Intelligent clothing power supply control method and system based on temperature feedback

Abstract

The invention relates to the field of intelligent clothing power supply control, in particular to an intelligent clothing power supply control method and system based on temperature feedback. Real-time collected data in a multi-mode sensor are obtained, dynamic baseline calibration is conducted on the real-time collected data, then feature vectors are constructed, fusion calculation is conducted on the human body temperature according to a Pennes biological heat equation, calculation is conducted on the human body initial temperature based on a dynamic heat balance equation, and the system active regulation and control heat flux is obtained. Establishing an intelligent temperature prediction model; and inputting the characteristic data set and the system active regulation and control heat flux into a temperature intelligent prediction model for prediction, and carrying out temperature regulation and control on a power supply control system of the intelligent clothes according to a multi-step temperature prediction result. The precision and efficiency of temperature prediction are significantly improved, and accurate pre-judgment of multi-step temperature is realized.

Description

Technical Field

[0001] The present invention relates to the field of intelligent clothing power control, and particularly to an intelligent clothing power control method and system based on temperature feedback. Background Art

[0002] Intelligent clothing power control has become an important part of enterprise competitiveness. However, traditional intelligent clothing power control often relies on manual experience and rules, is greatly affected by subjective factors, and is difficult to meet the increasingly complex supply chain requirements. With the rapid development of information technology, especially the wide application of machine learning technology, and the expansion of intelligent clothing application scenarios, the precise temperature control of its power control system has become a technical focus. Traditional temperature control schemes have multiple limitations, have an unclear perception of human body temperature, have inaccurate temperature control, and cannot perform temperature control in advance when the human body is exercising. Therefore, how to improve the accuracy and precision in intelligent clothing power control is a technical problem that needs to be solved urgently at present. Summary of the Invention

[0003] The purpose of the present invention is to solve the above problems and design an intelligent clothing power control method and system based on temperature feedback.

[0004] In the intelligent clothing power control method based on temperature feedback, the intelligent clothing power control method includes the following steps:

[0005] Obtain the real-time acquisition data in the multi-modal sensor, perform dynamic baseline calibration on the real-time acquisition data, and construct a feature vector to obtain a feature data set;

[0006] Perform fusion calculation on the human body temperature according to the Pennes bioheat equation to obtain the initial human body temperature, and calculate the system's active regulation heat flux based on the dynamic heat balance equation;

[0007] Establish a CNN-LSTM hybrid model neural network model, train the hybrid model using the training data set, and optimize the model parameters through an improved IGWO gray wolf algorithm to obtain an IGWO-CNN-LSTM intelligent temperature prediction model;

[0008] Input the feature data set and the system's active regulation heat flux into the IGWO-CNN-LSTM intelligent temperature prediction model for prediction to obtain a multi-step temperature prediction result;

[0009] Perform temperature control on the power control system of the intelligent clothing according to the multi-step temperature prediction result.

[0010] Further, in the above-mentioned intelligent clothing power control method based on temperature feedback, the steps of acquiring real-time acquisition data from distributed temperature sensors and physiological sensors, performing dynamic baseline calibration on the real-time acquisition data, and constructing a feature vector to obtain a feature dataset include:

[0011] The multi-modal sensors at least include an infrared temperature sensor, an environmental temperature and humidity sensor, a six-axis accelerometer, and a pressure sensor;

[0012] Calibrate the real-time acquisition data using an adaptive calibration algorithm, update the baseline parameters every 500 ms, and obtain calibrated acquisition data;

[0013] Construct a multi-dimensional feature vector from the calibrated acquisition data to obtain a feature dataset; the multi-dimensional features at least include the average skin temperature at multiple points on the forehead, armpit, and neck, environmental temperature, environmental humidity, three-dimensional acceleration vector, metabolic equivalent based on acceleration, and posture recognition.

[0014] Further, in the above-mentioned intelligent clothing power control method based on temperature feedback, the steps of performing fusion calculation on the human body temperature according to the Pennes bioheat equation to obtain the initial human body temperature, and calculating the system's active regulation of heat flux based on the dynamic heat balance equation include:

[0015] Obtain the initial human body temperature data through the chest and back thermocouples, and estimate the initial human body temperature data based on the Pennes bioheat equation to obtain the human core temperature T core , where the core temperature conduction coefficient α of the human skin is 0.8 W / m·k;

[0016] Introduce the human body motion state weight, where the infrared weight is 0.7 when the human body is stationary and dynamically adjusted to 0.4 during motion, and perform weighted fusion on the human core temperature T core ;

[0017] Establish a dynamic heat balance equation:

[0018]

[0019] where h represents the convection coefficient, A represents the effective heat dissipation area, ∈ represents the emissivity, σ represents the Stefan-Boltzmann constant, T target represents the target temperature, represents the human skin temperature, represents the human body environmental temperature, Q active is the system's active regulation of heat flux, representing the magnitude of the heat flow involved in the active regulation process of the system;

[0020] Calculating the initial body temperature based on the dynamic heat balance equation to obtain the system's active regulation of heat flux.

