Multi-sensor data fusion temperature controller precise temperature control method

Abstract

The present application relates to the field of automatic temperature control, and in particular to a precise temperature control method for a temperature controller based on multi-sensor data fusion. A dynamic heat propagation partial differential equation is constructed based on space thermodynamic parameters of the temperature controller and coordinates of the multi-sensor, and heat conduction time delay characteristics between each sensor and the temperature controller actuator are calculated. An asynchronous temperature data stream of the multi-sensor is obtained, and each sampling value is deduced to a unified time reference along the time axis based on the heat conduction time delay characteristics, and the predicted temperature of each node at the current time is calculated. The predicted temperature is weighted and fused to obtain a space heat field fusion temperature, which is input into a closed-loop temperature control regulator to generate a temperature controller actuator control instruction. The present application aligns the heat conduction time delay of the multi-sensor caused by the difference in spatial position, eliminates the phase deviation of asynchronous data fusion, compensates for the measurement lag of thermal inertia, reduces the overshoot of the temperature control system, and speeds up the response speed to thermal disturbance.

Description

Technical Field

[0001] This invention relates to the field of automatic temperature control, and more specifically to a precise temperature control method for a temperature controller based on multi-sensor data fusion. Background Technology

[0002] Existing temperature control systems typically deploy multiple temperature sensors within the control space to collect ambient temperature data, with each sensor outputting a temperature data stream according to its own sampling period. Conventional multi-sensor data fusion schemes employ either synchronously reading data from each sensor at a fixed period or aligning data from different sampling times on the time axis using linear interpolation. Subsequently, they use averaging or Kalman filtering methods to weight and fuse the sensor readings, and then input the fused temperature value as the ambient temperature into the closed-loop control logic to generate control commands for the heating or cooling actuators.

[0003] Heat conduction within a temperature field exhibits physical inertia, and sensors at different spatial locations detect changes in the same heat source at different times. Existing conventional solutions directly fuse asynchronous temperature data with misaligned heat conduction delays, causing the fusion result to deviate from the true physical state of the spatial thermal field at the current moment, resulting in lag and temperature overshoot in the temperature control actuator. Summary of the Invention

[0004] The purpose of this invention is to provide a precise temperature control method for a thermostat based on multi-sensor data fusion, which can solve the problems mentioned in the background art.

[0005] To achieve the above objectives, the technical solution adopted by the present invention is as follows: A method for precise temperature control of a thermostat based on multi-sensor data fusion is characterized by the following steps: acquiring the thermodynamic parameters of the space where the thermostat is located and the spatial coordinates of multiple sensors; constructing a dynamic heat propagation partial differential equation based on the thermodynamic parameters and the spatial coordinates; calculating the heat conduction delay characteristics between each sensor and the temperature controller actuator according to the dynamic heat propagation partial differential equation; acquiring asynchronous temperature data streams output by the multiple sensors according to different sampling periods; based on the heat conduction delay characteristics, reversing each sample value in the asynchronous temperature data stream along the time axis and mapping it to a unified time reference to calculate the predicted temperature of each sensor spatial node at the current moment; weighting and fusing the predicted temperatures to obtain the spatial thermal field fused temperature at the current moment; inputting the spatial thermal field fused temperature into a closed-loop temperature controller to generate control commands for the temperature controller actuator.

[0006] Preferably, the step of constructing a dynamic heat propagation partial differential equation based on the thermodynamic parameters and the spatial coordinates includes: obtaining the distribution location of each medium in the space and the corresponding thermal diffusivity; dividing the space into a multi-scale grid, and calibrating the boundary conditions of the multi-scale grid, wherein the boundary conditions include adiabatic boundaries and convective heat transfer boundaries; establishing a dynamic heat propagation partial differential equation for the heterogeneous medium based on the thermal diffusivity and the boundary conditions; determining the spatial discrete distance between each node in the dynamic heat propagation partial differential equation and the temperature control actuator according to the spatial coordinates of the multi-sensor, substituting the spatial discrete distance into the dynamic heat propagation partial differential equation to solve for the heat conduction time delay feature.

[0007] Preferably, the step of calculating the heat conduction delay characteristics between each sensor in the multi-sensor system and the temperature control actuator based on the dynamic heat propagation partial differential equation includes: acquiring airflow velocity vector distribution data in the space; introducing the airflow velocity vector distribution data as a convection term into the dynamic heat propagation partial differential equation to construct a convection-diffusion coupling partial differential equation; solving the eigenvalues ​​of the convection-diffusion coupling partial differential equation to extract the heat propagation master mode; and calculating the dynamic heat conduction delay characteristics between each sensor in the multi-sensor system and the temperature control actuator based on the attenuation coefficient and oscillation frequency of the heat propagation master mode, wherein the dynamic heat conduction delay characteristics are dynamically updated as the airflow velocity vector distribution data changes.

[0008] Preferably, the step of acquiring the asynchronous temperature data stream output by the multiple sensors according to different sampling periods includes: receiving the original temperature sampling sequence output by the multiple sensors; for any sensor channel in the original temperature sampling sequence, calculating the time interval between adjacent sampling points; when the time interval exceeds a preset multiple of the standard sampling period corresponding to the sensor channel, determining that there is a missing data segment; within the missing data segment, extracting the adjacent historical valid sampling values ​​of the sensor channel, calculating the heat flux change rate in combination with the heat conduction delay characteristics, and performing interpolation reconstruction based on the heat flux change rate and the historical valid sampling values ​​to generate a completed asynchronous temperature data stream.

