An air conditioning control method and system
By acquiring occupant body surface temperature and air conditioning actuator data to construct a normalized regional stimulus input, individual dynamic thermal response parameters are identified, solving the air conditioning control deviation problem caused by differences in occupant thermal response and achieving the accuracy and reliability of individualized air conditioning control.
Patent Information
- Application Number
- CN202611574871.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-10-12
- Publication Date
- 2026-11-17
AI Technical Summary
Existing air conditioning control technology cannot effectively identify individual differences in the speed and intensity of thermal response among occupants, resulting in a discrepancy between the control effect and the actual thermal sensation. Furthermore, individual models are prone to distortion, leading to frequent occurrences of overcooling, overheating, and frequent adjustments.
By acquiring occupant body surface temperature and actual air conditioning actuator data, a normalized region stimulus input is constructed to confirm effective air conditioning stimuli. When the sample quality reaches a preset threshold, individual dynamic thermal response parameters are identified, and these parameters are used to recursively predict future thermal states and update the model.
Individualized air conditioning control was achieved, which improved the accuracy and reliability of control, reduced the risk of model contamination caused by low-quality samples, and reduced the occurrence of overcooling, overheating and frequent adjustments.
Smart Images

Figure CN123104336A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of vehicle air conditioning control technology, and in particular to an air conditioning control method and system. Background Technology
[0002] Existing technologies include solutions for air conditioning control based on infrared body surface temperature or environmental parameters, as well as solutions for recalling historical settings based on occupant identity and establishing personalized comfort models based on physiological parameters.
[0003] The aforementioned solutions can identify occupants, detect thermal status, or record occupant adjustment actions. However, they are insufficient in reflecting individual differences in the speed and intensity of thermal response among different occupants. The comfort models used are mostly based on uniform calibration or group statistics, resulting in discrepancies between the control effect and actual thermal sensation. Some solutions use manual adjustments by occupants as the learning basis, but manual adjustments reflect occupant operating habits, which are inconsistent with the actual thermal effects experienced by the human body, making the resulting individual models prone to distortion. Furthermore, the measurement quality of body surface temperature, obstruction, and changes in ambient temperature, sunlight, and vehicle speed are not differentiated, and changes in body surface temperature are easily attributed to the air conditioning effect. The online learning of individual models lacks data quality screening, with low-quality data also participating in the learning process, causing the model to gradually deviate from the true response after long-term use. The control mostly reacts based on the current body surface temperature, lacking prediction of subsequent hot and cold trends, easily leading to overcooling, overheating, and frequent adjustments.
[0004] Therefore, there is an urgent need to provide a technical solution to address the above problems. Summary of the Invention
[0005] To address the aforementioned technical problems, this invention provides an air conditioning control method and system.
[0006] Firstly, the present invention provides an air conditioning control method applied to a vehicle cabin, the technical solution of which is as follows: Acquire the surface temperature of each body region of the occupant in the target seat and the actual execution data of the air conditioning actuator; construct a directional normalized regional stimulus input to act on the occupant based on the actual execution data; when the change amplitude of the regional stimulus input reaches a preset excitation threshold and lasts for a preset time, and the actual execution direction is consistent with the control command, determine the candidate air conditioning stimulus, and establish a baseline window before stimulation and a response window after stimulation. When the candidate air conditioning stimulus is confirmed as a valid air conditioning stimulus, the identity confidence, thermal zone registration confidence, body surface temperature measurement quality, stimulus quality, context quality, and perturbation quality are summarized according to a preset summary relationship to obtain the sample quality. When the sample quality reaches a preset quality threshold, based on the body surface temperature within the regional stimulus input and the baseline window and response window, the individual dynamic thermal response parameters are identified separately for cooling and heating, and the corresponding covariance and confidence are obtained. The individual dynamic thermal response parameters include the input normalized response gain, response delay, and time constant. The future human thermal state under each candidate control sequence is recursively predicted using the individual dynamic thermal response parameters, and the air conditioning control quantity is determined and executed. The prediction error is determined based on the actual body surface response after execution, and the sample quality is reassessed. When the sample quality reaches a preset update threshold, the individual dynamic thermal response parameters and corresponding covariance and confidence level are updated using the prediction error; otherwise, they remain unchanged.
[0007] The beneficial effects of the air conditioning control method of the present invention are as follows: The method of this invention constructs a directional normalized region stimulus input using the actual execution data of the air conditioner actuator, enabling stimulus confirmation based on actual execution changes rather than solely on control commands. This establishes a correspondence between actual thermal effects and body surface temperature responses. By identifying individual dynamic thermal response parameters only when sample quality reaches a preset quality threshold, and updating parameters only when sample quality reaches a preset update threshold, model parameters are updated only under high-quality sample conditions, reducing contamination of model parameters by low-quality samples. By identifying individual dynamic thermal response parameters, including input normalized response gain, response delay, and time constant, separately for cooling and heating, the identified parameters reflect the dynamic thermal response characteristics of individual occupants, thus providing an individualized model basis for subsequent predictive control.
[0008] Secondly, the present invention provides an air conditioning control system, the air conditioning control system comprising: The stimulus detection module is used to acquire the surface temperature of each human body area of the occupant in the target seat and the actual execution data of the air conditioning actuator. Based on the actual execution data, a directional normalized regional stimulus input is constructed to act on the occupant. When the change amplitude of the regional stimulus input reaches the preset excitation threshold and lasts for a preset time, and the actual execution direction is consistent with the control command, a candidate air conditioning stimulus is determined and a baseline window before stimulation and a response window after stimulation are established. The gating identification module is used to summarize the identity confidence, thermal zone registration confidence, body surface temperature measurement quality, excitation quality, context quality, and perturbation quality according to a preset summary relationship when the candidate air conditioning stimulus is confirmed as a valid air conditioning stimulus, to obtain the sample quality. When the sample quality reaches a preset quality threshold, the module identifies the individual dynamic thermal response parameters according to cooling and heating based on the body surface temperature within the regional stimulus input, baseline window, and response window, and obtains the corresponding covariance and confidence. The individual dynamic thermal response parameters include input normalized response gain, response delay, and time constant. The predictive control module is used to recursively predict the future human thermal state under the action of each candidate control sequence using the individual dynamic thermal response parameters, and to determine and execute the air conditioning control quantity. The update module is used to determine the prediction error based on the actual body surface response after execution and to re-determine the sample quality. When the sample quality reaches a preset update threshold, the prediction error is used to update the individual dynamic thermal response parameters and the corresponding covariance and confidence level; otherwise, they remain unchanged.
[0009] The beneficial effects of the air conditioning control system of the present invention are as follows: The system of the present invention, through the coordinated work of the stimulus detection module, gating identification module, predictive control module and update module, enables the system to identify individual dynamic thermal response parameters based on the actual execution data of the air conditioning actuator, and update the parameters when the sample quality meets the conditions, thereby achieving the accuracy and reliability of individualized air conditioning control at the system level.
[0010] The above description is merely an overview of the technical solution of the present invention. In order to better understand the technical means of the present invention and to implement it in accordance with the contents of the specification, and in order to make the above and other objects, features and advantages of the present invention more apparent and understandable, specific embodiments of the present invention are described below. Attached Figure Description
[0011] The accompanying drawings are for illustrative purposes only and are not intended to limit the invention. Furthermore, the same reference numerals denote the same parts throughout the drawings. In the drawings: Figure 1 This is a flowchart illustrating the air conditioning control method according to an embodiment of the present invention; Figure 2 This is a schematic diagram of an air conditioning control closed loop driven by individual thermal response according to an embodiment of the present invention; Figure 3 This is a schematic diagram illustrating the individual cooling and heating dynamic response curves and parameter meanings of an embodiment of the present invention; Figure 4 This is a schematic diagram of the dual verification and state transition of the verification-based preference memory according to an embodiment of the present invention; Figure 5This is a structural block diagram of the air conditioning control system according to an embodiment of the present invention. Detailed Implementation
[0012] Exemplary embodiments of the invention will now be described in more detail with reference to the accompanying drawings. While exemplary embodiments of the invention are shown in the drawings, it should be understood that the invention can be implemented in various forms and should not be limited to the embodiments set forth herein.
[0013] Example 1: like Figure 1 As shown, this embodiment provides an air conditioning control method applied to a vehicle cabin. The target seat is the seat of the occupant currently undergoing personalized control, which can be the driver's seat or other occupant seats; the human body area refers to the temperature measurement areas of the occupant's face, neck, chest, hands, etc. Body surface temperature, area stimulus input, individual dynamic thermal response parameters, and actual body surface response are all associated with the occupant. For a single occupant, the personalized air conditioning control quantity is determined based on the occupant's needs; for multiple occupants, each occupant is identified and predicted separately, then the shared cooling or heating capacity is coordinated, and local actuators are adjusted according to the vehicle configuration. In both cases, safety requirements take priority. This method can be used in fuel vehicles, pure electric vehicles, or hybrid vehicles with corresponding sensing and adjustment capabilities, without requiring each seat to have an independent air conditioning unit or controller. Figure 2 As shown, this method forms a closed loop of perception, identification, prediction, and updating, and includes the following steps.
[0014] Step S100: Obtain the surface temperature of each body region of the occupant in the target seat and the actual execution data of the air conditioning actuator; construct a directional normalized regional stimulus input to act on the occupant based on the actual execution data; when the change amplitude of the regional stimulus input reaches the preset excitation threshold and lasts for a preset time, and the actual execution direction is consistent with the control command, determine the candidate air conditioning stimulus, and establish the baseline window before stimulation and the response window after stimulation.
[0015] Specifically, the surface temperature of each human body region can be obtained through infrared array sensors, thermal imaging sensors, contact temperature sensors, or multi-source fusion. Human body regions can include thermal sensing areas such as the face, forehead, neck, chest, and hands. When acquiring the surface temperature of each human body region, invalid pixels and abrupt changes are first removed. Then, a representative temperature for each human body region is calculated based on pixel quality and region weight. ; in, Let r be the representative temperature of human body region r at time k; Let be the set of valid pixels in the human body region r at time k; p is a single pixel in the set of valid pixels. Let p be the quality coefficient of pixel p at time k; The region weight for pixel p; This represents the temperature value of pixel p at time k. Invalid pixels may include sensor dead pixels, obstructed pixels, and pixels with abnormal readings due to reflection or sunlight; abrupt changes may include pixels where the temperature difference between adjacent sampling times exceeds a preset range.
[0016] Then, the representative temperatures of each region are combined according to their regional base weights and the current measurement quality of that region to form the overall surface temperature, i.e.: ; in, Let i be the overall surface temperature of occupant i at time k; For regional basic weights; The measurement quality of region r at time k. The basic weights of each region can be preset based on its impact on overall thermal comfort; for example, the weight of the face and neck can be higher than that of the hands. Measurement quality can be determined based on factors such as the effective pixel ratio and continuous visibility time of the region. If the denominator is lower than the minimum effective weight threshold, the overall body surface temperature at this time is marked as invalid, no division operation is performed, and it is not used for the identification and updating of individual dynamic thermal response parameters. When the number of effective regions, continuous visibility time, or thermal zone registration confidence is lower than the threshold, the overall body surface temperature is only used for display or basic control. The overall body surface temperature is a composite result of the body surface temperatures of each human body region, used to define subsequent response quantities and calculate predicted thermal comfort indices; it has a composite relationship with the body surface temperatures of each human body region.
[0017] Actual execution data of an air conditioning actuator refers to feedback data reflecting the actual operating status of the actuator, distinct from control commands. Control commands are target settings issued by the controller, such as adjusting the set temperature to 24°C. Actual execution data reflects the actual thermal output of the actuator, such as the actual power of the compressor or heat pump, the actual speed or airflow of the fan, the actual outlet air temperature, the airflow direction or zone valve position, the seat heating current, and the hot zone airflow status. There may be discrepancies between control commands and actual thermal effects; for example, a command might require increased cooling capacity, but the compressor might not increase its power synchronously due to protection logic. Therefore, constructing zone stimulus inputs using actual execution data can establish a correspondence between actual thermal effects and body surface temperature responses. When constructing zone stimulus inputs, the actual execution data is associated with the occupant by combining the hit coefficient of the airflow to the corresponding body area of the occupant in the target seat. Sharing the compressor's power feedback does not mean that all occupants experience the same zone stimulus input.
[0018] Based on actual execution data, a directional, normalized regional stimulus input is constructed to act on the occupants. Directionality means that the direction of heating is taken as positive, and the direction of cooling is taken as negative, allowing both heating and cooling effects to be distinguished and processed within the same framework. Normalization means mapping the actual execution data to a uniform numerical range, facilitating comparability of parameters across different vehicle models and operating conditions.
[0019] When the amplitude of the change in the regional stimulus input reaches a preset excitation threshold and lasts for a preset time, and the actual execution direction is consistent with the control command, a candidate air conditioning stimulus is determined. The excitation threshold and duration can be calibrated by combining sensor performance, vehicle thermal inertia, and sampling period. For example, the effective change in set temperature can be calibrated to 0.5–1.5℃, the effective change in actual outlet air temperature can be calibrated to 1–3℃, and the effective change in normalized air volume can be calibrated to 0.10–0.25. The above values are for calibration reference and are not the only values for the protection range. The actual execution direction being consistent with the control command means that when the control command requires heating, the actual execution data also reflects the heating effect, and when the control command requires cooling, the actual execution data also reflects the cooling effect. Changes in the set value of the air conditioning control quantity or changes in the control command are only used to initiate candidate observation and are not considered as valid air conditioning stimuli.