[0021] Further, in the above intelligent clothing power control method based on temperature feedback, establishing the CNN-LSTM hybrid model neural network model and training the hybrid model using the training dataset includes:

[0022] Establishing the CNN-LSTM hybrid model neural network model, where the CNN feature extraction module is a 2D convolutional layer. The first layer has 16 3×1 convolutional kernels for extracting local time patterns, and the second layer has 32 5×1 convolutional kernels for capturing larger time range features;

[0023] The LSTM time series modeling module is a forward + backward LSTM memory network for capturing long-term dependencies, with a hidden layer dimension of 64 units to balance model capacity and complexity; the attention mechanism is a time attention layer for weighted focusing on key time points;

[0024] Normalizing the temperature features, humidity features, acceleration features, and wind speed features in the training dataset, and after time series reconstruction of the normalized data, inputting it into the CNN-LSTM hybrid model neural network model;

[0025] The loss function of the model is a composite loss function, where L = 0.7MAE + 0.3MAPE. The initial value of the model's adaptive learning rate is set to 0.001, which decays by 20% every 5 epochs, and the gradient clipping threshold of the model is ±0.5.

[0026] Further, in the above intelligent clothing power control method based on temperature feedback, optimizing the model parameters through the improved IGWO gray wolf algorithm to obtain the IGWO-CNN-LSTM temperature intelligent prediction model includes:

[0027] Optimizing the non-linear convergence factor of the GWO gray wolf algorithm, and its calculation formula is as follows:

[0028]

[0029] Among them, a(t) represents the non-linear convergence factor, t represents the current iteration number, T represents the maximum iteration number, and the cubic decay accelerates the convergence speed in the later stage;

[0030] Updating the dynamic weight position of the GWO gray wolf algorithm, and its calculation formula is as follows:

[0031]

[0032] Among them, represents the position of the new generation of gray wolf individuals, Indicates the position of the individual with the highest fitness, Indicates the position of the individual with the second highest fitness, Indicates the position of the individual with the third highest fitness, ω1 represents the weight of the highest fitness, ω2 represents the weight of the second highest fitness, ω3 represents the weight of the third highest fitness, and α, β, δ represent the positions of the wolves;

[0033] Weight calculation:

[0034]

[0035] Among them, Represents the fitness function value, which is used to evaluate the quality of individuals, Represents the predicted positions of each leading wolf;

[0036] Obtain the gray wolf position vector after position update, map the parameters of the CNN-LSTM hybrid model neural network model to the gray wolf position vector, and obtain the IGWO-CNN-LSTM temperature intelligent prediction model.

[0037] Furthermore, in the above intelligent clothing power control method based on temperature feedback, the step of inputting the feature data set and the system's active regulation of heat flux into the IGWO-CNN-LSTM temperature intelligent prediction model for prediction to obtain multi-step temperature prediction results includes:

[0038] Obtain the feature data set and the system's active regulation of heat flux, divide the feature data set into time series sample data according to a sliding window, the length of the sliding window is 60s, the step size is 1s, and normalize the time series sample data to the interval [0,1];

[0039] Take the system's active regulation of heat flux as an external covariate and splice it with the feature data to form a model input tensor;

[0040] Use the Seq2Seq structure encoder-decoder architecture, and the decoder iteratively outputs the predicted values for the next 5 steps in an autoregressive manner to obtain multi-step temperature prediction results.

[0041] Furthermore, in the above intelligent clothing power control method based on temperature feedback, the step of performing temperature regulation on the power control system of the intelligent clothing according to the multi-step temperature prediction results includes:

[0042] Preset three-level temperature states, including cold state, hot state, and normal state, and judge the temperature of the intelligent clothing according to the multi-step temperature prediction results;

[0043] When it is judged to be in the cold state, start the heating module of the intelligent clothing, and the power supply priority of the heating module of the intelligent clothing is: thermal energy conversion to electrical energy > solar energy > rechargeable battery;

[0044] When it is judged to be in the hot state, start the thermoelectric refrigeration module of the intelligent clothing, and at the same time convert the body surface waste heat into electric energy for storage;

[0045] When it is judged to be in the normal state, turn off the active temperature control module of the intelligent clothing, only maintain the power supply of the sensor, and start the thermoelectric module to recover the ambient temperature difference electric energy.

[0046] To achieve the above object, the technical solution of the present invention is, further, in the intelligent clothing power control system based on temperature feedback, the intelligent clothing power control system includes:

[0047] A data acquisition module, which is used to acquire real-time acquisition data in the multi-modal sensor, perform dynamic baseline calibration on the real-time acquisition data, construct a feature vector, and obtain a feature data set;

[0048] A heat flux acquisition module, which is used to perform fusion calculation on the human body temperature according to the Pennes bioheat equation to obtain the initial human body temperature, and calculate the system's active regulation heat flux based on the dynamic heat balance equation;

[0049] A model establishment module, which is used to establish a CNN-LSTM hybrid model neural network model, train the hybrid model using the training data set, and optimize the model parameters through an improved IGWO gray wolf algorithm to obtain an IGWO-CNN-LSTM temperature intelligent prediction model;

[0050] A result prediction module, which is used to input the feature data set and the system's active regulation heat flux into the IGWO-CNN-LSTM temperature intelligent prediction model for prediction to obtain a multi-step temperature prediction result;

[0051] A temperature regulation module, which is used to perform temperature regulation on the power control system of the intelligent clothing according to the multi-step temperature prediction result.