[0009] Preferably, the step of calculating the predicted temperature of each sensor spatial node at the current moment by back-deriving each sampled value in the asynchronous temperature data stream along the time axis based on the heat conduction delay characteristic and mapping it to a unified time reference includes: obtaining the thermal capacity response hysteresis coefficient of each of the multiple sensors; for each sampled value in the asynchronous temperature data stream, adding the heat conduction delay characteristic and the thermal capacity response hysteresis coefficient to obtain the comprehensive hysteresis time; using the comprehensive hysteresis time as the back-deriving step size, pushing the sampled value forward along the time axis, and combining the integral of the dynamic heat propagation partial differential equation within the comprehensive hysteresis time to calculate the predicted temperature of each sensor spatial node at the current moment after eliminating the spatial heat conduction delay and sensor thermal capacity hysteresis.

[0010] Preferably, the step of inputting the space thermal field fusion temperature into a closed-loop temperature controller to generate control commands for the temperature controller actuator includes: calculating the temperature deviation between the space thermal field fusion temperature and a preset target temperature; obtaining the system thermal inertia parameters corresponding to the dynamic heat propagation partial differential equation, wherein the system thermal inertia parameters are determined based on the total heat capacity of the medium in space and the airflow convection intensity; dynamically adjusting the proportional gain and integral time constant of the closed-loop temperature controller according to the system thermal inertia parameters; and calculating the temperature deviation based on the adjusted proportional gain and integral time constant to generate control commands for the temperature controller actuator.

[0011] Preferably, the process of dividing the space into multi-scale grids and calibrating the boundary conditions of the multi-scale grids, including both adiabatic and convective heat transfer boundaries, includes: acquiring real-time detection signals of the opening and closing states of doors and windows within the space; locating the affected boundary grid regions in the space when the opening and closing states of doors and windows change; dynamically switching the boundary conditions of the affected boundary grid regions from adiabatic to convective heat transfer boundaries, or vice versa; recalculating the thermal resistance network between the switched boundary grid regions and adjacent grid regions; and updating the boundary constraint matrix of the dynamic heat propagation partial differential equation based on the thermal resistance network.

[0012] Preferably, the step of calculating the dynamic heat conduction delay characteristics between each sensor in the multi-sensor system and the temperature control actuator based on the attenuation coefficient and oscillation frequency of the primary heat propagation mode includes: performing time-series sliding window monitoring on the airflow velocity vector distribution data and calculating the variance change of the airflow velocity within the sliding window; triggering a modal mutation response mechanism when the variance change exceeds a preset steady-state threshold; extracting secondary modes in the primary heat propagation mode whose attenuation coefficient is lower than a preset attenuation lower limit under the modal mutation response mechanism; weighting and superimposing the delay parameters corresponding to the secondary modes with the delay parameters corresponding to the primary heat propagation mode, and using the superposition result as the dynamic heat conduction delay characteristics under the mutation condition.

[0013] Preferably, the steps of extracting historical valid sampled values ​​adjacent to the sensor channel within the data missing segment, calculating the heat flux change rate based on the heat conduction delay characteristics, and performing interpolation reconstruction based on the heat flux change rate and the historical valid sampled values ​​to generate a completed asynchronous temperature data stream include: determining whether the time span of the data missing segment is greater than the single-channel reconstruction limit threshold; if so, obtaining reference sampled values ​​of at least one reference sensor channel spatially adjacent to the sensor channel within the data missing segment; constructing spatial interpolation weight coefficients based on the spatial distance between the sensor channel and the reference sensor channel and the correlation with historical temperature; spatially mapping the reference sampled values ​​according to the spatial interpolation weight coefficients, and performing spatiotemporal joint interpolation reconstruction based on the historical valid sampled values ​​to generate a completed asynchronous temperature data stream.

[0014] Preferably, the step of obtaining the system thermal inertia parameters corresponding to the dynamic heat propagation partial differential equation, wherein the system thermal inertia parameters are determined based on the total heat capacity of the medium in space and the airflow convection intensity, includes: real-time acquisition of the rate of change between the output power of the temperature control actuator and the fusion temperature of the space thermal field; online identification of the system thermal inertia parameters at the current moment based on the ratio of the output power to the rate of change; calculation of the inertia deviation between the system thermal inertia parameters at the current moment and the initial nominal thermal inertia parameters; if the inertia deviation exceeds the allowable inertia range, calculation of the feedforward compensation amount based on the inertia deviation; superimposing the feedforward compensation amount onto the temperature deviation value, and then inputting it into the closed-loop temperature controller after parameter adjustment to generate a control command for the temperature control actuator to suppress overshoot.

[0015] Compared with the prior art, the beneficial effects of the present invention are as follows: 1. This invention calculates the heat conduction delay characteristics between each sensor and the temperature control actuator by constructing a dynamic heat propagation partial differential equation. It then maps the asynchronous temperature data stream backward along the time axis to a unified time reference, calculates the predicted temperature of each node at the current moment, and performs data fusion. This method aligns the heat conduction delay caused by spatial differences among multiple sensors before data fusion, eliminates phase deviations in asynchronous data fusion, compensates for measurement lag caused by thermal inertia of the temperature field, and ensures that the fused temperature value restores the true thermal field state at the current moment. This reduces the overshoot of the temperature control system and accelerates the response speed to thermal disturbances.

