Passive radiation refrigeration film reverse design method based on machine learning

Abstract

The invention discloses a reverse design method of a passive radiation refrigeration film based on machine learning, which comprises the following steps of: collecting optical characteristic data of a base material through experiments, establishing a basic data set, and dividing into a training set and a test set; on the basis, a prediction model from the components and the thickness to the optical characteristic data and a reverse design model from the optical characteristic data to the components and the thickness are established. Training the prediction model and the reverse design model, calculating initial design values of components and thickness under target optical characteristics through the reverse design model, comparing the initial design values with optical characteristic data obtained by optical simulation, and finishing the design of the passive radiation refrigeration film when a mean square error meets requirements.

Description

Technical Field

[0001] The present invention belongs to the technical field of passive energy utilization, and in particular, relates to a passive radiation cooling film reverse design method based on machine learning Background Art

[0002] As global warming and energy crisis become increasingly serious, energy conservation and emission reduction have gradually become important research directions in various fields. Traditional active cooling methods face problems such as low efficiency and environmental pollution. Passive radiative cooling technology uses the high emissivity of materials in the mid-infrared range to transfer heat to space by radiation, thereby achieving natural cooling. As a new type of energy utilization technology, passive radiative cooling does not rely on traditional electric drive systems, has significant energy-saving advantages and environmental protection characteristics, and has shown broad application prospects in the fields of building energy conservation, cold chain transportation, and electronic equipment heat dissipation. However, the manufacturing of passive radiative cooling films still faces many challenges. In the traditional passive radiative cooling film manufacturing process, the selection of film substrates and the design of films mainly rely on experience and experiments, lacking systematic theoretical guidance, and the existing design methods usually rely on a single material property, such as emissivity or refractive index, ignoring the interaction of multiple optical properties and material thickness, resulting in greater uncertainty in the design process. In addition, the selection and optimization of substrates by traditional design methods often rely on a large amount of experimental data and repeated verification, resulting in low design efficiency and difficulty in coping with changing environmental conditions and changing application requirements.

[0003] In the prior art, Su et al. [1] Two structures, a quadrilateral prism metasurface and a truncated cone metasurface, were constructed, and the K nearest neighbor algorithm was used to predict their optimal structural parameters and corresponding absorptivity / emissivity in the solar spectrum and atmospheric window, providing a new idea for the design of passive radiation coolers. However, when constructing the model, the description of the data preprocessing process is relatively simple, and there may be problems such as data noise and data imbalance that have not been explored in depth, which may affect the generalization ability of the model. Ding et al. [2] The complex relationship between bionic metamaterials and their spectral responses is revealed through deep learning models, which accelerates the design and optimization of robust bionic radiative cooling metamaterials and makes significant progress in standardized passive radiative cooling applications.

[0004] In response to the above problems, the present invention proposes a passive radiative cooling film reverse design method based on machine learning. This method combines machine learning technology with the design process of passive radiative cooling film, optimizes the selection of appropriate material types and thicknesses based on specific spectral emissivity requirements, and verifies its cooling effect through optical simulation. Compared with traditional design methods, the present invention significantly improves the efficiency and accuracy of design by establishing an efficient performance prediction model and reverse design model, while effectively reducing experimental and development costs.

[0005] References

[0006] [1]Wei Su; Zhipeng Ding; Yinlong Luo; Lipengan Ye; Hong Wu; HongbingYao. Machine learning-enabled design of metasurface based near-perfect daytimeradiative cooler[J]. Solar Energy Materials and Solar Cells, 2023, 260: 112488.

[0007] [2] Zhenmin Ding; Xin Li; Qingxiang Ji; Yunce Zhang; Honglin Li; Hulin Zhang; Lorenzo Pattelli; Yao Li; Hongbo Xu; Jiupeng Zhao. Machine-learning-assisted design of a robust biomimetic radiative cooling metamaterial[J]. ACS Materials Letters, 2024, 6(6): 2416-2424. Summary of the invention

[0008] The purpose of the present invention is to propose a reverse design method for passive radiative cooling films based on machine learning. By combining machine learning technology with the design requirements of passive radiative cooling films, the design process of passive radiative cooling films is optimized to reduce the time and cost of the design process.