[0052] Further, in the above-mentioned intelligent clothing power control system based on temperature feedback, the heat flux acquisition module includes the following units:

[0053] A data acquisition unit, which is used to obtain the initial human body temperature data through the chest and back thermocouples, and estimate the initial human body temperature data according to the Pennes bioheat equation to obtain the human core temperature T core , where the core temperature conduction coefficient α of the human skin is 0.8 W / m·k;

[0054] A weighted fusion unit, which is used to introduce the human body motion state weight, where the infrared weight is 0.7 when the human body is stationary and dynamically adjusted to 0.4 by the IMU when the human body is moving, and perform weighted fusion on the human core temperature T core ;

[0055] A heat balance calculation unit for establishing a dynamic heat balance equation:

[0056]

[0057] where h represents the convection coefficient, A represents the effective heat dissipation area, ∈ represents the emissivity, σ represents the Stefan-Boltzmann constant, T target represents the target temperature, represents the human skin temperature, represents the human environmental temperature, Q active is the active regulation heat flux of the system, indicating the magnitude of the heat flow involved in the active regulation process of the system;

[0058] A heat flux obtaining unit for calculating the initial human temperature based on the dynamic heat balance equation to obtain the active regulation heat flux of the system.

[0059] Furthermore, in the above intelligent clothing power control system based on temperature feedback, it is characterized in that the temperature regulation module includes the following units:

[0060] A temperature state judgment unit for presetting three-level temperature states, including cold state, hot state and normal state, and judging the temperature of the intelligent clothing according to the multi-step temperature prediction result;

[0061] A cold state execution unit for starting the heating module of the intelligent clothing when it is judged to be in the cold state, and the power supply priority of the heating module of the intelligent clothing is: thermal energy conversion to electrical energy > solar energy > rechargeable battery;

[0062] A hot state execution unit for starting the thermoelectric refrigeration module of the intelligent clothing when it is judged to be in the hot state, and simultaneously converting the body surface waste heat into electrical energy for storage;

[0063] A normal state execution unit for turning off the active temperature control module of the intelligent clothing when it is judged to be in the normal state, only maintaining the power supply of the sensor, and starting the thermoelectric module to recover the environmental temperature difference electrical energy.

[0064] Its beneficial effects are as follows. By collecting real-time data through multi-modal sensors and performing dynamic baseline calibration, a precise feature dataset is constructed, effectively improving the accuracy of raw data processing and providing a reliable data basis for subsequent temperature analysis. Secondly, by integrating the Pennes bioheat equation and the dynamic heat balance equation, the initial human body temperature and the system's active regulation of heat flux are accurately calculated, realizing the scientific nature of temperature calculation and meeting the actual needs of the human body from the perspective of bioheat physics. Furthermore, based on the CNN-LSTM hybrid model optimized by the IGWO grey wolf algorithm, which not only uses CNN to extract spatial features of data and LSTM to process temporal features, but also optimizes the model parameters through an improved algorithm, significantly improving the accuracy and efficiency of temperature prediction and achieving precise prediction of multi-step temperature. Finally, applying the prediction results to the temperature control of the power control system forms a closed-loop system of "data collection - calculation and analysis - precise prediction - intelligent regulation", enabling the intelligent clothing power system to more accurately adapt to the human body temperature needs, improving the temperature control comfort, energy utilization efficiency, and usage safety of wearable devices. Description of the Drawings

[0065] By reading the detailed description of the preferred embodiments below, various other advantages and benefits will become clear to those of ordinary skill in the art. The drawings are only for the purpose of showing the preferred embodiments and are not considered to be a limitation of the present invention.

[0066] Figure 1 Schematic diagram of the first embodiment of the intelligent clothing power control method based on temperature feedback in the embodiments of the present invention;

[0067] Figure 2 Schematic diagram of the second embodiment of the intelligent clothing power control method based on temperature feedback in the embodiments of the present invention;

[0068] Figure 3 Schematic diagram of the third embodiment of the intelligent clothing power control method based on temperature feedback in the embodiments of the present invention;

[0069] Figure 4 Schematic diagram of the first embodiment of the intelligent clothing power control system based on temperature feedback in the embodiments of the present invention. Detailed Embodiments

[0070] In order to make the objectives, technical solutions, and advantages of the present invention clearer, the following further describes the present invention in detail with reference to the drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0071] Those skilled in the art can understand that, unless specifically stated otherwise, the singular forms "a", "an", "the", and "said" used herein may also include the plural forms. It should be further understood that the term "comprising" used in the specification of the present invention means the presence of the stated features, integers, steps, operations, elements, and / or components, but does not exclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and / or their groups.

[0072] The present invention will be specifically described below with reference to the accompanying drawings. As Figure 1 shown, for the intelligent clothing power control method based on temperature feedback, the intelligent clothing power control method includes the following steps:

[0073] Step 101: Obtain the real-time acquisition data in the multi-modal sensor, perform dynamic baseline calibration on the real-time acquisition data, and construct a feature vector to obtain a feature data set;

[0074] Specifically, in this embodiment, the multi-modal sensor at least includes an infrared temperature sensor, an environmental temperature and humidity sensor, a six-axis accelerometer, and a pressure sensor; the adaptive calibration algorithm is used to calibrate the real-time acquisition data, and the baseline parameters are updated every 500 ms to obtain the calibrated acquisition data; a multi-dimensional feature vector in the calibrated acquisition data is constructed to obtain a feature data set; the multi-dimensional features at least include the mean values of the skin temperatures at multiple points on the forehead, under the armpit, and on the neck, the environmental temperature, the environmental humidity, the three-dimensional acceleration vector, the metabolic equivalent based on acceleration, and posture recognition.