[0020] While identifying candidate air conditioning stimuli, a baseline window before stimulation and a response window after stimulation are established. The baseline window is a data window used to determine the baseline temperature before stimulation; the response window is a collective term for all observation windows after stimulation, including the delay window, dynamic response window, and stability verification window. The delay window corresponds to the duration between the moment the stimulus input is applied to the region and the first credible body surface temperature response; the dynamic response window is used to extract the body surface temperature response and identify individual dynamic thermal response parameters; the stability verification window is used to confirm whether the occupant has entered a stable and comfortable state. These windows are established simultaneously when identifying candidate air conditioning stimuli. The duration of each window can be calibrated based on sensor performance, vehicle model thermal inertia, and sampling period. For example, the baseline window can be 30–180 s, the dynamic response window can be 120–900 s, and the stability verification window can be 120–600 s. These values are for calibration reference and are not the sole values for the protection range.
[0021] Step S200: When a candidate air conditioning stimulus is confirmed as a valid air conditioning stimulus, the identity confidence, thermal zone registration confidence, body surface temperature measurement quality, stimulus quality, context quality, and perturbation quality are summarized according to a preset summary relationship to obtain the sample quality. When the sample quality reaches a preset quality threshold, the individual dynamic thermal response parameters are identified according to cooling and heating based on the body surface temperature within the regional stimulus input, baseline window, and response window, and the corresponding covariance and confidence are obtained. The individual dynamic thermal response parameters include input normalized response gain, response delay, and time constant.
[0022] Specifically, the conditions for confirming a candidate air conditioning stimulus as a valid air conditioning stimulus may include: at least one of the following execution feedbacks in the actual execution data—compressor or heat pump actual power, actual outlet air temperature, actual air volume, air direction or zone valve position, seat heating current, and hot zone air supply status—is continuously valid, and the signal-to-noise ratio of the zone stimulus input reaches a preset threshold; no second significant change in air conditioning execution occurs within the baseline window and response window; the change in body surface temperature within the response window relative to the baseline temperature within the baseline window exceeds a preset measurement noise threshold, and the change in body surface temperature is consistent with the zone stimulus input in time and direction of action. When all the above conditions are met, the candidate air conditioning stimulus is confirmed as a valid air conditioning stimulus; when any condition used to confirm a valid air conditioning stimulus is no longer met within the baseline window or response window, the current sample is terminated or re-segmented. See Example 2 for the specific meaning of the above conditions.
[0023] Sample quality is a gating condition used to determine whether the stimulus-response data is suitable for identifying individual dynamic thermal response parameters. Higher sample quality means that the identified individual dynamic thermal response parameters better reflect the occupant's true dynamic thermal response characteristics. Insufficient sample quality may result in contamination from factors such as identity mismatch, occlusion of the human thermal sensing area, weak stimulus, or sudden environmental changes, making it unsuitable for identifying and updating individual dynamic thermal response parameters. Identity confidence characterizes the reliability of the match between the occupant's identifier and the target seat, while thermal zone registration confidence characterizes the reliability of the registration between the occupant's human thermal sensing area and the sensor detection area. The determination methods for surface temperature measurement quality, stimulus quality, context quality, and perturbation quality are described in Example 4.
[0024] When the sample quality reaches a preset quality threshold, based on the regional stimulus input and the body surface temperature within the baseline and response windows, individual dynamic thermal response parameters are identified separately for cooling and heating, and the corresponding covariance and confidence scores are obtained. Individual dynamic thermal response parameters include input normalized response gain, response delay, and time constant; their meanings and identification methods are described in Example 3. Covariance is used to measure the uncertainty of parameter estimation, and confidence scores are used to measure the usability of the parameters. Individual dynamic thermal response parameters for cooling and heating are stored separately, and parameter clusters can be set according to seat, season, or environmental range.
[0025] Step S300: The future human thermal state under the action of each candidate control sequence is recursively predicted using individual dynamic thermal response parameters, and the air conditioning control quantity is determined and executed.
[0026] Specifically, for each candidate control sequence, the surface temperature, rate of change of surface temperature, and regional temperature difference of each human body region are recursively predicted using individual dynamic thermal response parameters. Based on the deviation of surface temperature from a predetermined individual thermal neutrality baseline, the rate of change of surface temperature, the regional temperature difference, and the environmental heat load, a predicted thermal comfort index characterizing the individual's thermal comfort state is determined. Then, based on the predicted thermal comfort index, changes in air conditioning control quantities, and energy consumption corresponding to each candidate control sequence, the air conditioning control quantities are solved under safety constraints with the goal of minimizing the weighted sum of comfort deviation, control changes, and energy consumption. A preset dead zone, minimum hold time, and rate of change limit are applied to the air conditioning control quantities corresponding to the target candidate control sequence before execution. The specific process of recursive prediction, predicting thermal comfort indices, and solving control quantities is detailed in Example 5.
[0027] Step S400: Determine the prediction error based on the actual body surface response after execution, and re-determine the sample quality; when the sample quality reaches the preset update threshold, update the individual dynamic thermal response parameters and the corresponding covariance and confidence level using the prediction error, otherwise keep them unchanged.
[0028] Specifically, after executing the air conditioning control, the actual body surface response of the occupant in the target seat is obtained through a body surface temperature sensor. This response is then compared with the future human thermal state predicted recursively for the same occupant in step S300 to determine the prediction error. The prediction error reflects the accuracy of the current individual dynamic thermal response parameters in characterizing the actual thermal response characteristics of the occupant. At the same time, the sample quality is re-determined based on the actual body surface response after execution.
[0029] When the sample quality reaches a preset update threshold, the individual dynamic thermal response parameters and their corresponding covariance and confidence level are updated using the prediction error. Specifically, the gating update relationship can be expressed as: ; in, This is a gating indicator function; when the sample quality reaches the update threshold q... up The value is 1 if the condition is met, and 0 otherwise; θ is the vector of parameters to be identified; y k These are actual observed values of body surface temperature response; L is the regression vector; k The gain, or Kalman gain, is estimated using recursive least squares or Kalman filtering. Regression vector. Constructed from historical observations of regional stimulus input and body surface temperature.
[0030] When the gating condition is not met, the individual dynamic thermal response parameters, covariance, and confidence level remain unchanged. Furthermore, it is not permissible to freeze only the individual dynamic thermal response parameters while continuing to reduce the covariance; otherwise, artificially reducing the covariance would affect the estimated gain L during subsequent updates. kIf the values are too small, the individual dynamic thermal response parameters cannot effectively respond to new observation data, reducing the model's ability to adapt to changes in occupant thermal response characteristics.
[0031] Through the above steps, this embodiment establishes a complete closed loop: constructing regional stimulus inputs using actual execution data of the air conditioning actuator, identifying individual dynamic thermal response parameters under sample quality gating, recursively predicting and determining air conditioning control quantities using individual dynamic thermal response parameters, and updating individual dynamic thermal response parameters under gating conditions based on actual body surface responses. This establishes a correspondence between actual thermal effects and body surface temperature responses, reducing the risk of model contamination caused by identity mismatch, human thermal perception area occlusion, weak excitation, or sudden environmental changes, and providing a foundation for suppressing overcooling, overheating, and frequent adjustments.
[0032] Example 2: In some embodiments, constructing a directional, normalized region stimulus input acting on the occupant based on actual execution data includes the following steps.
[0033] Step S110: Set the direction of heating to positive and the direction of cooling to negative.
[0034] Specifically, heating tends to increase the occupant's body surface temperature, while cooling tends to decrease it; the two effects on body surface temperature are opposite in direction. By treating heating as positive and cooling as negative, both thermal effects can be distinguished and processed within the same numerical framework. When subsequently identifying individual dynamic thermal response parameters, the stimulus quantities in the heating and cooling directions can be extracted separately for the positive and negative values, thereby identifying the corresponding input normalized response gain, response delay, and time constant for heating and cooling respectively.
[0035] Step S120: When the normalized actual air conditioning thermal power can be obtained from the actual execution data, the region stimulation input is calculated using the normalized actual air conditioning thermal power and the hit coefficient of the air outlet to the corresponding human body area; when the normalized actual air conditioning thermal power cannot be obtained from the actual execution data, the region stimulation input is calculated using the difference between the normalized actual wind speed, the actual air outlet temperature and the body surface temperature of the human body area, and the hit coefficient obtained from the actual execution data.
[0036] Specifically, depending on the signal conditions available from the vehicle, the area stimulus input is implemented using one of two mutually exclusive methods to avoid repeated measurement of the same thermal effect.
[0037] To achieve the first goal, when the normalized actual air conditioning thermal power can be obtained, the regional stimulus input is calculated according to the following formula: ; in, The stimulus input for human body region r at time k; b is the hit coefficient of the air outlet on the human body area r at time k; Q This is the thermal power conversion factor; Let K be the normalized actual air conditioning thermal power at time k.
[0038] The hit coefficient characterizes the degree to which the air conditioning vent's thermal effect hits the human body area *r*. It can be determined based on the airflow direction of the vent, the air volume distribution ratio, and the spatial position of the human body area relative to the vent. For example, the hit coefficient for the chest area directly facing the vent can be higher than that for the hand area, which is away from the airflow direction. The normalized actual air conditioning thermal power characterizes the magnitude and direction of the actual air conditioning thermal effect. The actual cooling thermal power and actual heating thermal power can be directly obtained from the thermal power feedback provided by the vehicle. If they cannot be directly obtained, they are estimated based on actual execution data such as the actual power of the compressor or heat pump, the actual air volume, and the actual air outlet temperature, according to a pre-configured conversion relationship. Then, a sign is assigned according to the direction of the actual thermal effect, i.e., positive for heating and negative for cooling. After normalization, the parameters are comparable between different vehicle models.
[0039] Implementation 2: When the normalized actual air conditioning thermal power cannot be obtained, the regional stimulus input is calculated according to the following formula: ; Among them, b v This is the conversion factor for wind speed and temperature difference; Let k be the normalized actual wind speed at time k. Let K be the actual outlet air temperature at time k. Let be the overall body surface temperature of occupant i at time k. The normalized actual wind speed is obtained by normalizing the actual fan speed or actual air volume. The difference between the actual outlet air temperature and the overall body surface temperature represents the driving force of the temperature difference between the outlet air and the human body. The larger the temperature difference, the stronger the thermal effect of the outlet air on the human body area. When the actual outlet air temperature is higher than the body surface temperature, the difference is positive, corresponding to the heating effect; when the actual outlet air temperature is lower than the body surface temperature, the difference is negative, corresponding to the cooling effect, thus naturally achieving directionality.
[0040] It should be noted that the two implementations mentioned above are mutually exclusive; that is, only one method is used to calculate the regional stimulus input at any given time. This is because if both methods are used simultaneously and the calculation results are superimposed, the same thermal effect will be repeatedly measured, causing the amplitude of the regional stimulus input to be too large. Consequently, the normalized response gain of the identified input will be too small, affecting the accuracy of the individual dynamic thermal response parameters. In practical applications, one implementation can be selected based on the signal conditions available to the vehicle model. For example, vehicles that can obtain the normalized actual air conditioning thermal power can use implementation one; vehicles that cannot obtain the normalized actual air conditioning thermal power but can obtain the normalized actual wind speed, actual air outlet temperature, and human body surface temperature can use implementation two.
[0041] After constructing the regional stimulus input, it is used according to direction: ; ; in, Let be the normalized stimulus quantity for the heating direction at time k. Let be the normalized stimulus quantity in the cooling direction at time k. When the regional stimulus input is positive, Take the positive value and Take zero; when the region stimulus input is negative, Take the opposite of the negative value. Both are set to zero, and are used respectively for subsequent identification of individual dynamic thermal response parameters based on cooling and heating.
[0042] Changes in the setpoint of the air conditioning control quantity or changes in the control command are only used to initiate candidate observations and are not considered as effective air conditioning stimuli.
[0043] Specifically, the control command is the target setting issued by the controller, such as adjusting the set temperature from 26℃ to 24℃. When the control command changes, the system can initiate candidate observation, that is, start monitoring whether the actual execution data changes accordingly to meet the excitation conditions. However, the control command itself does not represent that the actual thermal effect has occurred. For example, if the command requires increasing the cooling capacity but the compressor does not increase its power synchronously due to protection logic, or if the command requires adjusting the airflow direction but the zone valve does not actually operate, the actual thermal effect is insufficient to produce a identifiable body surface temperature response. Therefore, a candidate air conditioning stimulus is only determined when the change amplitude of the zone stimulus input constructed by the actual execution data reaches the preset excitation threshold and lasts for a preset time, and the actual execution direction is consistent with the control command.
[0044] The calibration reference values for the excitation threshold and duration are the same as those described in Example 1. The actual values can be adjusted according to the characteristics of the air conditioning system of the specific vehicle model, the thermal inertia of the cabin, and the measurement accuracy of the temperature sensor.
[0045] Step S210: The candidate air conditioning stimulus is confirmed as a valid air conditioning stimulus, including the following conditions.
[0046] The first condition is that at least one of the following execution feedbacks in the actual execution data—the actual power of the compressor or heat pump, the actual air outlet temperature, the actual air volume, the air direction or zone valve position, the seat heating current, and the hot zone air supply status—is continuously effective, and the signal-to-noise ratio of the zone stimulus input reaches a preset threshold.
[0047] Specifically, continuous and effective execution feedback means that the aforementioned execution feedback quantity is continuously available within the baseline window and response window without interruption or abnormal jumps. For example, the actual power signal of the compressor is continuously and effectively output within the response window. The signal-to-noise ratio (SNR) of the regional stimulus input is the ratio of the useful signal component to the noise component in the regional stimulus input. A SNR reaching a preset threshold indicates that changes in the regional stimulus input can be distinguished from measurement noise. If the SNR is insufficient, changes in the regional stimulus input may be drowned out by noise, and the parameters identified accordingly are unreliable.
[0048] The second condition is that no second significant change in air conditioning performance occurs within the baseline window and response window.
[0049] Specifically, a second significant change in air conditioning execution refers to another change in the air conditioning actuator that satisfies the excitation threshold within the baseline or response window of the current candidate air conditioning stimulus. If a second significant change in air conditioning execution occurs, the change in body surface temperature within the response window may be caused by the second thermal effect, leading to an unclear correspondence between stimulus and response. For example, if the compressor power increases significantly again within the response window of the cooling stimulus, the decrease in body surface temperature may be affected by both cooling effects simultaneously, making it difficult to attribute solely to the first stimulus.