[0016] 2. This invention introduces airflow velocity vectors to construct convection-diffusion coupled partial differential equations to extract the main mode of heat propagation. It then combines this with the sensor's thermal capacity response hysteresis coefficient to calculate the overall lag time, ensuring that the time delay compensation aligns with airflow disturbances and the sensor's own physical lag characteristics. For missing data segments, it uses interpolation based on the rate of change of heat flux for reconstruction or utilizes adjacent spatial reference channels for spatiotemporal joint reconstruction, maintaining the continuity of the asynchronous data stream. Based on the system's thermal inertia parameters, it dynamically adjusts the proportional gain and integral time constant of the closed-loop controller, and introduces feedforward compensation when the inertia deviation exceeds limits to suppress the impact of spatial medium thermal capacity changes on control stability. When spatial boundary conditions change or airflow variance abruptly changes, it dynamically switches the boundary conditions and superimposed secondary mode time delay parameters, enabling the time delay characteristics to adapt to sudden changes in the physical environment and suppressing temperature control instability caused by operating condition disturbances. Attached Figure Description

[0017] Figure 1 The flowchart of the main method for precise temperature control of a temperature controller based on multi-sensor data fusion of the present invention is shown below. Figure 2 This is a flowchart illustrating the construction of the dynamic heat propagation partial differential equation and the generation of heat conduction time delay features in this invention. Figure 3 This is a flowchart of the process for solving the convection-diffusion coupling equation and updating the dynamic heat conduction delay characteristics of the present invention. Figure 4 This is a flowchart of the asynchronous temperature data stream acquisition and spatiotemporal joint reconstruction of missing data according to the present invention. Figure 5 This is a flowchart of the asynchronous temperature data time axis reverse extrapolation and predicted temperature calculation of the present invention; Figure 6 This is a flowchart illustrating the adaptive adjustment of parameters and generation of control commands for the closed-loop temperature controller of the present invention. Detailed Implementation

[0018] refer to Figure 1 In one embodiment, the thermodynamic parameters of the space where the temperature controller is located and the spatial coordinates of multiple sensors are obtained. A dynamic heat propagation partial differential equation is then constructed based on the thermodynamic parameters and spatial coordinates. The thermodynamic parameters include the thermal diffusivity, specific heat capacity, density, and thermal conductivity of each medium within the space. The spatial coordinates of the multiple sensors are represented using a three-dimensional Cartesian coordinate system. A spatial rectangular coordinate system O-xyz is established with the geometric center of the temperature controller actuator as the origin, where the x-axis is along the length of the space, the y-axis is along the width of the space, and the z-axis is along the height of the space.

[0019] refer to Figure 2 The distribution location and corresponding thermal diffusivity of each medium within the space are obtained. The medium includes air, walls, floor, ceiling, and indoor furniture. Different media have different thermal diffusivity, with air having a thermal diffusivity of [missing value]. The thermal diffusivity of the wall The thermal diffusivity of the floor The thermal diffusivity of the ceiling The thermal diffusivity of furniture The space is divided into a multi-scale grid, and boundary conditions are defined for each grid, including adiabatic and convective heat transfer boundaries. An adaptive method is used for multi-scale grid division, employing finer grids in regions with larger temperature gradients and coarser grids in regions with smaller temperature gradients. The size of the grid cells ranges from 0.1m to 1.0m and is dynamically adjusted according to the temperature field distribution characteristics within the space.

[0020] Based on the aforementioned thermal diffusivity and boundary conditions, a partial differential equation for dynamic heat propagation in a heterogeneous medium is established. The partial differential equation for dynamic heat propagation in a heterogeneous medium is expressed as: ; in, For spatial location The density of the medium, expressed in kg / m³; For spatial location The specific heat capacity of the medium, expressed in J / (kg·K); The temperature at spatial location (x, y, z) at time t, in Kelvin; The thermal conductivity of the medium at spatial location (x, y, z) is expressed in W / (m·K). The intensity of the internal heat source at spatial location (x, y, z) at time t is expressed in W / m³. For gradient operators, It is a divergence operator.

[0021] Based on the spatial coordinates of the multiple sensors, the discrete spatial distances between each node in the dynamic heat propagation partial differential equation and the temperature control actuator are determined. These discrete spatial distances are then substituted into the dynamic heat propagation partial differential equation to solve for the heat conduction delay characteristic. Let the spatial coordinates of the i-th sensor be... The spatial coordinates of the temperature control actuator are The spatial discrete distance between the i-th sensor and the temperature control actuator is: based on the spatial coordinates of the multiple sensors, determine the spatial discrete distance between each node in the dynamic heat propagation partial differential equation and the temperature control actuator, substitute the spatial discrete distance into the dynamic heat propagation partial differential equation to solve, and generate the heat conduction delay feature. Let the spatial coordinates of the i-th sensor be... The spatial coordinates of the temperature control actuator are The spatial discrete distance between the i-th sensor and the temperature control actuator is: ; Thermal conduction delay characteristics This represents the time required for the temperature change generated by the temperature controller to propagate to the i-th sensor, obtained by solving the dynamic heat propagation partial differential equation. For the steady-state heat conduction process, the heat conduction time delay characteristic can be approximated as: ; in, The equivalent thermal diffusivity of the medium within the space is obtained by weighted averaging of the thermal diffusivity and volume fraction of each medium.

[0022] The heat conduction delay characteristics between each sensor and the temperature control actuator in a multi-sensor system are calculated based on the dynamic heat propagation partial differential equation. For unsteady-state heat conduction processes, the heat conduction delay characteristics change with time and need to be obtained by numerically solving the dynamic heat propagation partial differential equation. The dynamic heat propagation partial differential equation is discretized using the finite volume method, discretizing the spatial domain into N control volumes and the time domain into M time steps. The discretized equation is expressed as: ; in, The coefficient for the current control volume P. The coefficient for adjacent control volumes nb. Let P be the temperature of the current control volume at the nth time step. Let b be the temperature of the adjacent control volume nb at the nth time step, and b be the source term.