[0075] Sensor array design: Integrate a skin temperature sensor (accuracy ±0.1 °C), an environmental temperature and humidity sensor (accuracy ±2% RH), a pressure sensor (range 0 - 10 kPa), and an accelerometer (3-axis dynamic monitoring); Dynamic baseline calibration algorithm: Use an adaptive sliding window (window size 50 ms) to calculate the baseline drift in real time, combine with Kalman filtering to eliminate motion artifacts, and separate physiological signals and environmental noise through wavelet transform; Feature engineering: Extract time-domain features (mean value, variance, peak-to-peak value), frequency-domain features (power spectral density), and time-frequency features (Gabor transform coefficients) to construct a 12-dimensional feature vector

[0076] Multi-modal Sensor Data Acquisition: Smart clothing is equipped with various types of sensors. For example, temperature sensors are used to sense the temperature of the environment and the parts close to the human body, humidity sensors monitor the environmental humidity conditions, and acceleration sensors capture the human body's motion state, etc. According to their respective working principles, the sensors collect real-time data on the surrounding environment or the human body state at a specific frequency. For example, thermopile-type temperature sensors convert the temperature difference into an electrical signal output through the Seebeck effect; capacitive humidity sensors obtain humidity data based on the change in capacitance caused by environmental humidity changes. The data acquisition frequencies of different sensors vary according to requirements. For example, the temperature sensor may collect data multiple times per second, while the acceleration sensor collects data at a higher frequency when capturing intense movements.

[0077] Dynamic Baseline Calibration: Since sensors are subject to various factors of interference during actual use, such as environmental temperature drift, performance changes caused by long-term use, etc., there are deviations in the collected data. Dynamic baseline calibration is to monitor these interference factors in real time and correct the original collected data by establishing a mathematical model. For example, for temperature sensors, using a reference temperature source, regularly measure and compare the reference temperature with the sensor measurement value, construct a calibration curve based on the difference between the two, and adjust the subsequent collected temperature data in real time. Through dynamic calibration, the accuracy and stability of the collected data are ensured.

[0078] Step 102: Perform fusion calculation on the human body temperature according to the Pennes bioheat equation to obtain the initial human body temperature, and calculate the system's active regulation heat flux based on the dynamic heat balance equation.

[0079] Specifically, in this embodiment, the initial human body temperature data is obtained through the chest and back thermocouples, and the initial human body temperature data is estimated based on the Pennes bioheat equation to obtain the human core temperature T core , where the core temperature conduction coefficient α of the human skin is 0.8 W / m·k;

[0080] Introduce the weight of the human body motion state, where the infrared weight is 0.7 when the human body is stationary and dynamically adjusted to 0.4 during motion by the IMU, and perform weighted fusion on the human core temperature T core ;

[0081] Establish a dynamic heat balance equation:

[0082]

[0083] where h represents the convection coefficient, A represents the effective heat dissipation area, ∈ represents the emissivity, σ represents the Stefan-Boltzmann constant, T tar get represents the target temperature, represents the human skin temperature, represents the human body environmental temperature, Q activeThe active regulation of heat flux by the system represents the magnitude of the heat flow involved in the active regulation process of the system;

[0084] Based on the dynamic heat balance equation, the initial human body temperature is calculated to obtain the active regulation of heat flux by the system.

[0085] Multi-modal sensor data acquisition: Deploy a micro-thermocouple array (such as T-type or K-type thermocouples) in the chest and back areas of the smart clothing, and use the Seebeck effect to collect skin surface temperature data in real time. The thermocouple converts the temperature difference signal into an electrical signal through the linear relationship between the thermoelectric potential and temperature (such as the sensitivity of the T-type thermocouple is about 40 μV / °C). To ensure measurement accuracy, a differential amplifier circuit (such as the AD620 instrumentation amplifier) is used to condition the microvolt-level signal, and a ΔΣ analog-to-digital converter (such as the ADS1256) is used to achieve 24-bit high-precision digitization.

[0086] Parameter acquisition: Estimate the blood perfusion rate through simultaneously collected heart rate data (using photoplethysmography PPG), and combine the body surface temperature distribution to inversely infer the spatial distribution of tissue thermal conductivity k. Numerical solution method: Discretize the human torso model using the finite element method, use the chest and back thermocouple data as boundary conditions, and solve it in combination with the Dirichlet boundary condition (T = measured value) and the Neumann boundary condition. Use implicit time integration to handle transient terms to ensure the stability of the time step. Dynamic correction: Introduce the extended Kalman filter (EKF) to update the model parameters online to compensate for individual differences and environmental changes. For example, when a change in the motion state is detected (through an acceleration sensor), the estimated value of the metabolic heat production rate is automatically adjusted.

[0087] Multi-source data fusion: Spatiotemporally fuse the core temperature estimated by the Pennes equation with the body surface temperatures at multiple locations (forehead, armpit, etc.). Adopt a spatiotemporal Bayesian network model, and combine the anatomical parameters of human heat conduction (such as the fat layer thickness, blood vessel distribution density) to construct a personalized temperature field model. For example, use the deep temperature data of the chest and back (verified by implantable temperature sensors) to optimize the model weight matrix and improve the core temperature estimation accuracy to within ±0.3°C.