[0050] The third condition is that the change in body surface temperature within the response window relative to the baseline temperature within the baseline window exceeds a preset measurement noise threshold, and the change in body surface temperature is consistent with the regional stimulus input in terms of time and direction of action.
[0051] Specifically, the baseline temperature is a representative value of the overall body surface temperature within the baseline window, such as the average value of the overall body surface temperature within the baseline window. If the change in body surface temperature within the response window relative to the baseline temperature exceeds a preset measurement noise threshold, it indicates that the change in body surface temperature is sufficiently distinguishable from measurement noise. Consistency in time between the body surface temperature change and the regional stimulus input means that the moment the body surface temperature begins to respond is within the response delay range after the moment the regional stimulus input takes effect; consistency in direction means that a heating stimulus corresponds to an increase in body surface temperature, and a cooling stimulus corresponds to a decrease in body surface temperature. If the direction of the body surface temperature change is opposite to the direction of the regional stimulus input, it indicates that the change in body surface temperature may be caused by environmental disturbances or other factors, and should not be considered an effective air conditioning stimulus.
[0052] When all of the above conditions are met, the candidate air conditioning stimulus is confirmed as a valid air conditioning stimulus.
[0053] This embodiment confirms effective air conditioning stimuli from multiple dimensions, including the effectiveness of execution feedback, signal-to-noise ratio, absence of secondary significant execution changes, body surface temperature changes exceeding the measurement noise threshold, and consistency of time and direction. This ensures that the stimulus events used for identification have clear physical meaning and distinguishability, providing a data foundation for the accurate identification of individual dynamic thermal response parameters.
[0054] In some embodiments, the method further includes the following steps.
[0055] Step A100: When any condition used to confirm an effective air conditioning stimulus is no longer met within the baseline window or response window, terminate the current sample or re-segment it; when multiple thermal effects correspond within the response window, segment or distinguish the multiple thermal effects according to the execution time, effect area, or input direction.
[0056] Specifically, within the baseline or response window, if the feedback is interrupted, the signal-to-noise ratio drops below the threshold, a second significant change in air conditioning execution occurs, the change in body surface temperature does not exceed the measurement noise threshold, or the change in body surface temperature is inconsistent with the regional stimulus input in time or direction of action, then the current sample is terminated, meaning the stimulus-response data from this instance will no longer be used to identify individual dynamic thermal response parameters. In other cases, the response window can be re-segmented based on the time of condition failure, treating the data segment before the failure time as an independent sample and processing the data segment after the failure time separately.
[0057] When multiple thermal effects correspond to a response window, such as when an air conditioner actuator performs heating and cooling actions in different directions, or when it performs heating actions on different areas, the multiple thermal effects can be segmented or distinguished according to the execution time, the area of action, or the input direction. For example, the response window can be divided into multiple sub-windows according to the execution time, the response data of different human body areas can be assigned to the corresponding thermal effects, or the heating and cooling effects can be extracted according to the input direction. This establishes an independent correspondence between each thermal effect and the resulting body surface temperature response, avoiding the overlap of response data from multiple thermal effects.
[0058] Example 3: In some embodiments, individual dynamic thermal response parameters are identified using a first-order model with pure delay. The first-order model with pure delay uses the body surface temperature within the baseline window as the baseline, the body surface temperature offset relative to the baseline as the response quantity, and introduces an environmental compensation term.
[0059] Specifically, such as Figure 3 As shown in the figure, this diagram illustrates the dynamic response curves and parameter meanings of an individual's cooling and heating functions. The body surface temperature within the baseline window is initially used as the baseline T. base,i Define the body surface temperature offset: ; in, Let be the body surface temperature offset of occupant i at time k. Let T be the overall body surface temperature of occupant i at time k. base,iThe baseline is the body surface temperature of occupant i within the baseline window. The baseline window is a data window used to determine the baseline temperature before stimulation; baseline T is... base,i The average value of the overall body surface temperature within the baseline window, or the median or weighted average value, can be used.
[0060] Understandably, defining the offset with a baseline allows the identification model to respond with changes in body surface temperature relative to the baseline, thereby eliminating the influence of differences in basal body temperature between individual occupants and differences in ambient temperature on the identification results.
[0061] The first-order model with pure delay can be represented as: ; Among them, a i Let be the autoregressive coefficients of occupant i in the model, satisfying 0 i <1; Input the normalized response gain for the cooling direction of occupant i; Input the normalized response gain for the heating direction of occupant i; For a moment Normalized stimulus in the direction of cooling; For a moment Normalized stimulus in the direction of downward heating; The number of sampling steps corresponding to the response delay; Let i be the vector of environmental compensation coefficients for occupant i. Let k be the environment vector at time k; Let be the model residual at time k.
[0062] The relationship between the time constant and the model autoregressive coefficients is as follows: ; in, Let i be the time constant of crew member i; Let be the sampling period for human thermal data, and ln be the natural logarithm function. The relationship between response delay and the number of sampling steps is as follows: ,in The response delay for crew member i.
[0063] The input normalized response gain characterizes the steady-state change in body surface temperature caused by a unit normalized cooling stimulus or a unit normalized heating stimulus. Specifically, The value represents the change in steady-state body surface temperature caused by a unit normalized cooling stimulus. The larger the value, the more sensitive the occupant i is to cooling, and the greater the decrease in body surface temperature under the same cooling stimulus. It represents the change in steady-state body surface temperature caused by a normalized heating stimulus per unit. The larger the value, the more sensitive the occupant i is to heating, and the greater the increase in body surface temperature under the same heating stimulus. and The unit can be written as ℃ / normalized stimulus. If a dimensional input is used, it should be noted as ℃ / kW, ℃ / (m / s·℃) or other consistent units.
[0064] The response delay characterizes the time between the moment the stimulus input takes effect and the first credible surface temperature response.
[0065] Specifically, such as Figure 3 As shown, the input step time is t0, and the starting point of the first reliable body surface temperature response is t0+t. d The difference between the two is the response delay t. d The first reliable surface temperature response refers to the moment when the absolute value of the surface temperature deviation exceeds the measurement noise threshold and the direction of change is consistent with the direction of the regional stimulus input. Response delay is related to the occupant's physiological characteristics, surface blood flow distribution, and sensor response speed.
[0066] The time constant characterizes the time scale of the change in body surface temperature as the body surface temperature response approaches a steady state.
[0067] Specifically, time constant The timescale required for the body surface temperature response to reach a steady-state change of approximately 63.2%. Smaller occupants respond faster and can reduce air conditioning power in advance to avoid overshoot; Larger occupants respond more slowly, and the air conditioning power should not be continuously increased if no response is observed for a short period of time.
[0068] Environmental compensation items Used to eliminate the influence of environmental factors on body surface temperature.
[0069] Specifically, Let E(k) be the environmental compensation coefficient vector for occupant i, and E(k) be the environmental vector at time k. The environmental vector can include cabin and outdoor temperatures, humidity, sunlight, vehicle speed, window opening status, seat occupancy, evaporator temperature, and actual outlet air temperature. The role of the environmental compensation term is to separate the influence of environmental factors on body surface temperature from the body surface temperature offset, so that the identified input normalized response gain, response delay, and time constant can more accurately reflect the effect of the air conditioning stimulus itself on body surface temperature. For example, when the outside temperature is high and the sunlight is strong, the occupant's body surface temperature may rise due to the environmental heat load. Without environmental compensation, this part of the temperature rise caused by the environment will be attributed to the heating stimulus, resulting in an overestimation of the identified heating direction input normalized response gain.
[0070] When the regional stimulus input is zero, the amplitude of the regional stimulus input change is lower than the excitation threshold, or the signal-to-noise ratio of the regional stimulus input is insufficient, the response window is not used to identify the input normalized response gain.
[0071] Specifically, if the regional stimulus input is zero, there is no identifiable air conditioning stimulus within the response window. Changes in body surface temperature are primarily caused by environmental factors or the occupant's own physiological regulation. If this window is still used to identify the input normalized response gain, environmental disturbances or physiological regulation will be mistakenly attributed to the air conditioning stimulus, leading to distorted identification results. If the amplitude of the regional stimulus input change is below the excitation threshold, the body surface temperature change caused by the air conditioning stimulus may be drowned out by measurement noise. If the signal-to-noise ratio is insufficient, the ratio of the useful signal component to the noise component in the regional stimulus input is too low, making the parameters identified unreliable. In these cases, the response window is not used to identify the input normalized response gain, but it can be used to identify other parameters such as the environmental compensation coefficient.
[0072] like Figure 3 As shown in the left figure, different K values are displayed under the same cooling input. cool The difference in response curves, that is, under the same cooling input, the same delay, and similar time scales, the larger the absolute value of the steady-state cooling, the greater the K value. cool The larger the K, for example, the more sensitive occupant A is to cooling, the higher its K. cool The temperature is relatively high, so the system should reduce cooling earlier to avoid overcooling, while occupant B's K... cool If the τ value is relatively small, the system can appropriately extend or enhance the cooling time. The right figure shows the difference in response curves for different τ values under the same steady-state change. By keeping the steady-state change the same, we can characterize τ using a time scale of 63.2% of the steady-state change. For example, if occupant C has a smaller τ value, the system can reduce the power earlier to avoid overshoot. However, if occupant D has a larger τ value, the system should not continuously increase the heating capacity due to a short period of no response. Instead, it should maintain the minimum hold time and adjust the power only after confirming that the response has been established.
[0073] Step S270: Based on the regional stimulus input and the body surface temperature within the baseline window and response window, identify the individual dynamic thermal response parameters for cooling and heating respectively, and obtain the corresponding covariance and confidence level.
[0074] Specifically, the online estimation in step S270 can be achieved using recursive least squares, with the regression vector... Construct estimation relations, according to The parameter vector θ is adjusted successively. At this point, θ includes the model's autoregressive coefficient a. i Input the regression coefficient corresponding to the cooling direction. Input the regression coefficient corresponding to the heating direction. and environmental compensation coefficient .
[0075] in, , Therefore, after estimating the above regression coefficients, by and The normalized response gain for the cooling direction input and the normalized response gain for the heating direction input are calculated, and then converted using the model autoregressive coefficient α. i The time constant is obtained by converting the aforementioned time constant with respect to the model's autoregressive coefficients. k The actual observed value of body surface temperature response, i.e. ; The specific form of the prediction error in this regression estimate is called the prediction residual; L k To estimate the gain, the covariance matrix P is estimated from the parameters at the previous time step. k-1 Determine, for example Where λ is the forgetting factor, used to adjust the weights of historical data and new observation data, P k It is updated as the parameters are updated and used to measure the uncertainty of the parameter estimates.
[0076] When using Kalman filtering, the parameter vector is treated as a state variable, and the process is recursively applied in two steps: prediction and update. In the prediction step, the parameter estimate and its covariance at the current time are predicted based on the parameter estimate and its covariance from the previous time step. In the update step, the parameter estimate and its covariance are corrected based on the actual observed body surface temperature response at the current time step. Kalman filtering can output the covariance of the parameter estimate during the parameter estimation process, which facilitates subsequent assessment of the reliability of the parameter estimate.
[0077] In addition, other online estimation methods such as Bayesian estimation or lightweight neural networks can be used to obtain individual dynamic thermal response parameters, their covariance, and confidence scores. Bayesian estimation estimates parameters and their uncertainties by maintaining the posterior probability distribution of the parameters, while lightweight neural networks achieve online parameter estimation by training a network model that takes regional stimulus inputs and historical surface temperature data as inputs and individual dynamic thermal response parameters as outputs.
[0078] The covariance mentioned above measures the uncertainty of parameter estimation, while the confidence level measures the usability of the parameters. A smaller covariance indicates a more certain parameter estimate, and a higher confidence level indicates a more reliable parameter that can be directly used for predictive control. A lower confidence level indicates a less reliable parameter, requiring more high-quality samples to update individual dynamic thermal response parameters or a fallback to population parameters. Population parameters are dynamic thermal response reference parameters pre-obtained from historical samples of multiple occupants, used for fallback control when individual dynamic thermal response parameters are unreliable; specific fallback processing should comply with the operating state control rules in Example 7.
[0079] In some embodiments, the method further includes the following steps.
[0080] Step A200: Individual dynamic thermal response parameters are updated only when the response quantity meets the preset parameter confidence conditions; individual dynamic thermal response parameters for cooling and heating are stored separately, and parameter clusters are set according to seat, season or environmental range.
[0081] Specifically, parameter confidence conditions may include: the response magnitude exceeding a measurement noise threshold; the response duration covering a preset proportion of the dynamic response window; the aforementioned prediction residuals being within a preset range; the covariance of the parameter estimate being less than a preset threshold or the confidence level reaching a preset threshold; and the excitation magnitude and sampling window ensuring sufficient information content in the regression quantity. These conditions are used, in turn, to ensure that the body surface temperature response can be distinguished from measurement noise, that there are sufficient data points in the response process for parameter estimation, that the model's fit to the observed data meets requirements, that the reliability of the parameter estimation meets requirements, and that there is no severe collinearity among the components in the regression vector to ensure the identifiability of the parameter estimation.
[0082] Individual dynamic thermal response parameters for cooling and heating are stored separately, and parameter clusters are set according to seat, season, or environmental range. Specifically, the body surface temperature response characteristics of occupants may differ between cooling and heating conditions. For example, the response delay for heating may be greater than the response delay for cooling. Therefore, they are stored separately to avoid mutual interference. Setting parameter clusters by seat means storing separate parameter clusters for occupants in different seats. For example, the driver's seat, the front passenger seat, the left rear seat, and the right rear seat each have their own parameter clusters. Setting parameter clusters by season means storing parameters according to seasonal changes, such as summer parameter clusters, winter parameter clusters, and transitional season parameter clusters. Setting parameter clusters by environmental range means storing parameters according to environmental conditions such as outside temperature range and solar radiation intensity range, such as high temperature and high sunlight range parameter clusters and low temperature and low sunlight range parameter clusters. In this way, the identified parameters can adapt to the differences in thermal response under different operating conditions.