[0023] The asynchronous temperature data streams output by the multiple sensors at different sampling periods are acquired. The multiple sensors include N temperature sensors, each with a different sampling period. The system receives the raw temperature sampling sequence output by the multiple sensors, and the raw temperature sampling sequence is represented as follows: Where i is the sensor number and j is the sampling point number. Let be the sampling time of the j-th sampling point of the i-th sensor.

[0024] For any sensor channel in the original temperature sampling sequence, the time interval between adjacent sampling points is calculated. When the time interval exceeds a preset multiple of the standard sampling period corresponding to the sensor channel, a data missing segment is determined to exist. Let the standard sampling period of the i-th sensor channel be... The preset multiplier is Then when the time interval between adjacent sampling points At that time, it is determined that at time... to There are missing data segments.

[0025] Within the missing data segment, historical valid sampled values ​​adjacent to the sensor channel are extracted. The heat flux change rate is calculated based on the heat conduction delay characteristics. Interpolation and reconstruction are then performed based on the heat flux change rate and the historical valid sampled values ​​to generate a completed asynchronous temperature data stream. (Heat flux change rate) This represents the rate of change of heat flux density at the location of the i-th sensor per unit time, derived from the characteristics of heat conduction delay. Calculated from adjacent historical valid sample values: ; in, Let be the density of the medium at the location of the i-th sensor. Let be the specific heat capacity of the medium at the location of the i-th sensor.

[0026] refer to Figure 5 Based on the heat conduction delay characteristics, each sampled value in the asynchronous temperature data stream is deduced backward along the time axis and mapped to a unified time base to calculate the predicted temperature of each sensor spatial node at the current moment. The thermal capacity response hysteresis coefficients of each of the multiple sensors are then obtained. This indicates the temperature measurement lag time caused by the sensor's own heat capacity, which is determined by the sensor's heat capacity and thermal conductivity. ;

[0027] in, Let the mass of the i-th sensor be... Let i be the specific heat capacity of the i-th sensor. Let be the convective heat transfer coefficient between the i-th sensor and the surrounding medium. Let be the surface area of ​​the i-th sensor.

[0028] For each sampled value in the asynchronous temperature data stream, the heat conduction delay characteristic is added to the heat capacity response hysteresis coefficient to obtain the comprehensive hysteresis time: ; Using the comprehensive lag time as the backward extrapolation step size, the sampled values ​​are pushed forward along the time axis, and combined with the integral of the dynamic heat propagation partial differential equation within the comprehensive lag time, the predicted temperature of each sensor spatial node at the current moment is calculated after eliminating the spatial heat conduction delay and sensor thermal capacity lag. The current moment is... The i-th sensor at time i The sample value is Then the predicted temperature of the i-th sensor spatial node at the current time is: ; in, It is obtained from the solution of the partial differential equation of dynamic heat propagation at the i-th sensor spatial node.

[0029] The predicted temperatures are weighted and fused to obtain the current spatial thermal field fused temperature. The weighting fusion employs a weighting method based on sensor reliability, with weight coefficients determined by the sensor's measurement accuracy, spatial location importance, and historical data consistency. Let the weight coefficient of the i-th sensor be , then the spatial thermal field fused temperature is:

[0030] Among them, the weighting coefficient Satisfy the normalization condition: ; The space thermal field fusion temperature is input into the closed-loop temperature controller to generate control commands for the temperature control actuator. The temperature deviation between the space thermal field fusion temperature and the preset target temperature is calculated. ;

[0031] in, The preset target temperature.

[0032] The system thermal inertia parameter corresponding to the dynamic heat propagation partial differential equation is obtained. This parameter is determined based on the total heat capacity of the medium within the space and the airflow convection intensity. The system thermal inertia parameter J represents the heat required to raise the system temperature by 1 K, and is jointly determined by the total heat capacity of all media within the space and the airflow convection intensity. ;

[0033] in, This represents the number of types of media within the space. Let the density be the density of the k-th medium. Let k be the volume of the k-th medium. Let k be the specific heat capacity of the k-th medium. The convective heat transfer coefficient is... For convective heat transfer area, This represents the airflow velocity.

[0034] The proportional gain and integral time constant of the closed-loop temperature controller are dynamically adjusted based on the system's thermal inertia parameters. The closed-loop temperature controller employs a proportional-integral (PI) controller, and its transfer function is:

[0035] in, For proportional gain, The integral time constant is used. The proportional gain and integral time constant are dynamically adjusted based on the system's thermal inertia parameters. ; ; in, This is the initial proportional gain. Let be the initial integration time constant. This is the initial nominal thermal inertia parameter.

[0036] The temperature deviation value is calculated based on the adjusted proportional gain and integral time constant to generate control commands for the temperature control actuator. The output of the PI controller is: ; in, This is the control command for the temperature control actuator, ranging from 0 to 100%, representing the percentage of the actuator's output power relative to its rated power.

[0037] In this embodiment, the calibration results of multi-scale grid parameters and boundary conditions are shown in Table 1.

[0038] Table 1. Calibration of Multi-Scale Mesh Parameters and Boundary Conditions

[0039] Table 1 shows the meshing parameters and boundary condition calibration results for the spatial multi-scale grid. Due to the large temperature gradient in the central region, a fine grid of 0.1 m is used; the temperature gradients in the wall, floor, and ceiling regions are smaller, so a coarse grid of 0.5 m is used; and the window region, due to convective heat transfer, uses a medium grid of 0.2 m. Different medium regions correspond to different thermal diffusivity, and the boundary conditions are calibrated as either adiabatic or convective heat transfer boundaries based on the region characteristics.