[0088] Step 103: Establish a CNN-LSTM hybrid model neural network model, train the hybrid model using the training dataset, and optimize the model parameters through an improved IGWO grey wolf algorithm to obtain an IGWO-CNN-LSTM intelligent temperature prediction model;

[0089] Specifically, in this embodiment, a CNN-LSTM hybrid model neural network model is established, where the CNN feature extraction module is a 2D convolutional layer. The first layer has 16 3×1 convolutional kernels for extracting local time patterns, and the second layer has 32 5×1 convolutional kernels for capturing larger time range features;

[0090] The LSTM time series modeling module is a forward + backward LSTM memory network, which is used to capture long-term dependencies. The dimension of the hidden layer is 64 units, which is used to balance the model capacity and complexity; the attention mechanism is a time attention layer, which is used to weight and focus on key time points;

[0091] Normalize the temperature features, humidity features, acceleration features and wind speed features in the training dataset, and input the normalized data into the CNN-LSTM hybrid model neural network model after time series reconstruction;

[0092] The loss function of the model is a composite loss function, where L = 0.7MAE + 0.3MAPE. The initial value of the adaptive learning rate of the model is set to 0.001, and it decays by 20% every 5 steps. The gradient clipping threshold of the model is ±0.5.

[0093] Step 104: Input the feature dataset and the system's active regulation of heat flux into the IGWO-CNN-LSTM temperature intelligent prediction model for prediction to obtain multi-step temperature prediction results;

[0094] Specifically, in this embodiment, the nonlinear convergence factor of the optimized GWO gray wolf algorithm is calculated as follows:

[0095]

[0096] where a(t) represents the nonlinear convergence factor, t represents the current iteration number, T represents the maximum iteration number, and the cubic decay accelerates the convergence speed in the later stage;

[0097] Update the dynamic weight position of the GWO gray wolf algorithm, and its calculation formula is as follows:

[0098]

[0099] where, represents the position of the new generation of gray wolf individuals, represents the position of the individual with the highest fitness, represents the position of the individual with the second highest fitness, represents the position of the individual with the third highest fitness, ω1 represents the weight of the individual with the highest fitness, ω2 represents the weight of the individual with the second highest fitness, ω3 represents the weight of the individual with the third highest fitness, and α, β, δ represent the positions of the wolves;

[0100] Weight calculation:

[0101]

[0102] where, represents the fitness function value, which is used to evaluate the quality of individuals, Indicates the predicted positions of the leading wolves;

[0103] Obtain the position vector of the gray wolf after position update, map the parameters of the CNN-LSTM hybrid model neural network model to the position vector of the gray wolf, and obtain the IGWO-CNN-LSTM temperature intelligent prediction model.

[0104] Obtain the feature dataset and the system's active regulation of heat flux, divide the feature dataset into time-series sample data according to a sliding window, the length of the sliding window is 60s, the step size is 1s, and normalize the time-series sample data to the interval [0, 1];

[0105] Use the system's active regulation of heat flux as an external covariate to splice with the feature data to form a model input tensor;

[0106] Use the Seq2Seq structure encoder-decoder architecture, and the decoder iteratively outputs the predicted values for the next 5 steps in an autoregressive manner to obtain the multi-step temperature prediction result.

[0107] Step 105: Perform temperature regulation on the power control system of the intelligent clothing according to the multi-step temperature prediction result.

[0108] Specifically, in this embodiment, three-level temperature states are preset, including cold state, hot state, and normal state, and the temperature of the intelligent clothing is judged according to the multi-step temperature prediction result; when it is judged as the cold state, the heating module of the intelligent clothing is started, and the power supply priority of the heating module of the intelligent clothing is: thermal energy conversion to electrical energy > solar energy > rechargeable battery; when it is judged as the hot state, the thermoelectric refrigeration module of the intelligent clothing is started, and at the same time, the body surface waste heat is converted into electrical energy for storage; when it is judged as the normal state, the active temperature control module of the intelligent clothing is turned off, only the sensor power supply is maintained, and the thermoelectric module is started to recover the environmental temperature difference electrical energy.

[0109] Use a three-level temperature state intelligent decision-making mechanism to achieve accurate state judgment through an integrated multi-dimensional environmental perception module. The cold state determination is based on a dynamic temperature threshold algorithm, which is triggered when the body surface temperature continuously drops below the set value (such as 18°C ± 2°C) and the environmental humidity is higher than 60% RH, and the wind chill index is corrected by combining the data of the wind speed sensor. The hot state determination is based on the dual threshold conditions of analyzing the change rate of the body surface temperature gradient (ΔT / Δt ≥ 0.3°C / min) and the environmental temperature (≥ 28°C), and the infrared thermal imaging sensor is used to capture local overheating areas.

[0110] The energy management system adopts a three - level power supply priority strategy: First, activate the thermal energy conversion module, which converts the temperature difference between the human body and the environment into electrical energy through a thermoelectric generator (Seebeck effect). This module can achieve continuous power supply of 1.5W at a temperature difference of 5℃. Second, enable the flexible solar thin film, which can provide 3 - 5W of supplementary power when the light intensity is ≥20000 lux. Finally, the lithium polymer battery pack provides a stable power supply, supporting fast charging technology (80% capacity can be fully charged in 30 minutes). The system has an intelligent energy flow distribution algorithm to dynamically adjust the output power of each module to maximize the energy efficiency ratio.

[0111] The thermoelectric cooling module uses a thermoelectric cooler (TEC) combined with a microchannel liquid cooling technology, and starts the gradient cooling mode in the hot state: First, achieve a temperature difference cooling of - 5℃ through the TEC, and at the same time, the liquid cooling system conducts heat to the phase change material energy storage layer (PCM). This material can absorb 300 kJ / kg of heat when it undergoes a phase change at 25℃. The waste heat recovery system integrates an electromagnetic induction coil, uses the mechanical energy generated by human movement for piezoelectric power generation, and cooperates with a supercapacitor bank to achieve millisecond - level electrical energy storage.

