the basic availability of the device. The safety protection module is the final line of defense for the intelligent temperature control algorithm, ensuring that even in extreme cases where the prediction model fails or the control strategy deviates, the operational safety of the equipment can be fundamentally guaranteed.

[0087] To fully illustrate the technical solution of the present invention, the following detailed description is provided through embodiments and comparative examples.

[0088] Example 1: Performance of the System of the Invention In a typical application scenario, a tablet computer equipped with the intelligent temperature control algorithm of this invention runs a large-scale 3D open-world game (such as Genshin Impact, at the highest graphics settings) continuously for 60 minutes under an ambient temperature of 25°C. Before the game starts, all system parameters are in normal standby mode: CPU core temperature is approximately 32°C, GPU temperature is approximately 30°C, battery temperature is 28°C, and the temperature of the outer casing / grip area is 29°C. Power consumption monitoring shows that the instantaneous CPU power consumption is approximately 0.5W, and the instantaneous GPU power consumption is approximately 0.2W. After the game starts, the system detects a rapid increase in CPU and GPU load, with instantaneous power consumption reaching 8W and 12W respectively, and remaining at high levels. The data acquisition unit quickly transmits massive amounts of real-time data to the data processing and fusion unit, which performs time synchronization, filtering, normalization, and feature extraction to generate high-dimensional feature vectors. After receiving the feature vector, the thermal load prediction module, based on a pre-trained LSTM model, predicts that within the next 60 seconds, the CPU core temperature will rise from the current 35°C to 65°C, the GPU temperature from 33°C to 60°C, the battery temperature from 29°C to 38°C, and the casing grip area temperature from 30°C to 40°C. Upon receiving the prediction results, the adaptive control strategy generation module immediately initiates the MPC optimization algorithm. Considering the demands of sustained high-performance gaming, the weight (w_P) for maximizing performance in the MPC optimization function is increased, while ensuring a hard constraint that the core temperature does not exceed 90°C and the casing temperature does not exceed 45°C. Within the future prediction time domain, the MPC algorithm calculates the optimal control sequence: CPUDVFS: Maintain the large core frequency at 2.6GHz and the small core frequency at 1.8GHz, and slightly reduce the voltage by 0.05V.

[0089] GPUDVFS: Maintains the clock speed at 850MHz and dynamically fine-tunes the voltage based on the load.

[0090] Task scheduling: Ensure the game process has the highest priority and limit the total CPU usage of background applications to no more than 3%.

[0091] Charging Management: If the device is charging, the charging current will be reduced from 2A to 1.2A. If the device is not charging, this operation will not be performed.

[0092] Screen brightness: Maintains the user-set brightness. Upon receiving these control commands, the thermal execution unit immediately performs precise adjustment via the SoC interface, operating system API, and PMIC interface.

[0093] After 60 minutes of continuous gameplay, the system of this invention demonstrated excellent heat dissipation control: CPU average frequency: 2.58GHz, maintaining 92% of peak performance.

[0094] The average GPU frequency was 845MHz, maintaining 95% of its peak performance.

[0095] CPU maximum temperature: 78℃, far below the safe limit of 90℃.

[0096] Maximum battery temperature: 39℃, far below the 55℃ software safety limit.

[0097] The highest temperature in the grip area of ​​the outer shell is 42℃, which remains stable below the user comfort threshold of 45℃.

[0098] Average game frame rate: 58 FPS.

[0099] Frame rate stability (standard deviation): 2.5FPS, providing a smooth experience. During this process, users did not experience any noticeable throttling or stuttering, the surface temperature of the casing remained within a comfortable range, and the battery temperature was well controlled.

[0100] Comparative Example 1: Performance of Traditional Threshold Down System For comparison, another tablet with the same hardware configuration but equipped with a traditional static threshold-based frequency reduction cooling strategy was tested under the exact same environment and gaming scenarios. The traditional system only monitors temperature sensors, triggering a preset frequency reduction strategy when the temperature of any key hotspot reaches a preset threshold. The thresholds are set as follows: when the CPU core temperature reaches 75℃, the CPU frequency is immediately forced to drop to 1.8GHz; when the GPU core temperature reaches 70℃, the GPU frequency is forced to drop to 600MHz; and when the temperature of the outer casing's grip area reaches 43℃, the CPU / GPU frequencies are further reduced.