[0009] To achieve the purpose of the present invention, the present invention adopts the following technical solutions:

[0010] S10 collects optical property data of passive radiative cooling film substrates through experiments and establishes a basic data set;

[0011] S20 preprocesses the optical property data collected by S10 and divides the data into training set and test set using K-fold cross validation method;

[0012] S30 builds a prediction model from the composition and thickness of the passive radiation cooling film to the optical property data based on the basic data set and performs training;

[0013] S40 builds a reverse design model from optical property data to the composition and thickness of passive radiation cooling film based on the basic data set and performs training;

[0014] S50 inputs the target optical property data into the optical property data trained in S40 into the composition and thickness inverse design model of the passive radiation cooling film to obtain the initial design values ​​of the composition and thickness of the passive radiation cooling film;

[0015] S60 performs optical simulation on the initial design values ​​of the composition and thickness of the passive radiative cooling film obtained in S50, and predicts the emissivity of the passive radiative cooling film in the atmospheric window band and the reflectivity in the solar band;

[0016] S70 calculates the mean square error between the emissivity of the passive radiation cooling film in the atmospheric window band and the reflectivity in the solar band obtained in S60 and the target optical characteristic data in S50;

[0017] S80 determines whether the mean square error is less than 0.1. If the mean square error is less than 0.1, the design of the passive radiation cooling film is completed. If the mean square error is greater than or equal to 0.1, steps S30 to S80 are repeated.

[0018] In S10, the passive radiation cooling film substrate is one or more of Al, Al2O3, Au, BaF2, Be, CaF2, Cr, Cu, Fe, Mn, KBr, SiO2, Si3N4, SiC, Ti, TiO2, ZnO, ZnS, PDMS, PE, PVA, and PVDF.

[0019] The optical characteristic data include one or more of the refractive index, reflectivity, emissivity and extinction coefficient of the passive radiation cooling film substrate at a corresponding wavelength.

[0020] The collected basic data sets are divided into the following two matrices:

[0021] X i ={element i ,h i};

[0022] Y i ={n i ,R i ,ε i ,ki ,}

[0023] Among them, h i is the thickness of the passive radiation cooling film substrate; n i is the refractive index of the passive radiation cooling film substrate; R i is the reflectivity of the passive radiation cooling film substrate; ε i k is the emissivity of the passive radiative cooling film substrate; i is the extinction coefficient of the passive radiative cooling film substrate.

[0024] In S20, the optical characteristic data collected in S10 are preprocessed, and data cleaning, standardization and normalization are performed to remove outliers and missing data to ensure data quality.

[0025] First, a feature d is filled in by the mean method. j Fill in the missing values:

[0026]

[0027] Where M is the number of non-missing values.

[0028] Let a certain eigenvalue d j An outlier is defined as a value greater than the mean μ j 3 times the standard deviation σ j , and remove outliers:

[0029] d j -μ j |>3σ j

[0030] The Robust standardization method was used to standardize the variables and compress the data to the range of [-1.1];

[0031]

[0032] Among them, Q1 and Q3 are the 25th and 75th percentiles of the data respectively, and Q2 is the median.

[0033] Assume that feature d j The minimum and maximum values ​​are min j and max j , then the normalization formula is:

[0034]

[0035] The K-fold cross-validation method is used to divide the data set into a training set and a test set, and K=10. This is to reduce the risk of overfitting of the model on the basic data set, thereby improving the generalization ability of the model and ensuring the efficiency and accuracy of the machine learning model.

[0036] The step S30 specifically includes:

[0037] S31 takes the composition and thickness of the passive radiation cooling film in the basic data set as input and the optical characteristic data as output, and constructs a prediction model from the composition and thickness of the passive radiation cooling film to the optical characteristic data;

[0038] X i As the input feature vector of composition and thickness, Y i The corresponding optical characteristic data is used as an output vector.

[0039] The convolutional neural network model is used for prediction. For one-dimensional data, we use a one-dimensional convolution operation:

[0040] F1=X*W1+b1

[0041] Among them, X is the input data, * is the convolution operation, W1 is the convolution kernel, b1 is the bias term, and F1 is the output feature map of the convolution layer.

[0042] Use ReLU activation function for nonlinear transformation:

[0043] A1=ReLU(F1)

[0044] The ReLU function is defined as:

[0045] ReLU(x)=max(0,x)

[0046] Max pooling is used to select the maximum value in the local area, thereby reducing the dimension of the data and improving the robustness of the model:

[0047] P1=MaxPool(A1)

[0048] After the convolution and pooling operations, the extracted features are mapped to the final output through the fully connected layer. Each node is connected to all nodes in the previous layer:

[0049]

[0050] in, is the predicted optical property data of the model, W2 and b2 are the weights and biases of the fully connected layer.

[0051] Finally, a prediction model is constructed from the composition and thickness of the passive radiation cooling film to the optical properties data:

[0052]

[0053] in, is the optical property output by the model, x is the composition and thickness of the input, f is the mapping function learned by the convolutional neural network, and θ is the parameter (including weights and bias) of all convolutional layers, pooling layers, and fully connected layers.