[0112] In the normal state, the system enters the low - power operation mode: Only maintain the power supply of the micro - electro - mechanical system (MEMS) sensor array, including temperature and humidity, air pressure, heart rate, and blood oxygen sensors, and the overall power consumption is controlled below 0.5 mW. At the same time, start the ambient temperature difference power generation module, which converts the ambient temperature fluctuation into electrical energy using the Peltier effect. Under the condition of a 10℃ day - night temperature difference, it can achieve an average daily energy recovery of 0.8 Wh. The system also has an adaptive learning function, which optimizes the temperature control strategy through a neural network algorithm and gradually establishes a personalized thermal comfort model.

[0113] Its beneficial effects are as follows: By real - time data collection and dynamic baseline calibration of multi - modal sensors, a precise feature data set is constructed, effectively improving the accuracy of raw data processing and providing a reliable data basis for subsequent temperature analysis; Secondly, by integrating the Pennes bio - heat equation and the dynamic heat balance equation, the initial human body temperature and the system's active regulation of heat flux are accurately calculated, realizing the scientific nature of temperature calculation and meeting the actual needs of the human body from the biological thermophysical level; Furthermore, based on the CNN - LSTM hybrid model optimized by the IGWO gray wolf algorithm, which not only uses CNN to extract data spatial features and LSTM to process time - series features, but also optimizes the model parameters through the improved algorithm, significantly improving the accuracy and efficiency of temperature prediction and achieving precise multi - step temperature prediction; Finally, applying the prediction results to the temperature control of the power supply control system, forming a closed - loop system of "data collection - calculation analysis - precise prediction - intelligent regulation", enabling the intelligent clothing power supply system to more precisely adapt to the human body temperature needs, and improving the temperature control comfort, energy utilization efficiency, and use safety of wearable devices.

[0114] In this embodiment, please refer toFigure 2 , the second embodiment of the intelligent clothing power control method based on temperature feedback in the embodiments of the present invention. Input the feature data set and the system's active regulation of heat flux into the IGWO-CNN-LSTM temperature intelligent prediction model for prediction to obtain multi-step temperature prediction results, including the following steps:

[0115] Step 201: Obtain the feature data set and the system's active regulation of heat flux. Divide the feature data set into time series sample data according to a sliding window. The length of the sliding window is 60s, and the step size is 1s. Normalize the time series sample data to the interval [0, 1];

[0116] Step 202: Use the system's active regulation of heat flux as an external covariate and splice it with the feature data to form a model input tensor;

[0117] Step 203: Use the Seq2Seq structure encoder-decoder architecture. The decoder iteratively outputs the predicted values for the next 5 steps in an autoregressive manner to obtain multi-step temperature prediction results.

[0118] Its beneficial effect is that it can improve the accuracy of data prediction, significantly improve the accuracy and efficiency of temperature prediction, and achieve accurate prediction of multi-step temperature.

[0119] In this embodiment, please refer to Figure 3 , the third embodiment of the intelligent clothing power control method and system based on temperature feedback in the embodiments of the present invention. Temperature regulation of the power control system of the intelligent clothing according to the multi-step temperature prediction results includes the following steps:

[0120] Step 301: Preset three-level temperature states, including cold state, hot state, and normal state. Judge the temperature of the intelligent clothing according to the multi-step temperature prediction results;

[0121] Step 302: When it is judged to be in the cold state, start the heating module of the intelligent clothing. The power supply priority of the heating module of the intelligent clothing is: thermal energy conversion to electrical energy > solar energy > rechargeable battery;

[0122] Step 303: When it is judged to be in the hot state, start the thermoelectric refrigeration module of the intelligent clothing, and at the same time convert the body surface waste heat into electrical energy for storage;

[0123] Step 304: When it is judged to be in the normal state, turn off the active temperature control module of the intelligent clothing, only maintain the power supply of the sensor, and start the thermoelectric module to recover the environmental temperature difference electrical energy.

[0124] Its beneficial effect is that it saves the power of the intelligent clothing, can recover energy, and achieves energy recycling.

[0125] The above describes the intelligent clothing power control method based on temperature feedback provided by the embodiments of the present invention. Next, the intelligent clothing power control system based on temperature feedback of the embodiments of the present invention will be described. Please refer to Figure 4 In an embodiment of the intelligent clothing power control system in the embodiments of the present invention, it includes:

[0126] A data acquisition module, configured to obtain the real-time acquisition data in the multi-modal sensor, perform dynamic baseline calibration on the real-time acquisition data, and construct a feature vector to obtain a feature data set;

[0127] A heat flux acquisition module, configured to perform fusion calculation on the human body temperature according to the Pennes bioheat equation to obtain the initial human body temperature, and calculate the initial human body temperature based on the dynamic heat balance equation to obtain the system's active regulation heat flux;

[0128] A model establishment module, configured to establish a CNN-LSTM hybrid model neural network model, train the hybrid model using the training data set, and optimize the model parameters through an improved IGWO gray wolf algorithm to obtain an IGWO-CNN-LSTM temperature intelligent prediction model;

[0129] A result prediction module, configured to input the feature data set and the system's active regulation heat flux into the IGWO-CNN-LSTM temperature intelligent prediction model for prediction to obtain a multi-step temperature prediction result;

[0130] A temperature regulation module, configured to perform temperature regulation on the power control system of the intelligent clothing according to the multi-step temperature prediction result.