[0101] Approximately 15 minutes after the game started, the traditional system detected that the CPU core temperature reached 76℃ and the GPU core temperature reached 72℃. The system immediately triggered a frequency reduction strategy. CPU frequency: dropped sharply from 2.8GHz to 1.8GHz.

[0102] GPU frequency: drastically reduced from 890MHz to 600MHz. This throttling caused the game frame rate to drop sharply from 60FPS to around 40FPS within a short period, resulting in noticeable stuttering and performance loss for the user. Approximately 30 minutes later, due to continued high load operation, despite the throttling, the temperature of the grip area on the casing still reached 44℃, triggering an even stricter throttling strategy again. CPU frequency: further reduced to 1.2GHz.

[0103] GPU frequency: further reduced to 450MHz. At this point, the game frame rate drops to less than 30FPS, severely impacting the user experience and causing noticeable stuttering.

[0104] After 60 minutes of continuous gameplay, the traditional system performed as follows in terms of data: Average CPU frequency: 1.75GHz, only 62.5% of peak performance.

[0105] The average GPU frequency is 550MHz, which is only 62% of the peak performance.

[0106] The CPU's maximum temperature is 85℃, which is close to its safe limit.

[0107] Maximum battery temperature: 45℃.

[0108] The highest temperature in the grip area of ​​the shell is 44.5℃, which is close to the upper limit of user comfort.

[0109] Average game frame rate: 35 FPS.

[0110] Frame rate stability (standard deviation): 15.8 FPS, fluctuating wildly, resulting in a choppy experience.

[0111] The data comparison between the above embodiments and comparative examples clearly demonstrates the significant advantages of the tablet computer thermal management system based on the intelligent temperature control algorithm of this invention in providing excellent sustained performance output, ensuring user comfort, and effectively managing device thermal load. This invention, through forward-looking prediction and refined multi-objective optimization, avoids the performance waste and user experience degradation caused by traditional passive threshold frequency reduction strategies.

[0112] The table below summarizes the key performance indicators of the system of this invention and the traditional threshold down-conversion system in the above test scenario: index The system of the present invention (Example 1) Traditional threshold downsampling system (Comparative Example 1) Improve / Enhance Average CPU frequency (GHz) 2.58 1.75 +47.4% Average GPU frequency (MHz) 845 550 +53.6% Average frame rate (FPS) in the game 58 35 +65.7% Frame rate stability (FPS standard deviation) 2.5 15.8 Reduced by 84.2% CPU maximum temperature (°C) 78 85 -7℃ Maximum battery temperature (°C) 39 45 -6℃ Maximum temperature in the grip area of ​​the outer casing (°C) 42 44.5 -2.5℃ User-perceived performance interruption count 0 2 Avoid interruption The data above clearly shows that the system of this invention achieves significant improvements in the average operating frequency of both the CPU and GPU, directly translating to a 65.7% increase in the average frame rate of games. More importantly, frame rate stability is significantly improved, with the standard deviation reduced by 84.2%, meaning a substantial improvement in the smoothness experienced by users in actual gameplay, with virtually no stuttering. Simultaneously, while maintaining higher performance output, the system of this invention successfully keeps the peak temperatures of the CPU, battery, and user-held area at lower levels, ensuring device safety and user comfort. Traditional systems, due to their passive response mechanisms, while ultimately avoiding overheating damage, do so at the cost of significant performance sacrifice and severely compromised user experience.

[0113] In summary, the tablet PC heat dissipation control system based on intelligent temperature control algorithms provided by this invention, through its sophisticated sensor network, efficient data processing and fusion, forward-looking deep learning prediction, intelligent MPC strategy generation, and multi-dimensional heat dissipation execution mechanism, supplemented by self-learning feedback and independent safety protection, successfully solves the fundamental problem of balancing performance, heat dissipation, and user experience in existing technologies. This system is not merely simple temperature control, but a globally optimized, continuously learning intelligent management paradigm, laying a solid foundation for the high-performance and highly reliable operation of tablet PCs in future complex application scenarios.