[0083] This embodiment employs a first-order model with pure delay and defines the offset using the body surface temperature within the baseline window as the baseline, enabling the identification model to characterize the dynamic response process of occupant body surface temperature to air conditioning stimulation with fewer parameters. By introducing an environmental compensation term to eliminate the influence of environmental factors, and storing the individual dynamic thermal response parameters for cooling and heating separately, and setting parameter clusters according to seat, season, or environmental range, the applicability of the individual dynamic thermal response parameters under different operating conditions is improved.
[0084] Example 4: In some embodiments, determining sample quality includes the following steps.
[0085] Step S220: Determine the identity confidence and thermal zone registration confidence based on the binding relationship between the occupant's identifier and the target seat and the human body thermal sensing area; when the binding relationship is interrupted, stop writing individual dynamic thermal response parameters.
[0086] Specifically, identity confidence represents the reliability of the match between the occupant identifier and the target seat. The occupant identifier can be obtained through facial recognition, account login, digital key, mobile Bluetooth key, seat position memory, or a combination thereof. After the occupant enters the cabin and takes a seat, the system binds the occupant identifier to the target seat, for example, binding the identified occupant A to the driver's seat. Identity confidence can be determined comprehensively based on factors such as facial recognition similarity, account matching degree, and the degree of consistency between seat position and historical records. Hot zone registration confidence represents the reliability of the registration between the occupant's human thermal sensing area and the sensor detection area. The human thermal sensing area can include areas such as the face, forehead, neck, chest, and hands. For example, when the occupant adjusts their sitting posture or the seat position changes, the facial area detected by the sensor may shift, and the hot zone registration confidence will decrease accordingly.
[0087] When the binding relationship is interrupted, writing individual dynamic thermal response parameters should be stopped. Binding relationship interruption may include situations such as crew member change, crew member seat change, crew member leaving the cabin, or sensors being unable to continue to identify crew members. In this case, the currently acquired body surface temperature data cannot be determined to belong to any particular crew member. If it is still used to identify or update individual dynamic thermal response parameters, the thermal response characteristics of other crew members will be incorrectly written into the current crew member's individual dynamic thermal response parameters, causing parameter contamination. Therefore, writing should be stopped and resumed only after the binding relationship is re-established.
[0088] Step S230: Determine the body surface temperature measurement quality based on at least one of the proportion of effective pixels in the body surface temperature of each human body region and the number of effective regions.
[0089] Specifically, the quality of body surface temperature measurement is used to characterize the reliability of the currently acquired body surface temperatures for each human body region. The proportion of effective pixels refers to the percentage of effective pixels remaining within the sensor detection area corresponding to each human body region, after removing sensor defects, pixels obscured by clothing or objects, and pixels with abnormal readings due to reflection or sunlight, out of the total pixels in that area. The number of effective regions refers to the number of regions in the human body, such as the face, forehead, neck, chest, and hands, where effective temperature data can be obtained. For example, when an occupant wears a scarf to cover their neck, the proportion of effective pixels in the neck area may be below a threshold and not counted in the number of effective regions; when an occupant rests their chin on their hand, the proportion of effective pixels in the facial area may decrease.
[0090] This criterion corresponds to the synthesis process of body surface temperature. Only when the effective pixel ratio, the number of effective regions, and the continuous visibility time meet the requirements will the body surface temperature of each human body region participate in the synthesis of the region's representative temperature and the overall body surface temperature, and be used for the identification and updating of individual dynamic thermal response parameters. That is, when the effective pixel ratio is lower than a preset ratio threshold, the representative temperature of that region will not participate in the synthesis of the overall body surface temperature; when the number of effective regions is lower than a preset number threshold, the representativeness of the overall body surface temperature will be insufficient; and when the continuous visibility time is lower than a preset time threshold, the temperature data of that region will be unstable and will not be used for the identification and updating of individual dynamic thermal response parameters. Therefore, body surface temperature data with lower measurement quality will not have an adverse effect on the identification of individual dynamic thermal response parameters.
[0091] Step S240: When a candidate air conditioning stimulus is confirmed as a valid air conditioning stimulus, the stimulus quality is set to a non-zero value; when a candidate air conditioning stimulus is not confirmed as a valid air conditioning stimulus, the stimulus quality is set to zero, and the sample in this case is not included in the identification.
[0092] Specifically, the excitation quality characterizes whether the current air conditioning stimulus is sufficient to produce a identifiable body surface temperature response. When a candidate air conditioning stimulus is confirmed as a valid air conditioning stimulus, it indicates that the execution feedback is continuous and effective, the signal-to-noise ratio of the regional stimulus input reaches a preset threshold, no second significant change in air conditioning execution occurs within the baseline window and response window, and the change in body surface temperature within the response window exceeds a preset measurement noise threshold and is consistent with the regional stimulus input in time and direction of action. In this case, the excitation quality is set to a non-zero value. When a candidate air conditioning stimulus is not confirmed as a valid air conditioning stimulus, the excitation quality is set to zero, and the current sample is not included in the identification.
[0093] In other embodiments, an effective air conditioning input can be constructed based on at least one feedback quantity among the actual power of the compressor or heat pump, the actual speed or air volume of the fan, the actual outlet air temperature, the position of the zone valve, the seat heating current, and the hot zone air supply status. The excitation quality is non-zero only when its change amplitude reaches the excitation threshold, the duration reaches the preset requirement, the input signal-to-noise ratio reaches the threshold, and the actual execution direction is consistent with the control command. Here, the effective air conditioning input in this section refers to the aforementioned zone stimulus input, which is an input quantity constructed from the actual execution data of the actuator to characterize the actual thermal effect intensity; the input signal-to-noise ratio in this section refers to the signal-to-noise ratio of the zone stimulus input; the calibration reference values for the excitation threshold and duration are the same as those described in Embodiment 1; the input signal-to-noise ratio reaching the threshold means that the ratio of the useful signal component to the noise component in the effective air conditioning input reaches the preset requirement, ensuring that its change can be distinguished from the measurement noise.
[0094] Step S250: If no window opening, door opening, or passenger seat change occurs within the baseline window and response window, the context quality is determined to be a valid value; otherwise, it is determined to be a invalid value.
[0095] Specifically, context quality is used to characterize whether there are contextual events affecting the body surface temperature response within the baseline window and response window. Opening a window or door introduces outside air, causing rapid changes in cabin temperature that overlap with changes in body surface temperature caused by air conditioning stimulation, making it difficult to attribute the body surface temperature response solely to air conditioning stimulation. Occupant relocation can cause body surface temperature data originally belonging to one occupant to be incorrectly associated with another. Therefore, when no window, door, or occupant relocation occurs within the baseline window and response window, the context quality is considered valid; otherwise, it is considered invalid, and the current sample is not used for identification and updating.
[0096] Step S260: When the outside temperature, sunlight, vehicle speed, number of occupants or other heat sources inside the cabin within the baseline window and response window do not change by more than a preset threshold, or when the change can be eliminated by a pre-configured environmental compensation item, the disturbance quality is determined to be an effective value; otherwise, the disturbance quality is reduced or set to zero.
[0097] Specifically, disturbance quality is used to characterize the degree of environmental disturbance within the baseline and response windows. Changes in outside temperature, sunlight, vehicle speed, number of occupants, or other heat sources within the cabin can all affect occupant body surface temperature. For example, sudden changes in sunlight intensity when the vehicle enters a tunnel, sudden changes in vehicle speed during acceleration, or changes in the number of occupants due to the entry of other occupants into the cabin can all lead to changes in body surface temperature, thereby interfering with the extraction of the body surface temperature response to air conditioning stimulation.
[0098] When the aforementioned environmental factors do not undergo abrupt changes exceeding a preset threshold, the disturbance quality takes an effective value. When abrupt changes exceeding the preset threshold occur, if the abrupt change can be eliminated by a pre-configured environmental compensation item, the disturbance quality can still take an effective value. That is, gradual changes that can be compensated by the environmental compensation item are retained, while abrupt changes that cannot be compensated are reduced or set to zero. For example, the gradual increase in outside temperature as the vehicle moves is a gradual change that can be compensated by the outside temperature component in the environmental compensation item, and the disturbance quality can be retained. However, the sudden drop in solar radiation intensity when the vehicle enters a tunnel is an abrupt change that is difficult to completely eliminate by the environmental compensation item. In this case, the disturbance quality should be reduced or set to zero.
[0099] Sample quality can be expressed as: ; Where, q sample For sample quality; q id For identity confidence; q align Register confidence levels for hot zones; q temp For body surface temperature measurement quality; q excitation To incentivize quality; q context For context quality; q disturbThis represents the perturbation quality. The above summary relationship is in product form. When any of the six quality dimensions takes a zero value or a low value, the sample quality decreases accordingly. For example, when the stimulus quality takes a zero value, regardless of the quality of other dimensions, the sample quality is zero, and the sample will not be used for identification.
[0100] Update threshold q up The thresholds can be calibrated to 0.60–0.80, the identity confidence threshold to 0.80–0.95, and the hot zone registration threshold to 0.70–0.90. Each threshold can be calibrated based on sensor noise and vehicle thermal inertia, and is not the sole value for the protection range. These thresholds are used to determine whether the sample quality has reached a level suitable for updating individual dynamic thermal response parameters, whether the identity confidence level is credible, and whether the hot zone registration confidence level is credible. Specific values can be calibrated based on sensor accuracy, cabin thermal inertia, and project performance targets.
[0101] In some embodiments, the method further includes the following steps.
[0102] Step A300: Samples whose quality reaches the update threshold after execution are stored in the occupant's long-term account; samples whose quality is below the update threshold but not below the preset temporary use threshold are used only for the current driving session; samples whose quality is below the temporary use threshold are discarded.
[0103] Specifically, sample tiered utilization refers to applying different processing strategies to samples based on their quality. Samples whose quality meets the update threshold, satisfying high requirements in terms of identity matching, thermal zone registration, body surface temperature measurement, stimulus effectiveness, contextual effectiveness, and perturbation level, can be used to update individual dynamic thermal response parameters and stored in the occupant's long-term account. Samples whose quality is below the update threshold but not below the preset temporary use threshold are only used in the current driving session, meaning they can be temporarily used to assist in prediction or control during the current driving session and are not retained after the driving session ends. Samples whose quality is below the temporary use threshold are discarded. Through this tiered utilization strategy, it is ensured that the samples stored in the long-term account are of high quality, that samples of acceptable quality can play a role in the current driving session, and that low-quality samples do not contaminate individual dynamic thermal response parameters.
[0104] When the gating conditions are not met, i.e. the sample quality does not reach the preset quality threshold or the preset update threshold, the individual dynamic thermal response parameters, covariance, and confidence level remain unchanged. It is not possible to freeze the individual dynamic thermal response parameters and continue to reduce the covariance, so as to avoid the covariance being artificially reduced, resulting in the estimation gain being too small during subsequent updates, and the individual dynamic thermal response parameters being unable to effectively respond to new observation data.
[0105] This embodiment comprehensively determines sample quality from six dimensions: identity binding, hot zone registration, body surface temperature measurement, stimulus effectiveness, context effectiveness, and perturbation level. Samples are then classified according to quality and stored in a long-term account, used only in the current driving session, or discarded. This reduces the risk of model contamination caused by identity mismatch, occlusion, weak stimulus, or environmental abrupt changes, while allowing samples of acceptable quality to play a role in the current driving session.
[0106] Example 5: In some embodiments, the future human thermal state under the action of each candidate control sequence is recursively predicted using individual dynamic thermal response parameters, including the following steps.
[0107] Step S310: For each candidate control sequence, the surface temperature, surface temperature change rate, and regional temperature difference of each human body region are recursively predicted using individual dynamic thermal response parameters.
[0108] Specifically, a candidate control sequence refers to a sequence of candidate values for air conditioning control variables within the prediction time domain. These control variables may include at least one of the following: set temperature, compressor or heat pump power, airflow, outlet air temperature, airflow direction, internal / external circulation, zone valves, and seat heating or ventilation. For example, a sequence where the set temperature is 24°C, 24°C, 23°C, and 23°C for four future time periods. When zone adjustment capability is available, the airflow, airflow direction, zone valves, or seat heating and ventilation of the target seat can be adjusted. Which air conditioning control variables can be independently adjusted depends on the vehicle configuration; outlet air temperature or cooling capacity that cannot be independently controlled are not considered independent control variables for each seat. Recursive prediction refers to using the current body surface temperature as a basis and gradually calculating the body surface temperature, body surface temperature change rate, and regional temperature difference at future times using the individual dynamic thermal response parameters identified in Example 3. Specifically, the air conditioning control quantity at each time step in the candidate control sequence can be converted into the corresponding regional stimulus input, substituted into a first-order model with pure delay, and recursively calculated at each time step to obtain the body surface temperature offset. Then, the baseline temperature is superimposed to obtain the body surface temperature at each future time step. The body surface temperature change rate can be obtained by dividing the difference in body surface temperature between adjacent times by the sampling period. The regional temperature difference can be obtained by the difference in body surface temperature between different human body regions, such as the temperature difference between the face and the hands. Since the input normalized response gain, response delay, and time constant are different for different occupants, the future body surface temperature trajectory obtained by recursively calculating the same candidate control sequence for different occupants will also be different. Therefore, recursive prediction can reflect the dynamic thermal response characteristics of individual occupants.
[0109] Step S320: Based on the deviation of body surface temperature from a predetermined individual thermal neutrality baseline, the rate of change of body surface temperature, the regional temperature difference, and the environmental heat load, determine the predicted thermal comfort index used to characterize the individual's thermal comfort state.
