CN-122021367-A - Wall double-unknown thermodynamic parameter inversion method and system based on physical neural network

CN122021367ACN 122021367 ACN122021367 ACN 122021367ACN-122021367-A

Abstract

The invention provides a wall double-unknown thermal parameter inversion method and system based on a physical neural network, which belong to the technical field of building energy conservation and comprise the steps of acquiring relevant environment parameters and infrared image data at inversion time of a wall to be detected; the method comprises the steps of inputting infrared image data into a convolutional neural network model to obtain temperature field information of the outer wall surface of a wall to be detected, setting an initial curve of the temperature of the inner wall surface based on the smooth characteristic and the moderate hysteresis effect of indoor air temperature, giving an initial value of a heat conduction coefficient, constructing a physical neural network model together according to Fourier heat conduction law and the boundary conditions of convection heat exchange of the inner surface and the outer surface, inputting relevant environment parameters and the temperature field information of the outer wall surface into the physical neural network model, and carrying out forward solving to obtain the heat transfer coefficient and the temperature of the inner wall surface of the wall to be detected. The invention can obtain the stable and reliable wall body heat conductivity coefficient and the inner wall surface temperature.

Inventors

LIU YANFENG
QIU YANYI
ZHOU YONG
Dang Daifeng
Bai Haihang
WANG YINGYING

Assignees

西安建筑科技大学

Dates

Publication Date: 20260512
Application Date: 20260415

Claims (10)

1. The wall double-unknown thermodynamic parameter inversion method based on the physical neural network is characterized by comprising the following steps of: Acquiring relevant environmental parameters and infrared image data of inversion time of a wall to be detected; inputting the infrared image data into a convolutional neural network model to obtain temperature field information of the outer wall surface of the wall to be detected; Setting an initial curve of the temperature of the inner wall surface based on the smooth characteristic and the moderate hysteresis effect of the indoor air temperature, and giving an initial value of a heat conduction coefficient; And inputting the relevant environmental parameters and the temperature field information of the outer wall surface into a physical neural network model, and carrying out forward solving to obtain the heat transfer coefficient and the inner wall surface temperature of the wall body to be measured.
2. The method for inverting the wall double unknown thermal parameters based on the physical neural network according to claim 1, wherein the related environmental parameters comprise indoor air temperature, outdoor air temperature, solar radiation intensity, wind speed, wall thickness, material density and specific heat capacity of the wall.
3. The wall double-unknown thermal parameter inversion method based on the physical neural network according to claim 1, wherein the convolutional neural network model comprises an input layer, a first convolutional layer, a first pooling layer, a second convolutional layer, a second pooling layer, a first full-connection layer, an activation function layer, a second full-connection layer and an output layer which are sequentially connected, and a feature map of an input image is extracted; And (3) outputting a pixel-by-pixel surface temperature value through temperature regression and correction by adopting a surface area statistical method, and summarizing the temperature results of the pixel level according to a time sequence to obtain temperature field information of the outer wall surface.
4. The inversion method of the wall double unknown thermal parameters based on the physical neural network according to claim 1, wherein the physical neural network model is constructed together according to the fourier heat conduction law and the boundary conditions of the convective heat transfer of the inner surface and the outer surface, specifically: The method comprises the steps of constructing a physical neural network model taking a physical information neural network as a framework, wherein the model takes space coordinates x and time t as independent variables to represent temperature field information of an outer wall surface, defining a heat conduction coefficient k as a material parameter, applying positive value and interval constraint, defining an indoor air temperature time sequence as a time boundary quantity, and simultaneously applying amplitude range, change rate and smoothness physical feasibility constraint.
5. The wall double-unknown thermal parameter inversion method based on the physical neural network as claimed in claim 1, wherein the unified evaluation function is: Wherein, the 、、、、 As the weight coefficient of the light-emitting diode, For the residual loss of the fourier equation, Is the boundary residual error of the inner wall surface, The residual is constrained for the data and, For the loss of the boundary condition of the outer wall surface, Is a smoothness constraint term.
6. The wall double-unknown thermal parameter inversion method based on the physical neural network according to claim 5, wherein the loss of the boundary condition of the outer wall surface is: Wherein, the The heat conductivity coefficient of the wall body; The temperature field of the wall body is represented as the temperature of the wall body at the position x and the time t; The thickness of the wall body is; v is the outdoor wind speed; The temperature of the outer wall surface is calculated after iteration; time series for outdoor air temperature; The solar radiation absorptivity of the wall body; is the intensity of solar radiation.
7. The physical neural network-based wall double-unknown thermodynamic parameter inversion method of claim 1, wherein the staged double-parameter inversion strategy comprises: The method comprises the steps of fixing the temperature of an inner wall surface, solving and optimizing the heat conductivity coefficient, wherein under the condition of ensuring that the temperature sequence of the inner wall surface is kept unchanged a priori, the minimum processing is carried out by taking Fourier equation residual loss, outer wall surface boundary condition loss and data constraint residual as sub-targets mainly formed, and then the heat conductivity coefficient of a wall body output by a parameter network is updated; Setting the heat conductivity coefficient and keeping the heat conductivity coefficient fixed, adopting smoothness requirement, amplitude range limitation and change rate constraint, carrying out segmentation or spline type small-amplitude correction operation on an inner wall temperature curve, and carrying out minimization treatment on sub-targets taking Fourier equation residual loss, inner wall boundary residual and data constraint residual as main bodies so as to update the inner wall temperature output by a parameter network.
8. Wall body double unknown thermal parameter inversion system based on physical neural network, which is characterized by comprising: the data acquisition module is used for acquiring relevant environmental parameters and infrared image data of inversion time of the wall to be detected; The outer wall temperature extraction module is used for inputting the infrared image data into a convolutional neural network model to obtain temperature field information of the outer wall surface of the wall to be detected; The model construction module is used for setting an initial curve of the temperature of the inner wall surface based on the smooth characteristic and the moderate hysteresis effect of the indoor air temperature and giving an initial value of a heat conduction coefficient; And the inversion module is used for inputting the relevant environmental parameters and the temperature field information of the outer wall surface into a physical neural network model, and carrying out forward solution to obtain the heat transfer coefficient and the inner wall surface temperature of the wall body to be measured.
9. An electronic device comprising a memory and a processor, and computer instructions stored on the memory and running on the processor, which when executed by the processor, perform the physical neural network-based wall double unknown thermodynamic parameter inversion method of any one of claims 1-7.
10. A computer readable storage medium storing computer instructions which, when executed by a processor, perform the physical neural network based wall double unknown thermal parameter inversion method of any one of claims 1-7.