[0114] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely illustrative of the principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the present invention as claimed. The scope of protection of the present invention is defined by the appended claims and their equivalents.

Claims

1. A tablet computer heat dissipation control system based on an intelligent temperature control algorithm, characterized in that, The application relates to a tablet computer thermal management system, which comprises the following: a data acquisition unit arranged in a tablet computer and used for acquiring multi-dimensional running data inside and outside the tablet computer in real time; a data processing and fusion unit electrically connected with the data acquisition unit and realized by using a special coprocessor embedded in a tablet computer main control chip or an independent low-power microcontroller, which is used for receiving and processing original data from the data acquisition unit, the processing including time synchronization, data filtering and denoising, data normalization and feature extraction, so as to form a multi-dimensional feature vector; a thermal load prediction module electrically connected with the data processing and fusion unit, which is mainly a deep learning model based on a long short-term memory network and is used for making a forward-looking prediction on temperature variation trends of each key thermal point of the tablet computer in a future period of time based on the feature vector, the model being trained offline through large-scale historical running data and then being deployed into a tablet computer memory to perform real-time reasoning; an adaptive control strategy generation module electrically connected with the thermal load prediction module, which is mainly an optimization algorithm based on model predictive control and is used for dynamically generating an optimal heat dissipation control strategy in a rolling optimization period based on future temperature prediction results and multi-dimensional running data, the strategy being realized by solving a multi-objective optimization function which comprehensively considers core temperature constraints, shell surface temperature constraints, continuous performance output, battery health and endurance and user experience perception; a heat dissipation execution unit electrically connected with the adaptive control strategy generation module and used for receiving and executing control instructions, so as to realize fine dynamic regulation and control of heat dissipation performance and system power consumption by controlling processor frequency and voltage level, operating system task scheduling priority, background application resource limitation, battery charging current limitation and display screen brightness; a feedback and learning module electrically connected with the heat dissipation execution unit and the data acquisition unit respectively, which is used for receiving actual running data, comparing the actual running data with prediction results and target output, periodically updating deep learning model parameters and adjusting weight coefficients and constraint conditions of a multi-objective optimization function, so as to realize system adaptive learning and continuous optimization; a safety protection module electrically connected with the data acquisition unit and directly communicating with a main control chip of the heat dissipation execution unit, which sets hardware-level and software-level double over-temperature protection thresholds and is used for triggering a forced safety protection measure immediately when the temperature exceeds a preset threshold.

2. The tablet computer heat dissipation control system based on intelligent temperature control algorithm according to claim 1, characterized in that, The data acquisition unit comprises: a plurality of temperature sensors which communicate with the data processing and fusion unit through a bus protocol, monitor temperature variation in real time and send data; a power consumption monitoring unit integrated in a power management integrated circuit and composed of a high-precision current sensor and a voltage sensor, which is used for monitoring instantaneous power consumption data of main power consumption components in real time and transmitting the data through an interface, wherein the power consumption data includes actual working current and voltage of each component; an environment sensor integrated in a machine body and communicating with the data processing and fusion unit through a bus protocol, which is used for monitoring actual temperature of an environment. The system state monitoring interface obtains real-time running state data of the tablet computer through operating system standard API or kernel level interface, including processor utilization, main frequency and voltage, memory usage, memory read / write speed, running application identification and resource occupation, screen brightness setting, battery charging state, health state and charging current, and pushes the data to the data processing and fusion unit at a preset frequency.

3. The tablet computer heat dissipation control system based on intelligent temperature control algorithm according to claim 2, characterized in that, The multiple temperature sensors constitute a distributed temperature sensor network, and the thermistors are arranged in the micro-gap of the processor core heat dissipation path, the surface of the power management integrated circuit, and between the battery cells.