[0054] S32 uses the training set to train the prediction model of the passive radiation cooling film from the composition and thickness to the optical characteristic data constructed in S31; uses the test set to test the accuracy of the prediction model of the passive radiation cooling film from the composition and thickness to the optical characteristic data after the training;

[0055] S33 selects mean square error as the loss function to measure the gap between the predicted value and the actual value:

[0056]

[0057] Select the Adam optimizer and adaptively adjust the learning rate. The update formula of Adam is:

[0058]

[0059]

[0060] Where α is the learning rate, β1 and β2 are the decay coefficients of the first and second order moments, and ε is a constant to prevent division by zero.

[0061] The training method selected by S34 is one of forward propagation, loss function calculation, back propagation and parameter update; the prediction model of the composition and thickness of the passive radiation cooling film to the optical property data is trained to an accuracy of not less than 0.9.

[0062] Among them, the forward propagation will input the optical characteristic data y i The data is passed into the convolutional neural network and after passing through each layer (convolutional layer, pooling layer, fully connected layer), a prediction value is generated. That is, composition and thickness data.

[0063] The backpropagation algorithm calculates the gradient of the loss function with respect to each parameter, propagates the error from the output layer back to the input layer via the chain rule, and updates each weight in the network.

[0064] The Adam optimizer is used to update the parameters of the convolutional neural network according to the calculated gradients.

[0065] Repeat the above steps until the loss function converges and the training accuracy reaches the requirement (not less than 0.9). The training goal is to make the inverse design model from optical property data to composition and thickness data perform well on the test set.

[0066] The step S40 specifically includes:

[0067] S41 takes the optical characteristic data in the basic data set as input and the composition and thickness of the passive radiation cooling film as output, and constructs a reverse design model from the optical characteristic data to the composition and thickness of the passive radiation cooling film;

[0068] Y i As the input feature vector of optical property data, X i as the corresponding composition and thickness as output vectors.

[0069] The convolutional neural network model is used for prediction, and the model building process is similar to S31.

[0070] S42 uses the training set to train the inverse design model of the composition and thickness of the passive radiative cooling film from the optical characteristic data constructed in S41; and uses the test set to test the accuracy of the inverse design model of the composition and thickness of the passive radiative cooling film from the trained optical characteristic data. The model training process is similar to S32;

[0071] S43 selects mean square error as the loss function, selects Adam optimizer, and adaptively adjusts the learning rate;

[0072] The training method selected by S44 is one of forward propagation, loss function calculation, back propagation and parameter update; the optical property data is trained into an inverse design model of the composition and thickness of the passive radiation cooling film to an accuracy of not less than 0.9.

[0073] In S50, the target optical characteristic data are: the emissivity of the passive radiation cooling film in the atmospheric window band and the reflectivity in the solar band.

[0074] As a preferred example, the emissivity of the passive radiation cooling film in the atmospheric window band is not less than 0.9, and the reflectivity in the solar band is not less than 0.9.

[0075] In S60, the optical simulation uses a transfer matrix method to simulate the optical response of the passive radiative cooling film to calculate the optical characteristic data of the passive radiative cooling film substrate at different wavelengths.

[0076] Assuming that light waves are incident from air onto the passive radiative cooling film, and there is another medium (such as air or substrate) underneath the film, the entire system can be regarded as a three-layer structure.

[0077] For passive radiative cooling films, the core of the transfer matrix method for calculating reflectivity and transmittance lies in the optical impedance and phase change of each layer.

[0078] Optical impedance Z0 in air: The impedance of air is generally known and is approximately a constant Z0=1 (in light wave propagation).

[0079] Optical impedance Z in passive radiative cooling films film : The impedance of the film and its refractive index n film and the incident angle θ film Related, expressed as:

[0080]

[0081] Transfer matrix M of passive radiative cooling film film for:

[0082]

[0083] Where δ = k film d film is the phase change of the passive radiative cooling film, is the wave number in the film, d film is the thickness of the passive radiation cooling film, λ is the wavelength of the incident light, Z film is the optical impedance of the film, n film is the refractive index of the film.

[0084] Reflectivity refers to the energy ratio of incident light reflected back from the film surface. The reflection between the film and air can be calculated by the transfer matrix method:

[0085]

[0086] Where Z0 is the optical impedance of air, Z film is the optical impedance of the film, θ air and θ film are the incident angles of light in air and film, respectively.

[0087] Transmittance refers to the energy ratio of light that passes through the film, and the calculation formula is:

[0088]

[0089] Emissivity reflects the ability of a film material to radiate heat energy in a specific wavelength band. It can be calculated using the following formula:

[0090] ε=1-RA

[0091] A=1-RT

[0092] Where A is the absorptivity of the film and ε is the emissivity of the film.

[0093] The extinction coefficient is related to the refractive index of the film and the absorption characteristics of the material. Its calculation formula is:

[0094]

[0095] Among them, Im(n film ) is the imaginary part of the film's complex refractive index.