[0110] Specifically, the predicted thermal comfort index can be expressed as: ; in, c is the predicted thermal comfort index for occupant i at time k; T This is the weighting coefficient for body surface temperature deviation; Let be the overall surface temperature of occupant i at time k; c serves as the individual thermal neutrality baseline for occupant i; v This is the weighting coefficient for the rate of change of body surface temperature. Let c be the rate of change of body surface temperature of occupant i at time k; d This is the regional temperature difference weighting coefficient; Let c be the temperature difference in the region for occupant i at time k; E E(k) is the environment vector weight coefficient vector; E(k) is the environment vector at time k.
[0111] The individual thermal neutrality baseline refers to the reference value of body surface temperature when an occupant feels neither too hot nor too cold. It can be determined in advance based on the occupant's historical comfort data. For example, after an occupant has repeatedly adjusted the air conditioning to a certain state and the body surface temperature has stabilized, the overall body surface temperature under this stable state is recorded as the individual thermal neutrality baseline. Body surface temperature deviation characterizes the degree to which the current body surface temperature deviates from the individual thermal neutrality baseline; a positive deviation indicates that it is too hot, and a negative deviation indicates that it is too cold. The body surface temperature change rate characterizes the trend of body surface temperature change. Even if the current individual thermal comfort state has not deviated from the preset comfort zone, a large change rate indicates that the occupant will soon feel uncomfortable. Regional temperature difference characterizes the temperature difference between different areas of the human body. For example, if the temperature difference between the face and the hands is too large, even if the individual thermal comfort state corresponding to the overall body surface temperature is within the preset comfort zone, the occupant may still feel local discomfort. Environmental heat load is applied to the predicted thermal comfort index through environmental vector weighting coefficients.
[0112] Predicted thermal comfort index S i S is a normalized index. i Below the lower boundary of the preset comfort zone indicates it's too cold, within the preset comfort zone indicates it's comfortable, and above the upper boundary of the preset comfort zone indicates it's too hot. The preset comfort zone can be set to a normalized value of -0.3 to 0.3, i.e., S. i A value between -0.3 and 0.3 is considered to indicate that the occupants are in a comfortable state. These values are for calibration reference only and are not the sole determining factor for the protection range.
[0113] Step S330: Based on the predicted thermal comfort index, air conditioning control quantity change and energy consumption corresponding to each candidate control sequence, determine the comfort deviation, control change and energy consumption corresponding to each candidate control sequence.
[0114] Specifically, comfort deviation refers to the deviation between the predicted thermal comfort index and the target thermal comfort state. The target thermal comfort state can be set to zero, that is, the occupants are expected to be in a neutral state of neither being too hot nor too cold. It can also be set to a value that is slightly colder or slightly warmer according to the occupants' preferences. Control change refers to the change in air conditioning control quantity between adjacent moments, such as the change in set temperature, the change in compressor power, or the change in air volume. Energy consumption refers to the energy consumption of the air conditioning system under the action of the candidate control sequence, which can be estimated based on the power of the compressor or heat pump, the power of the fan, and the power of seat heating or ventilation.
[0115] Step S340: With the goal of minimizing the weighted sum of comfort deviation, control variation and energy consumption, solve the air conditioning control quantities corresponding to each candidate control sequence under safety constraints, and determine the candidate control sequence with the minimum weighted sum as the target candidate control sequence.
[0116] Specifically, the objective function can be expressed as: ; Where J is the objective function value; H is the prediction time domain length; h is the time index within the prediction time domain; w s This is the comfort deviation weighting coefficient; The predicted thermal comfort index of passenger i at time k+h is obtained according to the aforementioned recursive prediction. For target thermal comfort state; w u To control the changing weighting coefficients; Let w be the change in the air conditioning control quantity at time k+h; e Energy consumption weighting coefficient; Let K be the energy consumption of the air conditioning system at time k+h.
[0117] The objective function J consists of a weighted sum of three terms. The first term... The first term is the comfort deviation term, representing the sum of squared deviations between the predicted thermal comfort index and the target thermal comfort state; the second term... The control term represents the sum of squares of the changes in the air conditioning control quantity. A smaller term indicates a smoother change in the air conditioning control quantity, which helps suppress frequent oscillations in the air conditioning control quantity; the third term... This is the energy consumption item, representing the energy consumption of the air conditioning system. By adjusting w... s w u and w e The three weighting coefficients allow for a trade-off between comfort, smoothness of control, and energy consumption. For example, the weighting coefficient can be increased when occupants are more sensitive to temperature fluctuations. s When energy consumption needs to be reduced, the power consumption can be increased by w. e .
[0118] Specifically, safety constraints may include upper and lower limit constraints on air conditioning control quantities, rate of change constraints, and safety protection constraints. The upper and lower limit constraints on air conditioning control quantities mean that air conditioning control quantities such as set temperature, compressor power, and air volume must be within the allowable range. For example, the set temperature must be between 16°C and 30°C. The rate of change constraint means that the change in air conditioning control quantities within a unit of time must be within the allowable range. The safety protection constraint means that when anti-fogging defrosting, power battery thermal management, or driving safety requirements are triggered, the air conditioning control quantities must prioritize meeting the safety requirements.
[0119] Under safety constraints, the air conditioning control variables corresponding to each candidate control sequence are solved separately. That is, for each candidate control sequence, its objective function value J is calculated under the premise of satisfying the safety constraints, and the candidate control sequence with the smallest objective function value J is determined as the target candidate control sequence. Since the objective function J takes into account comfort deviation, control variation and energy consumption, the target candidate control sequence can ensure passenger comfort while taking into account control smoothness and energy consumption.
[0120] Step S350: Apply preset dead zone, minimum hold time and rate of change limits to the air conditioning control quantity corresponding to the target candidate control sequence, and execute the limited air conditioning control quantity.
[0121] Specifically, the dead zone refers to the situation where the air conditioning control quantity will not be adjusted when the change is less than a preset range. For example, if the difference between the set temperature corresponding to the target candidate control sequence and the current set temperature is less than 0.5℃, no adjustment will be made to avoid frequent actions of the air conditioning control quantity due to small deviations. The minimum hold time refers to the shortest time that the air conditioning control quantity must remain after adjustment, for example, it can be 30 to 120 seconds. That is, the air conditioning control quantity must remain for at least this duration after adjustment before it is allowed to be adjusted again. The rate of change limit refers to the maximum change of the air conditioning control quantity per unit time, for example, the change of the set temperature per minute must not exceed 1℃. The above values are for calibration reference and are not the only values for the protection range.
[0122] The restricted air conditioning control parameters are executed by the air conditioning controller. During execution, the air conditioning controller controls the compressor or heat pump, fan, damper, zone valve, and seat heating or ventilation actuators according to the restricted air conditioning control parameters to output the corresponding heat effect to the cabin.
[0123] This embodiment utilizes individual dynamic thermal response parameters to recursively predict body surface temperature, rate of change of body surface temperature, and regional temperature difference under each candidate control sequence, and determines the predicted thermal comfort index accordingly. This allows control decisions to be based on predictions of future human thermal states, rather than solely on feedback control based on current body surface temperature. Since individual dynamic thermal response parameters reflect the occupant's input normalized response gain, response delay, and time constant, for occupants with large time constants, it avoids continuously increasing air conditioning power due to a short period without observed response; for occupants with large input normalized response gain, it allows for earlier reduction of air conditioning power to avoid overcooling or overheating. By using the weighted sum and minimum of comfort deviation, control variation, and energy consumption as the optimization objective, and imposing dead zone, minimum hold time, and rate of change constraints, it balances comfort, control smoothness, and energy consumption, suppressing frequent oscillations in the control quantity.
[0124] In some embodiments, a preset stability threshold is used to determine whether the absolute value of the rate of change of body surface temperature meets the stability condition, and can be taken as 0.01 to 0.05 °C / min. When the absolute value of the rate of change of body surface temperature is lower than the preset stability threshold, the body surface temperature is considered to have stabilized. The above values are for calibration reference and are not the only values for the protection range.
[0125] Example 6: In some embodiments, the method further includes the generation, verification, state management, and contextualized invocation of verification-based preference memories. For example... Figure 4 As shown in the figure, this diagram illustrates the relationship between double verification and state transition in verification-based preference memory.
[0126] Step A400: When an occupant actively changes the air conditioning control setting, the environmental conditions, body surface temperature, and individual thermal comfort status before and after the change are recorded, and candidate preferences are generated.
[0127] Specifically, occupant-initiated changes to air conditioning control settings refer to occupants actively adjusting settings such as temperature, airflow, airflow direction, or operating mode via the air conditioning control panel, central touchscreen, voice commands, or gestures. Unlike the aforementioned automatic determination and execution of air conditioning control settings by the controller, the occupant-initiated changes and candidate preferences recorded in this step are associated with the occupant's identifier and target seat. When multiple passengers are traveling together, the actions of one occupant are not directly written into the long-term preference memory of other occupants. The system records the environmental conditions, body surface temperature, and individual thermal comfort status before and after the change. The environmental conditions may include the cabin temperature, outside temperature, humidity, sunlight intensity, vehicle speed, and window opening status before and after the change. The body surface temperature may include the body surface temperature of each body area before and after the change, as well as the overall body surface temperature. The individual thermal comfort status is characterized by thermal comfort indices obtained by calculating the measured body surface temperature, body surface temperature change rate, regional temperature difference, and environmental data before and after the change according to the index calculation relationship in Example 5. The index obtained by substituting the recursive prediction data into this calculation relationship is called the predicted thermal comfort index. The above recorded content is used for subsequent verification of candidate preferences.
[0128] Step A410: When the identity confidence and thermal zone registration confidence are determined to be valid based on the binding relationship between the occupant identification, the target seat and the human body thermal sensing area, and the body surface temperature measurement quality is valid and the sample quality reaches the preset quality threshold, the candidate preference is verified. Among them, the change is not triggered by defogging, air quality, noise limitation, safety protection or temporary mandatory demand, the changed air conditioning control quantity is maintained for the preset verification time, and the change amplitude of the regional stimulus input reaches the preset preference verification threshold.
[0129] Specifically, valid identity confidence and thermal registration confidence mean that the binding relationship between the occupant identifier and the target seat and the human body thermal sensing area is not interrupted, and both reach the identity confidence threshold and thermal registration threshold described in Example 4, respectively; valid body surface temperature measurement quality means that the proportion of valid pixels and the number of valid areas in the body surface temperature of each human body area meet the requirements described in Example 4; and sample quality reaching the preset quality threshold means that the sample quality calculated according to the summary relationship described in Example 4 reaches the preset quality threshold.
[0130] The changes are not triggered by defogging, air quality, noise restrictions, safety protection, or temporary mandatory needs. Specifically, when a defogging need is triggered, occupants may adjust the airflow direction to blow towards the windshield for defogging purposes, not to adjust thermal comfort; when an air quality need is triggered, occupants may switch to recirculation to reduce the entry of external pollutants; when a noise restriction is triggered, occupants may reduce the airflow to reduce fan noise; when a safety protection is triggered, the system may forcibly adjust the air conditioning operation to protect the battery or compressor, an operation not initiated by the occupants; when a temporary mandatory need is triggered, occupants may quickly adjust the air conditioning control for temporary needs such as getting in and out of the vehicle or making phone calls, an operation that does not represent the occupants' long-term preferences. Changes in air conditioning control in the above situations are not used to generate candidate preferences, thus avoiding misjudging non-thermal comfort-driven operations as preferences.
[0131] The modified air conditioning control setting should be maintained for a preset verification time. Specifically, if the occupant adjusts the air conditioning control setting again shortly after the change, it indicates that the change may be temporary or exploratory and is insufficient as a basis for confirming preferences. The preset verification time can be calibrated based on the vehicle model's thermal inertia and the occupant's operating habits; for example, it can be set to 120–600 seconds.
[0132] The amplitude of the change in the regional stimulus input reaches the preset preference verification threshold. Specifically, after an occupant changes the air conditioning control quantity, the air conditioning actuator needs to actually execute the change and generate a thermal effect sufficient to cause a body surface temperature response. If the amplitude of the change in the regional stimulus input after the change does not reach the preset preference verification threshold, it indicates that the actual thermal effect is insufficient to generate a identifiable body surface temperature response, and the candidate preference cannot be verified accordingly. The preset preference verification threshold can be calibrated with reference to the excitation threshold described in Example 2.
[0133] The verification of candidate preferences includes physiological thermal response verification and user behavior verification. When an individual's thermal comfort state remains within the preset comfort zone within the preset verification window and the absolute value of the rate of change of body surface temperature is lower than the preset stability threshold, the physiological thermal response verification is deemed successful. When no change in air conditioning control quantity opposite to the direction of change or further correction of the change is detected within the preset observation window, the user behavior verification is deemed successful.
[0134] Specifically, physiological thermal response verification is used to determine whether the occupant's actions have produced the expected physiological thermal response. The preset verification window, as described in Example 1, is used to confirm whether the occupant has entered a stable comfort state. The individual's thermal comfort state remaining within the preset comfort zone means that within the stable verification window, the thermal comfort index calculated using the measured body surface temperature, body surface temperature change rate, and regional temperature difference data within that window, according to the determination method described in Example 5, remains within the preset comfort zone. For example, the normalized index is set to -0.3 to 0.3, meaning this determination is based on the measured body surface temperature response rather than the predicted result. The absolute value of the body surface temperature change rate being lower than the preset stability threshold means that the absolute value of the measured body surface temperature change rate is lower than the preset stability threshold, indicating that the occupant's body surface temperature has stabilized. The preset stability threshold can be 0.01–0.05 °C / min; these values are calibration references and not the sole value for the protection range.