[0040] In this embodiment, the propagation characteristics of the temperature field in space are accurately described by constructing a dynamic partial differential equation for heat propagation in a heterogeneous medium. Based on the spatial coordinates of multiple sensors, the heat conduction delay characteristics between each sensor and the temperature control actuator are calculated. The asynchronous temperature data stream is then extrapolated backward along the time axis to a unified time reference, eliminating the influence of spatial heat conduction delay and sensor thermal hysteresis. A weighted fusion method based on sensor reliability is used to obtain the fused temperature of the spatial thermal field. Combined with the system's thermal inertia parameters, the parameters of the PI controller are dynamically adjusted to generate control commands for the temperature control actuator. This method can accurately reproduce the true state of the spatial thermal field at the current moment, improving the control accuracy of the temperature control system.

[0041] refer to Figure 3 In a preferred embodiment, airflow velocity vector distribution data within the space is collected. This airflow velocity vector distribution data is obtained through wind speed sensors positioned at different locations within the space, with each wind speed sensor outputting a three-dimensional airflow velocity vector. ,in Let x be the airflow velocity component in the x-direction. Let be the airflow velocity component in the y-direction. This represents the airflow velocity component in the z-direction. The sampling period for the airflow velocity vector distribution data is synchronized with the sampling period of the temperature sensor.

[0042] The airflow velocity vector distribution data is introduced as a convection term into the dynamic heat propagation partial differential equation to construct a convection-diffusion coupled partial differential equation. The convection-diffusion coupled partial differential equation is expressed as: ; Where v is the airflow velocity vector, with units of m / s; For convection, it represents heat transfer caused by airflow.

[0043] The eigenvalues ​​of the convection-diffusion coupled partial differential equations are solved to extract the principal modes of heat propagation. The orthogonal decomposition (POD) method is then used to perform mode decomposition on the solutions of the convection-diffusion coupled partial differential equations. First, temperature field data at different times are collected to construct a temperature field snapshot matrix X: ; in, The number of snapshots. Calculate the covariance matrix C of the temperature field snapshot matrix: ; Perform eigenvalue decomposition on the covariance matrix C: ; in, For eigenvalues, These are the corresponding eigenvectors, i.e., the POD basis functions. The eigenvalues ​​are arranged in descending order, and the eigenvectors corresponding to the K largest eigenvalues ​​constitute the principal modes of heat propagation. The value of K is determined according to the energy retention criterion, typically retaining more than 99% of the energy. ; Based on the attenuation coefficient and oscillation frequency of the dominant heat propagation mode, the dynamic heat conduction delay characteristics between each sensor in the multi-sensor system and the temperature control actuator are calculated. These dynamic heat conduction delay characteristics are dynamically updated as the airflow velocity vector distribution data changes. (Attenuation coefficient of the dominant heat propagation mode) and oscillation frequency From eigenvalues Determining the real and imaginary parts: ; Where j is the imaginary unit. The time delay parameter corresponding to the i-th dominant heat propagation mode is: ; The dynamic heat conduction delay characteristic between the k-th sensor and the temperature control actuator in a multi-sensor system is a weighted sum of the delay parameters of each principal mode: ; in, The weighting coefficient of the i-th principal mode at the k-th sensor position is determined by the value of the POD basis function at the k-th sensor position.

[0044] The airflow velocity vector distribution data is monitored using a time-series sliding window, and the variance change of the airflow velocity within the sliding window is calculated. A sliding window of length L is used to monitor the magnitude of the airflow velocity vector, and the variance of the airflow velocity magnitude within the window is calculated. ; in, Let be the magnitude of the airflow velocity vector at time i. The average value of the airflow velocity modulus within the window: ; Calculate the change in variance between two adjacent windows: ; When the variance change exceeds a preset steady-state threshold, a modal mutation response mechanism is triggered. (Preset steady-state threshold) The value is determined based on the normal fluctuation range of airflow within the space, typically taken as twice the variance under normal operating conditions. When > When a sudden change in airflow state is detected, the modal change response mechanism is triggered.

[0045] Under the modal abrupt response mechanism, secondary modes with attenuation coefficients lower than a preset attenuation lower limit are extracted from the primary heat propagation mode. Preset attenuation lower limit The value is determined based on the system's dynamic response characteristics, and is typically taken as 1 / 10 of the average attenuation coefficient of the principal mode. Extract all parameters that satisfy this value. The secondary modes constitute the secondary mode set S.

[0046] The time delay parameters corresponding to the secondary mode and the primary mode of heat propagation are weighted and superimposed, and the superposition result is used as the dynamic heat conduction time delay feature under abrupt change conditions. The dynamic heat conduction time delay feature between the k-th sensor and the temperature control actuator under abrupt change conditions is as follows:

[0047] in, This is a weighting coefficient, with a value ranging from 0.5 to 0.8, dynamically adjusted according to the severity of sudden airflow changes.

[0048] In this embodiment, the calculation results of the main heat propagation mode parameters and time delay are shown in Table 2.

[0049] Table 2 Calculation of Dominant Mode Parameters and Time Delay for Heat Propagation

[0050] Table 2 shows the parameters and time delay calculation results of the main heat propagation modes. The total energy proportion of the first six main modes reaches 99.99%, which meets the energy retention criterion. Mode 1 is a steady-state heat conduction mode with an oscillation frequency of 0 and the highest energy proportion, reaching 85.32%. Modes 2 to 6 are oscillating modes with gradually increasing attenuation coefficients and gradually decreasing time delay parameters.