[0131] The above shows and describes 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 by the above embodiments. The above embodiments and the descriptions in the specification are only preferred examples of the present invention and are not used to limit the present invention. Without departing from the spirit and scope of the present invention, the present invention will have various changes and improvements, and these changes and improvements all fall within the scope of the present invention claimed.

Claims

1. An intelligent clothing power supply control method based on temperature feedback, characterized in that The intelligent clothing power control method includes the following steps: Obtain the real-time acquisition data in the multimodal sensor, perform dynamic baseline calibration on the real-time acquisition data, and then construct a feature vector to obtain a feature dataset; Perform fusion calculation on the human body temperature according to the Pennes bioheat equation to obtain the initial human body temperature, and calculate the initial human body temperature based on the dynamic heat balance equation to obtain the system's active regulation of heat flux; Establish a CNN-LSTM hybrid model neural network model, train the hybrid model using the training dataset, and optimize the model parameters through an improved IGWO gray wolf algorithm to obtain an IGWO-CNN-LSTM temperature intelligent prediction model; Input the feature dataset and the system's active regulation of heat flux into the IGWO-CNN-LSTM temperature intelligent prediction model for prediction to obtain a multi-step temperature prediction result; Perform temperature regulation on the power control system of the intelligent clothing according to the multi-step temperature prediction result.

2. The intelligent clothing power supply control method based on temperature feedback according to claim 1, wherein, The step of obtaining the real-time acquisition data in the multimodal sensor, performing dynamic baseline calibration on the real-time acquisition data, and then constructing a feature vector to obtain a feature dataset includes: The multimodal sensor includes at least an infrared temperature sensor, an environmental temperature and humidity sensor, a six-axis accelerometer, and a pressure sensor; Use an adaptive calibration algorithm to calibrate the real-time acquisition data, update the baseline parameters every 500 ms to obtain calibrated acquisition data; Construct a multi-dimensional feature vector in the calibrated acquisition data to obtain a feature dataset; the multi-dimensional features at least include the average skin temperature at multiple points on the forehead, armpit, and neck, environmental temperature, environmental humidity, three-dimensional acceleration vector, metabolic equivalent based on acceleration, and posture recognition.

3. The intelligent clothing power supply control method based on temperature feedback according to claim 1, characterized in that The step of performing fusion calculation on the human body temperature according to the Pennes bioheat equation to obtain the initial human body temperature, and calculating the initial human body temperature based on the dynamic heat balance equation to obtain the system's active regulation of heat flux includes: Obtain the initial human body temperature data through the thermocouples on the chest and back, and estimate the initial human body temperature data based on the Pennes bioheat equation to obtain the human core temperature T core , where the core temperature conduction coefficient α of the human skin is 0.8 W / m·k; Introduce the weight of the human body motion state, where the infrared weight is 0.7 when the human body is stationary and dynamically adjusted to 0.4 by the IMU when the human body is moving, and weight and fuse the human core temperature T core Perform weighted fusion; Establish a dynamic heat balance equation: Among them, h represents the convection coefficient, A represents the effective heat dissipation area, ∈ represents the emissivity, σ represents the Stefan-Boltzmann constant, T target represents the target temperature, represents the human skin temperature, represents the human ambient temperature, Q active is the active regulation heat flux of the system, representing the magnitude of the heat flow involved in the active regulation process of the system; Calculate the initial human body temperature based on the dynamic heat balance equation to obtain the system's active regulation of heat flux.

4. The intelligent clothing power supply control method based on temperature feedback according to claim 1, characterized in that The step of establishing a CNN-LSTM hybrid model neural network model and training the hybrid model using the training dataset includes: Establish a CNN-LSTM hybrid model neural network model, where the CNN feature extraction module is a 2D convolutional layer. The first layer has 16 3×1 convolutional kernels for extracting local time patterns, and the second layer has 32 5×1 convolutional kernels for capturing larger time range features; The LSTM time series modeling module is a forward + backward LSTM memory network for capturing long-term dependencies. The hidden layer dimension is 64 units for balancing model capacity and complexity; the attention mechanism is a time attention layer for weighted focusing on key time points; Normalize the temperature features, humidity features, acceleration features, and wind speed features in the training dataset, and after performing time series reconstruction on the normalized data, input it into the CNN-LSTM hybrid model neural network model; The loss function of the model is a composite loss function, where L = 0.7MAE + 0.3MAPE. The initial value of the adaptive learning rate of the model is set to 0.001, and it decays by 20% every 5 steps. The gradient clipping threshold of the model is ±0.

5.

5. The intelligent clothing power supply control method based on temperature feedback according to claim 1, wherein, The above-mentioned is optimized by the improved IGWO grey wolf algorithm to obtain the IGWO-CNN-LSTM temperature intelligent prediction model, including: Optimize the non-linear convergence factor of the GWO grey wolf algorithm, and its calculation formula is as follows: Among them, a(t) represents the non-linear convergence factor, t represents the current iteration number, T represents the maximum iteration number, and the cubic decay accelerates the convergence speed in the later stage; Update the dynamic weight position of the GWO grey wolf algorithm, and its calculation formula is as follows: Among them, represents the position of the new generation of gray wolf individuals, represents the position of the individual with the highest fitness, represents the position of the individual with the second highest fitness, represents the position of the individual with the third highest fitness, ω1 represents the weight of the highest fitness, ω2 represents the weight of the second highest fitness, ω3 represents the weight of the third highest fitness, and α, β, δ represent the positions of the wolves; Weight calculation: Among them, represents the fitness function value, which is used to evaluate the quality of individuals, represents the predicted positions of each leading wolf; Obtain the grey wolf position vector after position update, map the parameters of the CNN-LSTM hybrid model neural network model to the grey wolf position vector, and obtain the IGWO-CNN-LSTM temperature intelligent prediction model.