[0135] User behavior verification is used to determine whether the occupant accepts the control result through subsequent actions. Specifically, it checks if the pre-defined observation window detects no change in the air conditioning control quantity in the opposite direction of the change, or no further correction to the change. A change in the air conditioning control quantity in the opposite direction refers to the occupant performing an operation in the opposite direction after changing the air conditioning control quantity, such as adjusting the set temperature from 26℃ to 24℃, and then adjusting it back to 26℃ or to 25℃ in the observation window. A further correction to the change refers to the occupant adjusting the same air conditioning control quantity in the same direction again, such as adjusting the set temperature from 26℃ to 24℃ and then to 23℃. Both of these situations indicate that the occupant is not satisfied with the change result, and the user behavior verification fails.
[0136] Physiological thermal response verification and user behavior verification constitute dual verification. If there is only occupant operation without a stable thermal response, it indicates that the occupant's operation may not have produced the expected physiological effect, for example, because the air conditioning actuator did not actually operate or environmental disturbances offset the thermal effect. If a stable response occurs but the occupant subsequently adjusts in the opposite direction, it indicates that the occupant is not satisfied with the control result. Neither of these two scenarios meets the dual verification requirement, and the candidate preference is not confirmed.
[0137] Step A420: When both physiological thermal response verification and user behavior verification are passed, and the candidate preference reaches the preset high confidence level in a single verification or the number of times both verifications are passed in similar environments and occupant states reaches the preset number of confirmations, the candidate preference is written into the occupant's long-term preference memory; otherwise, it is not written into the long-term preference memory, and the confidence level of the candidate preference is reduced.
[0138] Specifically, the preset high-confidence condition refers to a high level of confidence in both physiological thermal response verification and user behavior verification in a single verification. For example, within the stable verification window, the individual's thermal comfort state remains in the central area of the preset comfort zone rather than the edge area, the absolute value of the rate of change of body surface temperature is much lower than the preset stability threshold, and no reverse operation or further correction is detected within the observation window. The preset number of confirmations refers to the cumulative number of times that both physiological thermal response verification and user behavior verification have passed under similar environmental and occupant conditions, reaching a preset value. The number of confirmations can be 2 to 5, or it can be replaced by a single high-confidence verification. The confidence threshold in the preset high-confidence condition can be 0.70 to 0.90. The above values are for calibration reference and are not the only values for the protection scope.
[0139] If a candidate preference fails double verification, or if it passes double verification but fails to meet the preset high-confidence condition and the number of times both verifications pass does not reach the preset confirmation count, it will not be written into the long-term preference memory, and the confidence level of the candidate preference will be reduced. The purpose of reducing the confidence level is that if the same candidate preference is triggered again subsequently, its lower confidence level will make it more difficult to meet the preset high-confidence condition, and it will be used with lower weight when invoked as a control prior, thereby reducing the probability of unreliable preferences affecting air conditioning control.
[0140] In step A500, candidate preferences are managed using four states: candidate, confirmed, decayed, and invalid.
[0141] Specifically, such as Figure 4 As shown, candidate preferences are initially in the candidate state. When a candidate preference meets the conditions for being written into long-term preference memory, it is converted to the confirmed state. When a candidate preference has not been verified for a long time or the applicable scenario has changed, it is converted to the decay state and the confidence of the candidate preference is reduced. When it is continuously detected that the occupant has implemented an air conditioning control change in the opposite direction of the change, the occupant identification no longer matches, or the candidate preference has been inapplicable for a long time, the candidate preference is converted to the invalid state.
[0142] A candidate state refers to a state where a candidate preference has been generated but not yet written into long-term preference memory, i.e., a state that has not yet passed dual verification, or a state that has passed dual verification but has not yet met the preset high confidence condition and the number of times both verifications have passed has not reached the preset confirmation number. In this case, the candidate preference can be used as an auxiliary reference in the current driving session, but is not used as a long-term preference. A confirmed state refers to a state where the candidate preference has passed dual verification and been written into long-term preference memory. The preference record can be used as the initial target or control prior of the candidate control sequence. A decay state refers to a state where the confirmed preference record has been decayed due to long-term lack of re-verification or changes in the applicable scenario. The confidence level decreases in several states, including: Long-term non-verification refers to a preference record not being triggered and verified again within a preset time frame, such as exceeding one season or a preset number of days; Change in applicable scenario refers to a significant change in the environmental state corresponding to the preference record, such as the external temperature range changing from a summer high-temperature range to a winter low-temperature range; Invalidation refers to a preference record becoming invalid due to continuous reverse adjustments, identity mismatch, or long-term inapplicability, such as a preference record set at 24℃ but repeatedly adjusted back to 26℃ by the occupant, or the occupant identifier associated with the preference record being inconsistent with the current occupant identifier. In these states, the preference record is no longer invoked. In the decay state, the confidence level of the preference record decreases, and its weight as a control prior decreases accordingly.
[0143] Step A510 involves associating and storing the preference records in long-term preference memory with occupant identification, seat, environmental state, most recent verification time, verification result, and confidence level, and then calling the preference records according to scene similarity as the initial target or control prior of the candidate control sequence.
[0144] Specifically, the preference record includes at least the occupant identifier, seat, set temperature, air volume, air direction, operating mode, outside temperature range, sunlight range, initial cabin temperature, season, sample size, most recent verification time, verification result, and confidence level. The occupant identifier is used to identify the occupant to whom the preference record belongs, the seat is used to identify the seat location corresponding to the preference record, the most recent verification time is used to determine whether the preference record needs to be re-verified, the verification result is used to record whether the most recent verification passed, and the confidence level is used to characterize the reliability of the preference record.
[0145] The system invokes preference records based on scene similarity. This means determining the similarity between the current scene and a preferred record based on scene elements such as outside temperature range, sunlight range, season, initial cabin temperature, and seating position. When the similarity reaches a preset threshold, the preferred record is invoked. Specifically, the outside temperature range can be divided into high, medium, and low temperatures; the sunlight range can be divided into strong sunlight, weak sunlight, and no sunlight; the season can be determined based on the current date; the initial cabin temperature can be determined based on the cabin temperature when the occupants enter the cabin; and the seating position can be determined based on the occupants' seating location. Scene similarity is calculated by weighting the matching degree of each scene element. For example, if all scene elements of the current scene match a particular preference record, the scene similarity is high, and that preference record can be invoked; if only some scene elements match, the scene similarity is low, and the preference record's weight can be reduced or it can be omitted.
[0146] Using preference records as the initial target or control prior for candidate control sequences means using the set temperature, air volume, air direction, or operating mode in the preference records as the initial values of the candidate control sequences in Example 5, so as to reduce the search range of candidate control sequences and speed up the solution, or to use preference records as the initial reference of the objective function so that the air conditioning control quantity obtained by the solution is more inclined to be close to the occupants' preferences.
[0147] Step A520: When the real-time body surface thermal response is inconsistent with the preference record, the air conditioning control quantity is determined first based on the real-time body surface thermal response and safety constraints, and the preference record is suspended or its usage weight is reduced.
[0148] Specifically, real-time body surface thermal response refers to the occupant's current thermal response characteristics reflected by the body surface temperature acquired at the current moment and the individual dynamic thermal response parameters identified in Example 3. The preference record is generated based on historical data. If the occupant's physiological state, clothing, or health condition changes, the real-time body surface thermal response may differ from the preference record. For example, the preference record might show an occupant preferring a set temperature of 24°C in summer, but the occupant might currently be more sensitive to low temperatures due to discomfort, and the real-time body surface thermal response might show a rapid drop in body surface temperature. In this case, if the 24°C setting is still executed according to the preference record, the occupant might feel too cold. Therefore, when the real-time body surface thermal response differs from the preference record, the air conditioning control quantity is determined primarily based on the real-time body surface thermal response and safety constraints, and the preference record is either suspended or its weight is reduced. Safety constraints may include anti-fogging and defrosting, power battery thermal management, and driving safety requirements.
[0149] This embodiment first saves the occupant's active adjustments as candidate preferences, and then writes them into the long-term preference memory only after dual verification through physiological thermal response verification and user behavior verification. This ensures that the preferences written into the long-term preference memory have been confirmed by the stability and comfort of the human body state and the subsequent behavior of the occupant, reducing the probability of writing temporary needs or misoperations into the long-term preferences. The preference is managed through four states: candidate, confirmed, decayed, and invalid, so that the preference records can be dynamically adjusted with the time and scenario changes, avoiding the adverse effects of outdated or inapplicable preferences on the current control.
[0150] Example 7: In some embodiments, the method is performed separately for occupants in multiple seats within the vehicle cabin, and the method further includes the following steps.
[0151] Step A600 involves identifying individual dynamic thermal response parameters for each of the multiple occupants and determining the individual thermal comfort state of each occupant.
[0152] Specifically, when there are multiple occupants in the vehicle cabin, such as the driver and front-seat passenger riding together, each occupant's seat is designated as the target seat. Individual dynamic thermal response parameters are identified for each occupant, and their individual thermal comfort state is determined. The predicted results for each occupant are used for subsequent coordination, avoiding conflicting control quantities issued to the same shared actuator. Local actuators with independent adjustment capabilities are controlled according to the needs of the occupants in their respective seats. The process of identifying individual dynamic thermal response parameters for multiple occupants is consistent with the identification process described in Example 3; that is, for each occupant, based on their corresponding regional stimulus input and the body surface temperature within the baseline and response windows, the input normalized response gain, response delay, and time constant are identified separately for cooling and heating. The process of determining the individual thermal comfort state for each occupant is consistent with the determination process described in Example 5. Since the input normalized response gain, response delay, and time constant may differ among different occupants, and individual thermal neutrality benchmarks may also differ, the impact of the same air conditioning control variable on the thermal comfort of different occupants may vary. Identifying and determining these separately allows the thermal comfort needs of each occupant to be assessed independently.
[0153] Step A610: When the individual thermal comfort states of multiple occupants deviate from the preset comfort zone and the direction of deviation is the same, the occupant with the most severe deviation shall control the shared cooling or heating capacity.
[0154] Specifically, the preset comfort zone refers to the range within which an individual's thermal comfort state falls within a comfortable range, for example, a normalized index of -0.3 to 0.3. Same deviation direction means that the individual thermal comfort states of multiple occupants are all biased towards the warmer side or the cooler side. For example, if the predicted thermal comfort indices of the driver and the front passenger are both positive and exceed the upper boundary of the preset comfort zone, it indicates that both occupants feel too hot, and the deviation direction is the same. The most severe deviation refers to the largest deviation of the predicted thermal comfort index from the preset comfort zone, for example, the driver S... A =+0.7, Front passenger S B When the value is +0.5, the driver's deviation is more severe. In this case, the shared cooling capacity is controlled according to the occupant with the most severe deviation. That is, the shared actuators, such as the compressor or heat pump power and shared air volume, are controlled according to the driver's needs to prioritize alleviating the discomfort of the occupant with the most severe deviation. When the individual thermal comfort states of multiple occupants are all biased towards the colder side, the shared heating capacity is controlled according to the occupant with the most severe deviation.
[0155] Step A620: When the individual thermal comfort states of multiple occupants deviate from the opposite direction, the shared cabin temperature is maintained in a preset compromise range that balances safety and energy consumption, and is adjusted preferentially through the corresponding local air volume, air direction, zone valves, seat heating or seat ventilation actuators.
[0156] Specifically, the opposite direction of deviation means that the individual thermal comfort state of some occupants is biased towards the warmer side while that of other occupants is biased towards the cooler side. For example, driver A's S... A =+0.7 while the S of the front passenger B B =-0.5. In this situation, increasing the shared cooling capacity to meet the driver's preference for warmth would exacerbate the discomfort of the front-seat occupants due to cold, and vice versa. Therefore, the shared cabin temperature is maintained within a preset compromise range that balances safety and energy consumption. This means not adjusting it to meet the extreme needs of any occupant, but keeping it within a safe and energy-efficient middle range. Adjustments are prioritized through corresponding local actuators. For example, to meet the driver's preference for warmth, the local airflow on the driver's side is increased, and the airflow's hit coefficient on the corresponding body area is improved; the driver's seat ventilation is activated. To meet the front-seat occupants' preference for cold, the airflow on the occupant's side is reduced or the airflow direction is changed; if necessary, the occupant seat heating is activated. This alleviates the discomfort of each occupant without significantly altering the shared cabin temperature. If a seat does not have a corresponding local actuator, coordination is only performed within the vehicle's existing adjustable actuator range, without requiring independent temperature adjustment for each seat.
[0157] If fogging of the windshield is detected during the above coordination process, the safety defogging strategy overrides the coordination results.
[0158] In some embodiments, the method further includes the following steps.
[0159] Step A700: Determine the operating status based on factors such as occupant change or seat change, human body area occlusion, body surface temperature measurement drift, insufficient regional stimulus input, sudden environmental changes, continuous exceeding of prediction error limits, or communication failure. The operating status includes normal status, suspicious status, and failure status. In the normal status, recursive prediction and individual dynamic thermal response parameter updates are performed. In the suspicious status, recursive prediction is maintained and the updates of individual dynamic thermal response parameters and corresponding covariance and confidence are frozen. In the failure status, predictive control based on individual dynamic thermal response parameters is exited and the system switches to the pre-configured basic air conditioning strategy.
[0160] Specifically, the normal state refers to a state in which the system does not detect any abnormalities that affect the identification and control reliability of individual dynamic thermal response parameters. Under the normal state, the system continues to execute the recursive prediction and air conditioning control quantity determination process described in Example 5, and continues to execute the individual dynamic thermal response parameter update process based on prediction error and sample quality gating described in Example 1.
[0161] A suspicious state refers to an anomaly detected by the system that may affect the reliability of individual dynamic thermal response parameter identification but has not yet led to complete control failure. In a suspicious state, the system continues to make recursive predictions and determine air conditioning control quantities, but freezes the updates of individual dynamic thermal response parameters and their corresponding covariance and confidence levels. This is because samples obtained in a suspicious state may be contaminated by anomalous factors; if still used to update individual dynamic thermal response parameters, it could cause these parameters to deviate from the occupant's true dynamic thermal response characteristics.