[0051] In this embodiment, a convection-diffusion coupled partial differential equation is constructed by introducing an airflow velocity vector, considering the influence of airflow on temperature field propagation. An orthogonal decomposition method is used to extract the primary mode of heat propagation, and dynamic heat conduction time delay characteristics are calculated based on the attenuation coefficient and oscillation frequency of the primary mode, allowing the time delay characteristics to be dynamically updated with changes in airflow velocity. When a sudden change occurs in the airflow state, a modal change response mechanism is triggered, weighted and superimposed with the time delay parameters of the primary mode, improving the accuracy of time delay calculation under abrupt changes. This method can effectively compensate for changes in temperature field propagation time delay caused by airflow disturbances, improving the control stability of the temperature control system under varying airflow conditions.

[0052] In another preferred embodiment, real-time detection signals of the opening and closing status of doors and windows within the space are acquired. These signals are collected by position sensors mounted on the doors and windows. The position sensors output switching signals, where 0 indicates the doors and windows are closed and 1 indicates they are open. The detection period for the opening and closing status of doors and windows is synchronized with the sampling period of the temperature sensor.

[0053] When the opening and closing state of the doors and windows changes, the affected boundary grid region in the space is located. Changes in the opening and closing state of the doors and windows affect the boundary conditions of the space; the affected boundary grid region is the location of the doors and windows and their adjacent grid regions. The extent of the affected boundary grid region is determined based on the spatial coordinates and dimensions of the doors and windows.

[0054] The boundary conditions of the affected boundary grid region are dynamically switched from an adiabatic boundary to a convective heat transfer boundary, or vice versa. When a door or window changes from closed to open, the boundary conditions of the affected boundary grid region are switched from an adiabatic boundary to a convective heat transfer boundary; when a door or window changes from open to closed, the boundary conditions of the affected boundary grid region are switched from a convective heat transfer boundary to an adiabatic boundary.

[0055] The thermal resistance network between the switched boundary grid region and adjacent grid regions is recalculated, and the boundary constraint matrix of the dynamic heat propagation partial differential equation is updated based on this thermal resistance network. The thermal resistance network is composed of the thermal resistance between adjacent grid regions, and the formula for calculating the thermal resistance is: ; in, Let i be the thermal resistance between grid region i and grid region j. The distance between grid region i and grid region j. Let i be the equivalent thermal conductivity coefficient between grid region i and grid region j. Let be the contact area between grid region i and grid region j.

[0056] The boundary constraint matrix B represents the constraint effect of the boundary conditions on the temperature field, and its elements... This represents the heat flux contribution of boundary grid region i to internal grid region j. When the boundary conditions change, the elements of the boundary constraint matrix are recalculated, and the boundary conditions of the dynamic heat propagation partial differential equation are updated.

[0057] refer to Figure 4 Determine whether the time span of the missing data segment exceeds the single-channel reconstruction limit threshold. Single-channel reconstruction limit threshold. The sampling period of the sensor is determined based on the stability of the heat flux change rate, typically five times the standard sampling period. This applies when the time span of the data missing segment... When it is determined that single-channel interpolation reconstruction cannot guarantee the accuracy of the data, it is necessary to use spatially adjacent reference channels for spatiotemporal joint interpolation reconstruction.

[0058] If so, then obtain the reference sample value of at least one reference sensor channel spatially adjacent to the sensor channel within the data missing segment. Spatially adjacent reference sensor channels refer to sensor channels whose spatial distance from the sensor channel with missing data is less than a preset distance threshold. The preset distance threshold is determined based on the spatial correlation of the temperature field within the space, and is typically set to 2m.

[0059] Based on the spatial distance between the sensor channel and the reference sensor channel and its correlation with historical temperature, spatial interpolation weighting coefficients are constructed. The spatial interpolation weighting coefficients for the k-th reference sensor channel are then defined. for: ; in, This represents the historical temperature correlation coefficient between the sensor channel with missing data and the k-th reference sensor channel. The spatial distance between the sensor channel with missing data and the k-th reference sensor channel. The number of reference sensor channels.

[0060] Historical temperature correlation coefficient The Pearson correlation coefficients of the temperature sequences from the two sensor channels over a past period were calculated to obtain the following: ; in, The temperature value of the sensor channel with missing data at time t. Let be the temperature value of the k-th reference sensor channel at time t. The average temperature of the sensor channel with missing data. Let N be the average temperature of the k-th reference sensor channel, and N be the number of historical data points.

[0061] Spatially mapping the reference sample values ​​based on the spatial interpolation weighting coefficients, and combining this with historical valid sample values ​​for spatiotemporal joint interpolation reconstruction, generates the completed asynchronous temperature data stream. The formula for spatiotemporal joint interpolation reconstruction is: ; in, The reconstructed temperature value at time t for the sensor channel with missing data. This is a single-channel interpolation result based on historical valid sample values. This is the time interpolation weighting coefficient, with a value ranging from 0.2 to 0.8, which is dynamically adjusted according to the time span of the missing data segment.

[0062] refer to Figure 6 The system collects real-time data on the output power of the temperature control actuator and the rate of change of the fusion temperature of the space thermal field. The output power u(t) of the temperature control actuator is determined by the output command of the controller, and the rate of change of the fusion temperature of the space thermal field is... By measuring the fusion temperature of the space thermal field Numerical differentiation yields: ; in, The sampling period.