Description

Wall double-unknown thermodynamic parameter inversion method and system based on physical neural network Technical Field The invention belongs to the technical field of building energy conservation, and particularly relates to a wall double-unknown thermal parameter inversion method and system based on a physical neural network. Background The statements in this section merely provide background information related to the present disclosure and may not necessarily constitute prior art. Under the background that the building stock is huge in scale and the building new speed is high every year, the thermal performance of the building enclosure structure becomes a key factor affecting the building energy consumption and the indoor environment quality. The method directly determines the heating and refrigerating load and the stability of the indoor heat and humidity environment, and is an essential parameter in building energy-saving design, operation optimization and transformation evaluation. Therefore, it is necessary to develop a thermal inspection of the building envelope. On the one hand, the attenuation condition of the heat insulation performance of the building and the deviation problem in the construction process can be accurately identified through thermal detection. Based on the detection result, the potential of energy-saving reconstruction of the existing building can be quantified, and a scientific and reasonable energy-saving reconstruction scheme is formulated, so that the energy utilization efficiency of the building is improved. On the other hand, the thermal detection data can be used as an important calibration basis for green evaluation of the existing building, and a quantitative support is provided for green authentication of the building. At present, the thermal engineering detection of the enclosure structure mainly comprises a hot box method, a heat flow meter method, a plane heat source method and other methods. Although the method can acquire related information such as thermal parameters of the building envelope, a series of problems generally exist, namely firstly, the method has strict requirements on environmental conditions, often needs a long-time steady-state environment or a specific working condition, has long detection period, is easily influenced by external disturbance such as solar radiation, wind speed, humidity and the like, secondly, the operation such as sensor arrangement or open pore sampling is complex in the detection process, damage to a certain degree is caused to the building, thirdly, most of the method belongs to point location or small area measurement, the conditions such as large-area heterogeneous structures, thermal bridge effect and the like are difficult to comprehensively and accurately reflect, fourthly, the degree of dependence on priori information such as material layers, thickness and the like is higher, and when the on-site parameters are uncertain, the recognition accuracy is obviously reduced. Taken together, these problems lead to the difficulty of existing detection methods in meeting the requirements for efficient assessment of a vast array of existing buildings. Disclosure of Invention In order to overcome the defects of the prior art, the invention provides a wall double-unknown thermal parameter inversion method and system based on a physical neural network, which realize stable and accurate identification of the heat conductivity coefficient of a wall under the severe conditions of unsteady working condition, less prior information and nondestructive detection, synchronously acquire the time-varying change result of the temperature of the inner wall surface, and provide solid and reliable technical support for energy efficiency diagnosis and energy conservation reconstruction evaluation of the existing building. To achieve the above object, one or more embodiments of the present invention provide the following technical solutions: in a first aspect, the invention discloses a wall double unknown thermal parameter inversion method based on a physical neural network, which comprises the following steps: Acquiring relevant environmental parameters and infrared image data of inversion time of a wall to be detected; inputting the infrared image data into a convolutional neural network model to obtain temperature field information of the outer wall surface of the wall to be detected; Setting an initial curve of the temperature of the inner wall surface based on the smooth characteristic and the moderate hysteresis effect of the indoor air temperature, and giving an initial value of a heat conduction coefficient; And inputting the relevant environmental parameters and the temperature field information of the outer wall surface into a physical neural network model, and carrying out forward solving to obtain the heat transfer coefficient and the inner wall surface temperature of the wall body to be measured. In a second aspect, the invention di