4. The tablet heat dissipation control system based on intelligent temperature control algorithm of claim 1, wherein, The data processing and fusion unit performs the following processing on the multi-dimensional raw data: Time synchronization, a strategy combining hardware timestamp and software timestamp is adopted, data is aligned according to timestamp, and time series is aligned through interpolation method; Data filtering and denoising, Kalman filter is used to process temperature sensor data, and exponential weighted moving average filter is used to process power consumption data, and the smoothing factor of the filter is dynamically adjusted according to the data characteristics; Data normalization and feature extraction, different dimensional data are normalized, key features are extracted, the features are organized into fixed length feature vectors and output.

5. The smart temperature control algorithm based tablet heat dissipation control system of claim 4, wherein, The deep learning model of the heat load prediction module is a long short-term memory network with a multi-layer stacking structure; the model is trained by multi-dimensional historical running data under different environmental temperatures and different load types in the offline stage, and is trained by a back propagation algorithm and an optimizer to minimize the mean square error between the predicted temperature and the actual temperature; The model input is a multi-dimensional feature vector, and the output is a predicted temperature value sequence of each key hotspot in the next N time steps, and after the training is completed, the model is deployed to the memory of the tablet computer, and receives real-time data for forward inference during running.

6. The tablet heat dissipation control system based on intelligent temperature control algorithm of claim 1, wherein, The optimization algorithm of the adaptive control strategy generation module works in a rolling optimization period, and the target is to maximize the continuous performance output under the premise of meeting the device safety and user comfort constraints; the algorithm establishes a multi-objective optimization function, and comprehensively considers the core temperature constraint, the shell surface temperature constraint, the maximum continuous performance output, the battery health and endurance, and the user experience perception; in each control period, the future predicted temperature sequence and the current system state are used to solve the multi-objective optimization function through an iterative optimization algorithm to obtain the optimal control variable combination in the future control time domain.

7. The smart temperature control algorithm based tablet heat dissipation control system of claim 6, wherein, The control variables specifically include: Processor frequency and voltage level, realized through a dynamic voltage frequency adjustment mechanism; Operating system task scheduling priority, used to adjust the priority of foreground key application processes to the highest and reduce the priority of non-key background tasks; Background application resource limitation, used to set the upper limit of processor occupancy, the upper limit of memory usage or network bandwidth limitation of background applications; Battery charging current limitation, a continuous adjustable range from maximum charging current to stop charging.

8. The tablet heat dissipation control system based on intelligent temperature control algorithm of claim 1, wherein, The heat dissipation execution unit is mainly realized by the following ways: Processor dynamic voltage frequency adjustment control, the frequencies and voltages of each core of the processor are adjusted to the optimal set values by calling the firmware layer interface of the main control unit; Task scheduling and resource management dynamically adjust the priority of processes or threads through the operating system kernel API interface, and limit the resource occupation of non-critical background applications; Battery charging management dynamically adjusts the charging current by communicating with the power management integrated circuit; The passive cooling structure of the tablet computer cooperates with the heat dissipation execution unit, and the passive cooling structure includes a graphene heat dissipation film, a heat spreading plate, a heat pipe and a high thermal conductivity material filling, thereby enhancing the heat conduction efficiency.

9. The tablet heat dissipation control system based on intelligent temperature control algorithm of claim 1, wherein, The feedback and learning module is configured to: Real-time receive actual temperature data, power consumption data and system performance data monitored by the data acquisition unit; Calculate the prediction error of the actual data and the predicted temperature, and compare the actual performance output with the target performance output; Periodically update the deep learning model parameters using real-time running data; Adjust the weight coefficients and constraint conditions of the multi-objective optimization function according to the optimization effect.

10. The tablet heat dissipation control system based on intelligent temperature control algorithm of claim 1, wherein, The safety protection module includes: A hardware-level over-temperature protection mechanism, which sets up an independent hardware thermal switch or a special thermal protection circuit, monitors when the temperature exceeds the hardware-level absolute safety upper limit, triggers a non-maskable interrupt signal to the main control chip, forces the heat dissipation execution unit to reduce the processor frequency and triggers an emergency shutdown process; A software-level over-temperature protection mechanism, which contains an operating system kernel state independent daemon process, monitors when the temperature exceeds the software-level secondary safety upper limit, triggers a system-level warning, and forces the implementation of measures such as reducing the charging current, reducing the processor frequency, and shutting down the background application.

Images