[0063] Based on the ratio of the output power to the rate of change, the system thermal inertia parameters at the current moment are identified online. The formula for online identification of the system thermal inertia parameters is: ; in, This represents the rated power of the temperature control actuator. To improve the accuracy of identification, the recursive least squares method is used to estimate the system's thermal inertia parameters online. ; ; ; in, Here is the gain matrix. Let be the covariance matrix, and λ be the forgetting factor, with values ​​ranging from 0.95 to 0.99.

[0064] Calculate the inertia deviation between the current system thermal inertia parameter and the initial nominal thermal inertia parameter: ; If the inertia deviation exceeds the allowable inertia range, then the feedforward compensation amount is calculated based on the inertia deviation. The allowable inertia range is... ,in The allowable inertia deviation threshold is typically taken as 20% of the initial nominal thermal inertia parameter. > At that time, calculate the feedforward compensation amount: ; in, This is the feedforward compensation coefficient, with a value ranging from 0.1 to 0.5.

[0065] The feedforward compensation is added to the temperature deviation value and then input to the closed-loop temperature controller after parameter adjustment to generate a control command for the temperature actuator that suppresses overshoot. The temperature deviation value after adding the feedforward compensation is: ; Will The input is fed into the adjusted PI controller to generate control commands for the temperature control actuator.

[0066] In this embodiment, the data missing segment reconstruction method is compared as shown in Table 3.

[0067] Table 3 Comparison of Data Missing Segment Reconstruction Methods

[0068] Table 3 compares the reconstruction errors of the single-channel interpolation method and the spatiotemporal joint interpolation method under different data missing segment lengths. As the data missing segment length increases, the reconstruction errors of both methods gradually increase, but the error of the spatiotemporal joint interpolation method is significantly smaller than that of the single-channel interpolation method. When the data missing segment length is 6 seconds, the error reduction rate of the spatiotemporal joint interpolation method reaches 58.26%, significantly improving the accuracy of data reconstruction.

[0069] In this embodiment, by real-time detection of the opening and closing status of doors and windows, dynamically switching the boundary conditions of the space, and updating the boundary constraint matrix of the dynamic heat propagation partial differential equation, the heat propagation model can adapt to changes in spatial boundary conditions. For long periods of missing data, spatiotemporal joint interpolation reconstruction using adjacent spatial reference channels improves the accuracy of data completion. By identifying the system's thermal inertia parameters online, feedforward compensation is introduced when the inertia deviation exceeds limits, suppressing temperature overshoot caused by changes in system thermal inertia. This method effectively addresses complex operating conditions such as changes in spatial boundary conditions, missing data, and changes in system thermal inertia, improving the robustness and adaptability of the temperature control system.

Claims

1. A method for precise temperature control of a temperature controller based on multi-sensor data fusion, characterized in that, The process includes the following steps: obtaining the thermodynamic parameters of the space where the temperature controller is located and the spatial coordinates of multiple sensors, and constructing a dynamic heat propagation partial differential equation based on the thermodynamic parameters and the spatial coordinates; The thermal conduction delay characteristics between each sensor in the multi-sensor system and the temperature control actuator are calculated based on the dynamic heat propagation partial differential equation. Acquire asynchronous temperature data streams output by the multiple sensors according to different sampling periods; Based on the heat conduction delay characteristics, each sampled value in the asynchronous temperature data stream is deduced in reverse along the time axis and mapped to a unified time reference to calculate the predicted temperature of each sensor spatial node at the current moment. The predicted temperatures are weighted and fused to obtain the spatial thermal field fusion temperature at the current moment. The spatial thermal field fusion temperature is input into the closed-loop temperature controller to generate control commands for the temperature controller actuator. The step of calculating the heat conduction delay characteristics between each sensor and the temperature control actuator in the multi-sensor system based on the dynamic heat propagation partial differential equation includes: acquiring airflow velocity vector distribution data in the space; The airflow velocity vector distribution data is introduced as a convection term into the dynamic heat propagation partial differential equation to construct a convection-diffusion coupled partial differential equation. The eigenvalues ​​of the convection-diffusion coupled partial differential equations are solved to extract the main modes of heat propagation. Based on the attenuation coefficient and oscillation frequency of the main heat propagation mode, the dynamic heat conduction delay characteristics between each sensor in the multi-sensor system and the temperature control actuator are calculated. The dynamic heat conduction delay characteristics are dynamically updated as the airflow velocity vector distribution data changes.

2. The method for precise temperature control of a temperature controller based on multi-sensor data fusion according to claim 1, characterized in that, The step of constructing a dynamic heat propagation partial differential equation based on the thermodynamic parameters and the spatial coordinates includes: obtaining the distribution location of each medium in space and the corresponding thermal diffusivity; The space is divided into multi-scale grids, and the boundary conditions of the multi-scale grids are calibrated, including adiabatic boundaries and convective heat transfer boundaries. Based on the thermal diffusivity and the boundary conditions, a dynamic partial differential equation for heat propagation in a heterogeneous medium is established. Based on the spatial coordinates of the multi-sensor, the spatial discrete distance between each node in the dynamic heat propagation partial differential equation and the temperature control actuator is determined. The spatial discrete distance is then substituted into the dynamic heat propagation partial differential equation to solve for the heat conduction delay feature.

3. The method for precise temperature control of a temperature controller based on multi-sensor data fusion according to claim 1, characterized in that, The step of acquiring the asynchronous temperature data stream output by the multiple sensors according to different sampling periods includes: receiving the raw temperature sampling sequence output by the multiple sensors; For any sensor channel in the original temperature sampling sequence, the time interval between adjacent sampling points is calculated. When the time interval exceeds a preset multiple of the standard sampling period corresponding to the sensor channel, it is determined that there is a missing data segment. Within the missing data segment, historical valid sampled values ​​adjacent to the sensor channel are extracted, and the heat flux change rate is calculated in combination with the heat conduction delay characteristics. Based on the heat flux change rate and the historical valid sampled values, interpolation reconstruction is performed to generate a completed asynchronous temperature data stream.