[0162] A failure state refers to an abnormal situation detected by the system that prevents predictive control based on individual dynamic thermal response parameters from being executed reliably. In a failure state, the system exits predictive control based on individual dynamic thermal response parameters and switches to a pre-configured basic air conditioning strategy, that is, an air conditioning control strategy that does not rely on individual dynamic thermal response parameters, such as feedback control based on the set temperature, control based on a preset temperature curve, or control based on a fixed air volume and direction.
[0163] The example criteria for Level 3 operating states and their correspondence with the quality factors in Example 4 are as follows. A failure state is determined when identity or seat continuity is interrupted, communication times out, or sensor self-diagnosis fails. The former corresponds to the failure of identity confidence and thermal zone registration confidence, at which point it is impossible to determine which occupant the body surface temperature data belongs to. The latter indicates that the system cannot obtain the necessary sensor data or actuator feedback data, and predictive control cannot be reliably executed. A suspicious state is determined when the effective pixel ratio, thermal zone registration confidence, or body surface temperature measurement quality is below the threshold, the regional stimulus input is insufficient, environmental changes such as opening windows or doors or sudden changes in sunlight occur, the prediction error continuously exceeds the limit, or the individual dynamic thermal response parameters exceed the limit. These situations correspond to the insufficient body surface temperature measurement quality, insufficient thermal zone registration confidence, zero excitation quality, invalid context quality, or reduced disturbance quality described in Example 4, as well as a decrease in the accuracy of the current individual dynamic thermal response parameters in characterizing the actual thermal response characteristics of the occupant, or an estimated value of the individual dynamic thermal response parameters exceeding a reasonable range. A normal state is determined when none of the above conditions are met.
[0164] All the aforementioned abnormal situations are handled according to the aforementioned correspondence: when determined to be a normal state, recursive prediction continues and individual dynamic thermal response parameters are updated; when determined to be a suspicious state, recursive prediction is retained but updates of individual dynamic thermal response parameters and their corresponding covariance and confidence levels are frozen; when determined to be a failure state, predictive control based on individual dynamic thermal response parameters is exited and the pre-configured basic air conditioning strategy is switched to. Specifically, in a suspicious state, the confidence level reduction of individual dynamic thermal response parameters and the rollback of population parameters are not implemented; updates are resumed only after the anomaly is eliminated and the sample quality returns to the preset update threshold.
[0165] In step A710, the needs for anti-fog defrosting, power battery thermal management, and driving safety take precedence over the individual comfort needs of multiple occupants.
[0166] Specifically, the needs for anti-fogging and defrosting refer to the need to adjust the air conditioning direction, airflow, and temperature to quickly eliminate fog or frost when fog or frost appears on the windshield or side windows, ensuring clear driving visibility. The needs for power battery thermal management refer to the need to use the air conditioning system to cool or heat the power battery when its temperature is too high or too low, ensuring its performance and lifespan. Driving safety needs refer to air conditioning control requirements directly related to driving safety. All of these needs take precedence over the individual comfort needs of multiple occupants. That is, when these needs are triggered, the air conditioning control prioritizes meeting the needs for anti-fogging and defrosting, power battery thermal management, or driving safety, and only then considers the individual comfort needs of multiple occupants.
[0167] In some embodiments, the minimum holding time of the air conditioning control quantity can be 30 to 120 seconds. The above values are calibration references and are not the only values for the protection range. Specifically, in the process of multi-occupant coordinated control, when the shared cooling capacity or shared heating capacity is adjusted, the minimum holding time of the air conditioning control quantity is used to avoid frequent switching of the shared actuator between the needs of multiple occupants. For example, if the driver feels too hot and the front passenger feels too cold, frequent adjustments to the shared cooling capacity will cause fluctuations in the cabin temperature and exacerbate the discomfort of the two occupants. Therefore, the shared actuator should be held for at least 30 to 120 seconds after adjustment to help maintain a relatively stable cabin thermal environment.
[0168] This embodiment identifies individual dynamic thermal response parameters and determines individual thermal comfort states for multiple occupants, enabling independent assessment of each occupant's thermal comfort needs in multi-occupant scenarios and providing an individualized model foundation for differentiated comfort control. By controlling the shared actuator according to the occupant with the most severe deviation when demand directions are the same, priority is given to alleviating the most severe discomfort. By maintaining the shared cabin temperature within a compromise range and prioritizing the use of local actuators when demand directions are opposite, the adverse effects on other occupants caused by shared actuator adjustments are reduced. By setting three operating states—normal, suspicious, and failed—and adopting different control and learning strategies in different states, the system's safety and reliability under abnormal conditions are improved. In the suspicious state, freezing individual dynamic thermal response parameters and corresponding covariance and confidence updates avoids contamination of individual dynamic thermal response parameters by abnormal samples. In the failed state, switching to the basic air conditioning strategy ensures the system still provides basic air conditioning control functions. By prioritizing anti-fogging defrosting, power battery thermal management, and driving safety needs over personalized comfort needs, safety-related functions are always given priority.
[0169] Example 8: like Figure 5 As shown, this embodiment provides an air conditioning control system, including a stimulus detection module 1, a gate identification module 2, a prediction control module 3, and an update module 4.
[0170] Stimulus detection module 1 is used to acquire the surface temperature of each human body area of the occupant in the target seat and the actual execution data of the air conditioning actuator. Based on the actual execution data, a directional normalized regional stimulus input is constructed to act on the occupant. When the change amplitude of the regional stimulus input reaches the preset excitation threshold and lasts for a preset time, and the actual execution direction is consistent with the control command, candidate air conditioning stimuli are determined and a baseline window before stimulation and a response window after stimulation are established.
[0171] Specifically, the stimulus detection module 1 is the system's input-side module, responsible for acquiring the body surface temperature and related actual execution data of the occupant in the target seat, and constructing a regional stimulus input acting on the occupant. The body surface temperature of each human body region can be acquired through infrared array sensors, thermal imaging sensors, contact temperature sensors, or multi-source fusion. Human body regions may include thermal sensing areas such as the face, forehead, neck, chest, and hands. Actual execution data may include the actual power of the compressor or heat pump, the actual speed or airflow of the fan, the actual outlet air temperature, the airflow direction or zone valve position, the seat heating current, and the hot zone air supply status, etc., with the same meaning as described in Embodiment 1. Based on this, the stimulus detection module 1 constructs a directional, normalized regional stimulus input acting on the occupant in the target seat, and determines candidate air conditioning stimuli when the amplitude of the regional stimulus input reaches a preset excitation threshold and remains there for a preset time, and the actual execution direction is consistent with the control command. A baseline window before stimulation and a response window after stimulation are then established.
[0172] The gating identification module 2 is used to summarize the identity confidence, thermal zone registration confidence, body surface temperature measurement quality, excitation quality, context quality, and perturbation quality according to the preset summary relationship when the candidate air conditioning stimulus is confirmed as a valid air conditioning stimulus, so as to obtain the sample quality. When the sample quality reaches the preset quality threshold, based on the body surface temperature in the region stimulus input and the baseline window and response window, the individual dynamic thermal response parameters are identified separately for cooling and heating, and the corresponding covariance and confidence are obtained. The individual dynamic thermal response parameters include the input normalized response gain, response delay, and time constant.
[0173] Specifically, the gating identification module 2 is the identification module of the system, responsible for identifying the dynamic thermal response characteristics of the occupant in the target seat when the sample quality meets the conditions. The conditions for confirming a candidate air conditioning stimulus as a valid air conditioning stimulus, the determination method of the six quality factors in the sample quality, the meaning of each parameter in the individual dynamic thermal response parameters, and the meaning of covariance and confidence level are the same as those described in Examples 1 to 4. When the sample quality reaches the preset quality threshold, the gating identification module 2 identifies the individual dynamic thermal response parameters according to cooling and heating based on the regional stimulus input and the body surface temperature within the baseline window and response window, and obtains the corresponding covariance and confidence level.
[0174] The predictive control module 3 is used to recursively predict the future human thermal state under the action of each candidate control sequence using individual dynamic thermal response parameters, and to determine and execute air conditioning control quantities.
[0175] Specifically, the predictive control module 3 is the system's control output module, responsible for predicting and solving the air conditioning control quantity using the individual dynamic thermal response parameters identified by the gating identification module 2. For each candidate control sequence, the predictive control module 3 uses the individual dynamic thermal response parameters to recursively predict the surface temperature, surface temperature change rate, and regional temperature difference of each body region of the occupant in the target seat. Based on this, it determines the predicted thermal comfort index used to characterize the individual's thermal comfort state. With the goal of minimizing the weighted sum of comfort deviation, control change, and energy consumption, it solves the air conditioning control quantity under safety constraints. Then, it applies preset dead zone, minimum hold time, and change rate limits to the air conditioning control quantity corresponding to the target candidate control sequence before execution. The specific process is the same as described in Example 5.
[0176] The update module 4 is used to determine the prediction error based on the actual body surface response after execution and to re-determine the sample quality. When the sample quality reaches the preset update threshold, the prediction error is used to update the individual dynamic thermal response parameters and the corresponding covariance and confidence level; otherwise, they remain unchanged.
[0177] Specifically, update module 4 is the system's feedback update module, responsible for forming a closed-loop correction of individual dynamic thermal response parameters based on the actual body surface response after execution. After executing the air conditioning control quantity, update module 4 obtains the actual body surface response of the occupant in the target seat, compares it with the future human thermal state recursively predicted by predictive control module 3 to determine the prediction error, and re-determines the sample quality based on the actual body surface response after execution. When the sample quality reaches a preset update threshold, the individual dynamic thermal response parameters and the corresponding covariance and confidence level are updated. When the gating condition is not met, the individual dynamic thermal response parameters and the corresponding covariance and confidence level remain unchanged.
[0178] The aforementioned stimulus detection module 1, gating identification module 2, predictive control module 3, and update module 4 can be centrally deployed in the cockpit domain controller, or they can be implemented collaboratively by the cockpit domain controller, the air conditioning controller, and the cloud account service. For example, stimulus detection module 1 and gating identification module 2 can be deployed in the cockpit domain controller, predictive control module 3 can be deployed in the air conditioning controller, and update module 4 can be deployed in the cloud account service. The modules deployed in the vehicle interact with each other through the controller LAN bus or in-vehicle network such as in-vehicle Ethernet, while the modules deployed in the cloud interact with the in-vehicle modules through the vehicle network communication link. The four modules can be executed separately by the same controller for different target seats, and data and parameters are distinguished according to occupant identification and target seat. The modules are divided according to function, and it is not required that each seat be equipped with an independent controller. For the coordination of shared actuators and local actuators, see Example 7.
[0179] Example 9: The complete operation process of the air conditioning control method provided in this embodiment of the invention will be described below in conjunction with three typical application scenarios. All descriptions of individual dynamic thermal response parameter adjustments in the following scenarios are based on the identification results of the complete response time sequence; that is, the direction of change of individual dynamic thermal response parameters is based on the identification results of the complete response time sequence, and not directly inferred from single observation features.
[0180] Scenario 1: Summer cooling scenario. Occupant A sits in the driver's seat. The system uses the driver's seat as the target seat and establishes a binding relationship between the occupant's identifier, the driver's seat, and the human body's thermal sensing area through facial recognition. During cooling operation, the system constructs a directional, normalized regional stimulus input acting on the occupant in the target seat based on actual execution data such as compressor power, fan speed, outlet air temperature, and airflow direction or zone valve position. The direction of the cooling effect is negative. When the candidate air conditioning stimulus is confirmed as a valid air conditioning stimulus and the sample quality reaches a preset quality threshold, the system identifies the individual's dynamic thermal response parameters separately for cooling and heating. In this scenario, occupant A's body surface temperature drop under the same normalized cooling stimulus is greater than the predicted value. The system uses the body surface temperature within the baseline window as the baseline and the body surface temperature offset relative to the baseline as the response quantity, based on the complete response time sequence. It uses a first-order model with pure delay to fit the data and identifies that the normalized response gain of the updated cooling direction input is increased and the time constant is decreased. Based on this, the predicted future body surface temperature drop magnitude and drop rate are both greater than the prediction results corresponding to the original individual dynamic thermal response parameters. Therefore, the system reduces the cooling power in advance before the body surface temperature reaches the target body surface temperature value corresponding to the target thermal comfort state, thus suppressing the overcooling phenomenon.
[0181] Scenario 2: Winter Heating Scenario. After passenger D takes their seat, the system designates passenger D's seat as the target seat, establishing a binding relationship between the passenger identifier, the target seat, and the human thermal sensing area. During heating operation, a directional, normalized regional stimulus input is constructed to act on the passenger in the target seat, with the direction of the heating effect set to positive. In this scenario, passenger D has a relatively large response delay and time constant; that is, the time interval between the heating stimulus and the onset of a detectable response in the body surface temperature is long, and the time required for the body surface temperature to approach a steady-state change is also long. Under these circumstances, the system's preset minimum hold time for applying air conditioning control ensures that the system will not continuously increase the heating power if no temperature rise is observed in a short period. This avoids overheating caused by the heating power being stacked to a high level before the body surface temperature response is established, and the body surface temperature rapidly exceeding the target value corresponding to the target thermal comfort state after the response is established. Once the response is confirmed, the system determines the prediction error by comparing the actual body surface response with the predicted result and updates the individual dynamic thermal response parameters when the sample quality reaches a preset update threshold.