6. The intelligent clothing power supply control method based on temperature feedback according to claim 1, characterized in that The above-mentioned inputs the feature data set and the system's active regulation of heat flux into the IGWO-CNN-LSTM temperature intelligent prediction model for prediction, and obtains the multi-step temperature prediction result, including: Obtain the feature data set and the system's active regulation of heat flux, divide the feature data set into time series sample data according to a sliding window, the length of the sliding window is 60s, the step size is 1s, and normalize the time series sample data to the interval [0,1]; Use the system's active regulation of heat flux as an external covariate and splice it with the feature data to form a model input tensor; Use the Seq2Seq structure encoder-decoder architecture, and the decoder iteratively outputs the predicted values for the next 5 steps in an autoregressive manner to obtain the multi-step temperature prediction result.

7. The intelligent clothing power supply control method based on temperature feedback according to claim 1, characterized in that, The above-mentioned conducts temperature regulation on the power control system of the smart clothing according to the multi-step temperature prediction result, including: Preset three-level temperature states, including cold state, hot state and normal state, and judge the temperature of the smart clothing according to the multi-step temperature prediction result; When it is judged to be in the cold state, start the heating module of the smart clothing, and the power supply priority of the heating module of the smart clothing is: thermal energy conversion to electrical energy > solar energy > rechargeable battery; When it is judged to be in the hot state, start the thermoelectric refrigeration module of the smart clothing, and at the same time convert the body surface waste heat into electrical energy for storage; When it is judged to be in the normal state, turn off the active temperature control module of the smart clothing, only maintain the power supply of the sensor, and start the thermoelectric module to recover the ambient temperature difference electrical energy.

8. The intelligent clothing power control system based on temperature feedback is characterized in that The power control system of the smart clothing includes the following modules: A data acquisition module, which is used to obtain the real-time acquisition data in the multi-modal sensor, perform dynamic baseline calibration on the real-time acquisition data, and construct a feature vector to obtain a feature data set; A heat flux acquisition module, which is used to perform fusion calculation on the human body temperature according to the Pennes bioheat equation to obtain the initial human body temperature, and calculate the system's active regulation of heat flux based on the dynamic heat balance equation; A model establishment module, which is used to establish a CNN-LSTM hybrid model neural network model, train the hybrid model using a training data set, and optimize the model parameters through an improved IGWO grey wolf algorithm to obtain an IGWO-CNN-LSTM intelligent temperature prediction model; A result prediction module, which is used to input the feature data set and the system actively regulated heat flux into the IGWO-CNN-LSTM intelligent temperature prediction model for prediction to obtain multi-step temperature prediction results; A temperature regulation module, which is used to regulate the temperature of the power control system of the intelligent clothing according to the multi-step temperature prediction results.

9. The intelligent clothing power supply control system based on temperature feedback according to claim 8, characterized in that, The heat flux acquisition module includes the following units: A data acquisition unit, configured to obtain initial human body temperature data through chest and back thermocouples, and estimate the initial human body temperature data based on the Pennes bioheat equation to obtain the human core temperature T core , where the core temperature conduction coefficient α of the human skin is 0.8 W / m·k; A weighted fusion unit is used to introduce the weight of the human body's motion state, where the infrared weight is 0.7 when the human body is stationary and dynamically adjusted to 0.4 by the IMU when the human body is moving, and the core temperature T of the human body core is weighted and fused; A heat balance calculation unit, which is used to establish a dynamic heat balance equation: where h represents the convective coefficient, A represents the effective heat dissipation area, ∈ represents the emissivity, σ represents the Stefan-Boltzmann constant, T target represents the target temperature, represents the human skin temperature, represents the human ambient temperature, Q active is the active regulation heat flux of the system, indicating the magnitude of the heat flow involved in the active regulation process of the system; A heat flux obtaining unit, which is used to calculate the initial body temperature based on the dynamic heat balance equation to obtain the system actively regulated heat flux.

10. The intelligent clothing power supply control system based on temperature feedback according to claim 8, characterized in that, The temperature regulation module includes the following units: A temperature state judgment unit, which is used to preset three-level temperature states, including a cold state, a hot state, and a normal state, and judge the temperature of the intelligent clothing according to the multi-step temperature prediction results; A cold state execution unit, which is used to start the heating module of the intelligent clothing when it is judged to be in the cold state, and the power supply priority of the heating module of the intelligent clothing is: thermal energy conversion to electrical energy > solar energy > rechargeable battery; A hot state execution unit, which is used to start the thermoelectric refrigeration module of the intelligent clothing when it is judged to be in the hot state, and at the same time convert the body surface waste heat into electrical energy for storage; A normal state execution unit, which is used to turn off the active temperature control module of the intelligent clothing when it is judged to be in the normal state, only maintain the power supply of the sensor, and start the thermoelectric module to recover the ambient temperature difference electrical energy.

Images