4. The method for precise temperature control of a temperature controller based on multi-sensor data fusion according to claim 1, characterized in that, Based on the heat conduction delay characteristics, the step of reverse extrapolating each sampled value in the asynchronous temperature data stream along the time axis and mapping it to a unified time reference to calculate the predicted temperature of each sensor spatial node at the current moment includes: obtaining the thermal capacity response hysteresis coefficient of each of the multiple sensors; For each sampled value in the asynchronous temperature data stream, the heat conduction delay characteristic is added to the heat capacity response hysteresis coefficient to obtain the comprehensive hysteresis time; Using the comprehensive lag time as the backward extrapolation step size, the sampled values ​​are pushed forward along the time axis, and combined with the integral of the dynamic heat propagation partial differential equation within the comprehensive lag time, the predicted temperature of each sensor spatial node at the current moment after eliminating the spatial heat conduction delay and sensor thermal capacity lag is calculated.

5. The method for precise temperature control of a temperature controller based on multi-sensor data fusion according to claim 1, characterized in that, The step of inputting the space thermal field fusion temperature into the closed-loop temperature controller and generating the control command for the temperature controller actuator includes: calculating the temperature deviation value between the space thermal field fusion temperature and the preset target temperature; Obtain the system thermal inertia parameters corresponding to the dynamic heat propagation partial differential equation. The system thermal inertia parameters are determined based on the total heat capacity of the medium in space and the airflow convection intensity. The proportional gain and integral time constant of the closed-loop temperature controller are dynamically adjusted according to the system thermal inertia parameters. The temperature deviation value is calculated based on the adjusted proportional gain and integral time constant to generate control commands for the temperature control actuator.

6. The method for precise temperature control of a temperature controller based on multi-sensor data fusion according to claim 2, characterized in that, The steps of dividing the space into a multi-scale grid and calibrating the boundary conditions of the multi-scale grid, including the boundary conditions of adiabatic boundaries and convective heat transfer boundaries, include: acquiring real-time detection signals of the opening and closing status of doors and windows in the space. When the opening and closing state of the doors and windows changes, the affected boundary grid area in the positioning space is located; The boundary conditions of the affected boundary grid region are dynamically switched from an adiabatic boundary to a convective heat transfer boundary, or from a convective heat transfer boundary to an adiabatic boundary. The thermal resistance network between the switched boundary grid region and the adjacent grid region is recalculated, and the boundary constraint matrix of the dynamic heat propagation partial differential equation is updated based on the thermal resistance network.

7. The method for precise temperature control of a temperature controller based on multi-sensor data fusion according to claim 1, characterized in that, The step of calculating the dynamic heat conduction delay characteristics between each sensor in the multi-sensor and the temperature control actuator based on the attenuation coefficient and oscillation frequency of the main heat propagation mode includes: performing time-series sliding window monitoring on the airflow velocity vector distribution data and calculating the variance change of the airflow velocity within the sliding window; When the variance change exceeds a preset steady-state threshold, a modal mutation response mechanism is triggered. Under the modal mutation response mechanism, secondary modes with attenuation coefficients lower than a preset attenuation lower limit are extracted from the primary mode of heat propagation; The time delay parameters corresponding to the secondary mode and the time delay parameters corresponding to the main heat propagation mode are weighted and superimposed, and the superposition result is used as the dynamic heat conduction time delay feature under abrupt change conditions.

8. The method for precise temperature control of a temperature controller based on multi-sensor data fusion according to claim 3, characterized in that, The steps of extracting historical valid sampled values ​​adjacent to the sensor channel within the data missing segment, calculating the heat flux change rate in combination with the heat conduction delay characteristics, and performing interpolation reconstruction based on the heat flux change rate and the historical valid sampled values ​​to generate a completed asynchronous temperature data stream include: determining whether the time span of the data missing segment is greater than the single-channel reconstruction limit threshold. If so, then obtain the reference sample value of at least one reference sensor channel adjacent to the sensor channel space within the data missing segment; Based on the spatial distance between the sensor channel and the reference sensor channel and the correlation with historical temperature, spatial interpolation weighting coefficients are constructed. Based on the spatial interpolation weight coefficient, the reference sample value is spatially mapped, and combined with the historical valid sample value, spatiotemporal joint interpolation reconstruction is performed to generate a completed asynchronous temperature data stream.

9. The method for precise temperature control of a temperature controller based on multi-sensor data fusion according to claim 5, characterized in that, The step of obtaining the system thermal inertia parameters corresponding to the dynamic heat propagation partial differential equation, wherein the system thermal inertia parameters are determined based on the total heat capacity of the medium in space and the airflow convection intensity, includes: real-time acquisition of the output power of the temperature control actuator and the rate of change of the fusion temperature of the space thermal field; Based on the ratio of the output power to the rate of change, the system thermal inertia parameters at the current moment are identified online. Calculate the inertia deviation between the current system thermal inertia parameter and the initial nominal thermal inertia parameter; If the inertia deviation exceeds the allowable inertia range, then the feedforward compensation amount is calculated based on the inertia deviation; The feedforward compensation is superimposed on the temperature deviation value and then input to the closed-loop temperature controller after parameter adjustment to generate a control command for the temperature controller actuator that suppresses overshoot.

Images