[0182] Scenario 3: Preference Confirmation Scenario. An occupant in the target seat adjusts the air conditioning setting from 26℃ to 24℃. The system detects this change and records the environmental conditions, body surface temperature, and individual thermal comfort status before and after the change. Candidate preferences are generated, provided the change is not triggered by defogging, air quality, noise restrictions, safety protection, or temporary mandatory needs. The changed air conditioning setting remains unchanged for a preset verification duration, and the amplitude of the change in the regional stimulus input reaches a preset preference verification threshold. The system determines that the physiological thermal response verification is successful when the individual's thermal comfort state remains within the preset comfort zone and the absolute value of the rate of change of body surface temperature is lower than the preset stability threshold within the stable verification window. It also determines that the user behavior verification is successful when no change in air conditioning control quantity opposite to the change direction or a further correction of the change is detected within the preset observation window. When both verifications are successful and the number of successful verifications reaches the preset confirmation number under similar environments and occupant conditions, the system confirms the 24℃ setting in that scenario as a scenario-based preference and writes it into the occupant's long-term preference memory. Subsequently, when similar external temperature ranges, sunlight ranges, seasons, initial cabin temperatures, and seats are detected, this preference record is called according to scenario similarity as the initial target or control prior for candidate control sequences. When the real-time body surface thermal response is inconsistent with the preference record, the air conditioning control quantity is determined first based on the real-time body surface thermal response and safety constraints, and the preference record is suspended or its usage weight is reduced.
[0183] The above three scenarios respectively demonstrate the operation process of the air conditioning control method provided by the embodiments of the present invention under three typical working conditions: summer cooling, winter heating, and preference confirmation. Specifically, it updates the individual dynamic thermal response parameters according to the complete response time sequence to reduce the cooling power in advance, maintains the minimum hold time to avoid overheating based on the characteristics of large response delay and time constant, and writes the operation result confirmed by the stable comfort state into the long-term preference memory through dual verification.
[0184] Although embodiments of the present invention have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting the present invention. Those skilled in the art can make changes, modifications, substitutions and variations to the above embodiments within the scope of the present invention.
Claims
1. An air conditioning control method, characterized in that, Applied to vehicle cabins, including: Acquire the surface temperature of each body region of the occupant in the target seat and the actual execution data of the air conditioning actuator; construct a directional normalized regional stimulus input to act on the occupant based on the actual execution data; when the change amplitude of the regional stimulus input reaches a preset excitation threshold and lasts for a preset time, and the actual execution direction is consistent with the control command, determine the candidate air conditioning stimulus, and establish a baseline window before stimulation and a response window after stimulation. When the candidate air conditioning stimulus is confirmed as a valid air conditioning stimulus, the identity confidence, thermal zone registration confidence, body surface temperature measurement quality, stimulus quality, context quality, and perturbation quality are summarized according to a preset summary relationship to obtain the sample quality. When the sample quality reaches a preset quality threshold, based on the body surface temperature within the regional stimulus input and the baseline window and response window, the individual dynamic thermal response parameters are identified separately for cooling and heating, and the corresponding covariance and confidence are obtained. The individual dynamic thermal response parameters include the input normalized response gain, response delay, and time constant. The future human thermal state under each candidate control sequence is recursively predicted using the individual dynamic thermal response parameters, and the air conditioning control quantity is determined and executed. The prediction error is determined based on the actual body surface response after execution, and the sample quality is reassessed. When the sample quality reaches a preset update threshold, the individual dynamic thermal response parameters and corresponding covariance and confidence level are updated using the prediction error; otherwise, they remain unchanged.
2. The method according to claim 1, characterized in that, The candidate air conditioning stimuli are confirmed as valid air conditioning stimuli under the following conditions: The execution feedback of at least one of the following in the actual execution data is continuous and effective: actual power of compressor or heat pump, actual air outlet temperature, actual air volume, air direction or zone valve position, seat heating current and hot zone air supply status, and the signal-to-noise ratio of zone stimulus input reaches the preset threshold. No second significant change in air conditioning performance occurred within the baseline and response windows; The change in body surface temperature within the response window relative to the baseline temperature within the baseline window exceeds the preset measurement noise threshold, and the change in body surface temperature is consistent with the regional stimulus input in terms of time and direction of action. When all the conditions are met, the candidate air conditioning stimulus is confirmed as the effective air conditioning stimulus; The method further includes: terminating the current sample or re-segmenting when any condition used to confirm the effective air conditioning stimulus is no longer met within the baseline window or response window; When multiple thermal effects correspond to a response window, the multiple thermal effects are segmented or distinguished according to the execution time, the area of effect, or the input direction.
3. The method according to claim 1, characterized in that, The construction of a directional, normalized region stimulus input acting on the occupant based on actual execution data includes: The direction of heating is taken as positive, and the direction of cooling is taken as negative; When the normalized actual air conditioning thermal power can be obtained from the actual execution data, the regional stimulus input is calculated using the normalized actual air conditioning thermal power and the hit coefficient of the air outlet to the corresponding human body area; when the normalized actual air conditioning thermal power cannot be obtained from the actual execution data, the regional stimulus input is calculated using the difference between the normalized actual wind speed, the actual air outlet temperature and the body surface temperature of the human body area obtained from the actual execution data, and the hit coefficient. The change in the set value of the air conditioning control quantity or the change in the control command is only used to initiate candidate observation and is not considered as the effective air conditioning stimulus.
4. The method according to claim 1, characterized in that, The individual dynamic thermal response parameters are identified using a first-order model with pure delay. The first-order model with pure delay uses the body surface temperature within the baseline window as the baseline, the body surface temperature offset relative to the baseline as the response quantity, and introduces an environmental compensation term. The input normalized response gain characterizes the steady-state change in body surface temperature caused by a unit normalized cooling stimulus or a unit normalized heating stimulus. The response delay characterizes the duration between the moment the stimulus input to the region is applied and the first credible body surface temperature response; the time constant characterizes the time scale by which the body surface temperature response approaches the steady-state body surface temperature change. When the region stimulus input is zero, the change amplitude of the region stimulus input is lower than the excitation threshold, or the signal-to-noise ratio of the region stimulus input is insufficient, the response window is not used to identify the input normalized response gain. The method further includes: updating the individual dynamic thermal response parameters only when the response quantity meets the preset parameter confidence condition; The individual dynamic thermal response parameters for cooling and heating are stored separately, and parameter clusters are set according to seat, season, or environmental range.
5. The method according to claim 1, characterized in that, The determination of sample quality includes: The identity confidence level and the thermal zone registration confidence level are determined based on the binding relationship between the occupant's identifier and the target seat and the human body thermal sensing area; when the binding relationship is interrupted, the writing of the individual dynamic thermal response parameters is stopped; The quality of the body surface temperature measurement is determined based on at least one of the proportion of effective pixels and the number of effective regions in the body surface temperature of each human body region. When the candidate air conditioning stimulus is confirmed as a valid air conditioning stimulus, the stimulus quality is set to a non-zero value; when the candidate air conditioning stimulus is not confirmed as a valid air conditioning stimulus, the stimulus quality is set to zero, and the sample in this case is not included in the identification. When no window, door, or passenger seat change occurs within the baseline window and response window, the context quality is determined to be a valid value; otherwise, it is determined to be a invalid value. When the outside temperature, sunlight, vehicle speed, number of occupants, or other heat sources inside the cabin within the baseline window and response window do not experience sudden changes exceeding a preset threshold, or when the sudden changes can be eliminated by a pre-configured environmental compensation item, the disturbance quality is determined to be an effective value; otherwise, the disturbance quality is reduced or set to zero. The method further includes: storing samples whose quality reaches the update threshold after execution into the occupant's long-term account; using samples that are below the update threshold but not below the preset temporary use threshold only for the current driving session; and discarding samples that are below the temporary use threshold.
6. The method according to claim 1, characterized in that, Using the individual dynamic thermal response parameters to recursively predict the future human thermal state under each candidate control sequence includes: For each candidate control sequence, the individual dynamic thermal response parameters are used to recursively predict the surface temperature, surface temperature change rate, and regional temperature difference of each human body region. Based on the deviation of the body surface temperature from a predetermined individual thermal neutrality baseline, the rate of change of the body surface temperature, the regional temperature difference, and the environmental heat load, a predicted thermal comfort index is determined to characterize the individual's thermal comfort state. The determination and execution of the air conditioning control quantity includes: Based on the predicted thermal comfort index, changes in air conditioning control quantity and energy consumption corresponding to each candidate control sequence, determine the comfort deviation, control change and energy consumption corresponding to each candidate control sequence. With the goal of minimizing the weighted sum of the comfort deviation, the control variation, and the energy consumption, the air conditioning control quantity corresponding to each candidate control sequence is solved under safety constraints, and the candidate control sequence with the minimum weighted sum is determined as the target candidate control sequence. A preset dead zone, minimum hold time, and rate of change limit are applied to the air conditioning control quantity corresponding to the target candidate control sequence, and the limited air conditioning control quantity is executed.
7. The method according to claim 1, characterized in that, The method further includes: When passengers actively change the air conditioning control settings, the environmental conditions, body surface temperature, and individual thermal comfort status before and after the change are recorded to generate candidate preferences. When the identity confidence and thermal zone registration confidence are determined based on the occupant identification and the binding relationship between the target seat and the human body thermal sensing area, and the body surface temperature measurement quality is valid and the sample quality reaches the preset quality threshold, the candidate preference is verified; wherein, the change is not triggered by defogging, air quality, noise limitation, safety protection or temporary mandatory demand, the changed air conditioning control quantity is maintained for a preset verification duration, and the amplitude of the change in regional stimulus input reaches the preset preference verification threshold; The verification of the candidate preferences includes: when the individual's thermal comfort state remains within the preset comfort zone within the preset verification window and the absolute value of the rate of change of body surface temperature is lower than the preset stability threshold, the physiological thermal response verification is determined to be successful; when no change in air conditioning control quantity opposite to the change direction or further correction of the change is detected within the preset observation window, the user behavior verification is determined to be successful. When both the physiological thermal response verification and the user behavior verification pass, and the candidate preference reaches a preset high confidence level in a single verification or the number of times both verifications pass in similar environments and occupant states reaches a preset confirmation number, the candidate preference is written into the occupant's long-term preference memory; otherwise, it is not written into the long-term preference memory, and the confidence level of the candidate preference is reduced.
8. The method according to claim 7, characterized in that, The candidate preferences are managed using four states: candidate, confirmed, decayed, and invalid. The method further includes: When the candidate preference meets the conditions for being written into the long-term preference memory, the candidate preference is converted to a confirmed state; when the candidate preference has not been verified for a long time or the applicable scenario has changed, the candidate preference is converted to a decay state and the confidence of the candidate preference is reduced; when it is continuously detected that the occupant has made an air conditioning control change in the opposite direction of the change, the occupant identifier no longer matches, or the candidate preference has been inapplicable for a long time, the candidate preference is converted to an invalid state. The preference records in the long-term preference memory are associated and stored with the occupant identifier, seat, environmental state, most recent verification time, verification result and confidence level, and the preference records are called according to the scene similarity as the initial target or control prior of the candidate control sequence. When the real-time body surface thermal response is inconsistent with the preference record, the air conditioning control quantity is determined first based on the real-time body surface thermal response and safety constraints, and the use of the preference record is suspended or the usage weight of the preference record is reduced.
9. The method according to claim 1, characterized in that, The method is performed separately for occupants in multiple seats within the vehicle cabin, and the method further includes: The individual dynamic thermal response parameters of each occupant are identified, and the individual thermal comfort state of each occupant is determined. When the individual thermal comfort state of multiple occupants deviates from the preset comfort zone and the direction of deviation is the same, the occupant with the most severe deviation will share the cooling or heating capacity. When the individual thermal comfort states of multiple occupants deviate from the opposite direction, the shared cabin temperature will be maintained within a preset compromise range that balances safety and energy consumption, and will be adjusted first through the corresponding local air volume, air direction, zone valves, seat heating or seat ventilation actuators. The method further includes: The operating status is determined based on factors such as passenger changes or seat changes, human body area obstruction, body surface temperature measurement drift, insufficient stimulation input to the area, sudden environmental changes, continuous exceeding of prediction error limits, or communication failure. The operating status includes normal status, suspicious status, and failure status. In the normal status, recursive prediction and updates of the individual dynamic thermal response parameters are performed. In the suspicious status, recursive prediction is maintained and updates of the individual dynamic thermal response parameters and their corresponding covariance and confidence are frozen. In the failure status, predictive control based on the individual dynamic thermal response parameters is exited and the system switches to the pre-configured basic air conditioning strategy. Among these, anti-fogging and defrosting, power battery thermal management, and driving safety requirements take precedence over the individual comfort needs of multiple occupants.
10. An air conditioning control system, characterized in that, include: The stimulus detection module is used to acquire the surface temperature of each human body area of the occupant in the target seat and the actual execution data of the air conditioning actuator. Based on the actual execution data, a directional normalized regional stimulus input is constructed to act on the occupant. When the change amplitude of the regional stimulus input reaches the preset excitation threshold and lasts for a preset time, and the actual execution direction is consistent with the control command, a candidate air conditioning stimulus is determined and a baseline window before stimulation and a response window after stimulation are established. The gating identification module is used to summarize the identity confidence, thermal zone registration confidence, body surface temperature measurement quality, excitation quality, context quality and perturbation quality according to a preset summary relationship when the candidate air conditioning stimulus is confirmed as a valid air conditioning stimulus, to obtain the sample quality. When the sample quality reaches a preset quality threshold, based on the surface temperature within the region stimulus input and the baseline window and response window, the individual dynamic thermal response parameters are identified according to cooling and heating respectively, and the corresponding covariance and confidence are obtained. The individual dynamic thermal response parameters include input normalized response gain, response delay and time constant. The predictive control module is used to recursively predict the future human thermal state under the action of each candidate control sequence using the individual dynamic thermal response parameters, and to determine and execute the air conditioning control quantity. An update module is used to determine the prediction error based on the actual body surface response after execution, and to redetermine the sample quality; When the sample quality reaches a preset update threshold, the individual dynamic thermal response parameters and their corresponding covariance and confidence level are updated using the prediction error; otherwise, they remain unchanged.