[0096] Calculate the mean square error between the predicted emissivity of the passive radiative cooling film in the atmospheric window band and the reflectivity in the solar band obtained through optical simulation and the target optical property data. If the mean square error is less than 0.1, the design is completed; if the error is large, repeat steps S30-S80, readjust the composition and thickness, and further optimize the design.

[0097] The present invention has the following advantages and beneficial effects:

[0098] The present invention combines machine learning with the design of passive radiative cooling films to achieve precise optimization of the structure and composition of passive radiative cooling films according to the target spectral emissivity. Unlike traditional empirical design, this method uses experimental data and numerical simulation to automatically adjust the composition of the film substrate and the thickness of the film, thereby optimizing the cooling performance of the radiative film, significantly shortening the design cycle and reducing development costs. This technology not only improves design efficiency and accuracy, but also enables customized design according to different application requirements. It has broad application potential, especially in the fields of building energy conservation, electronic equipment heat dissipation, photovoltaic systems and air conditioning, promoting the development of low-carbon and green technologies. BRIEF DESCRIPTION OF THE DRAWINGS

[0099] Figure 1 It is a flowchart of the present invention;

[0100] Figure 2 is an implementation flow chart of an embodiment of the present invention;

[0101] Figure 3 Optical simulation diagram constructed for the present invention. DETAILED DESCRIPTION

[0102] The technical solution of the embodiment of the present invention is described in detail below with reference to the accompanying drawings.

[0103] This embodiment uses a machine learning algorithm to optimize the design of a passive radiative cooling film. The research goal is to achieve efficient cooling by optimizing the film's emissivity in the atmospheric window band and reflectivity in the solar band, especially to enable radiative cooling during both daytime and nighttime. The machine learning-based optimization method helps to automatically and accurately find the most suitable substrate composition and combination, as well as the film thickness.

[0104] Al2O3 (bauxite), PDMS (polydimethylsiloxane), SiO2 (silicon dioxide), and Al (aluminum) were selected as film substrates, and the optical property data of passive radiative cooling film substrates were collected through experiments to establish a basic data set;

[0105] The basic data set was preprocessed and the missing values ​​in the experimental data were interpolated to ensure the integrity of the data. The optical property data were normalized using the Robust normalization method to compress the data into the range of [-1,1], thereby reducing the impact of dimensional differences on the machine learning model. The data was divided into 10 subsets and K-fold cross validation (K=10) was used to ensure the generalization ability of the model.

[0106] Machine Learning Model Training:

[0107] 1) Prediction model from composition and thickness to optical property data:

[0108] The prediction model was constructed using composition and thickness as input and optical property data as output. The model was trained using a neural network, using the Adam optimizer and mean square error as the loss function. The training accuracy of the model reached above 0.9, ensuring its good prediction ability for the optical properties of passive radiative cooling film substrates.

[0109] 2) Reverse design model from optical property data to composition and thickness:

[0110] A reverse design model is constructed based on the optical property data. The reverse design model inputs the optical property data of the substrate and outputs the composition and thickness of the film. The reverse design model is trained through back propagation and achieves a prediction mean square error of less than 0.1.

[0111] The target optical property data (emissivity of the passive radiative cooling film in the atmospheric window band and reflectivity in the solar band) are input into the trained reverse design model to obtain the composition and thickness of the passive radiative cooling film substrate. The results are shown in the table below.

[0112] Table 1 Initial design results of passive radiative cooling films

[0113]

[0114] The transfer matrix method is used to perform optical simulation on the preliminarily designed film to predict the reflectivity and emissivity of the passive radiative cooling film at different wavelengths.

[0115] The mean square error between the simulated emissivity of the passive radiative cooling film in the atmospheric window band and the reflectivity in the solar band and the target value is calculated. If the error is greater than 0.1, the composition and thickness are readjusted for the next round of training and optimization.

[0116] After multiple rounds of iterations, the final designed film has an emissivity of no less than 0.9 in the atmospheric window band and a reflectivity of no less than 0.9 in the solar band, meeting the target optical property requirements.

[0117] The specific embodiments described above further illustrate the objectives, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above description is only a specific embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A passive radiation cooling film inverse design method based on machine learning, characterized in that: The following steps are involved: S10 collects optical property data of passive radiative cooling film substrates through experiments and establishes a basic data set; S20 preprocesses the optical property data collected by S10 and divides the data into training set and test set using K-fold cross validation method; S30 builds a prediction model from the composition and thickness of the passive radiation cooling film to the optical property data based on the basic data set and performs training; S40 builds a reverse design model from optical property data to the composition and thickness of passive radiation cooling film based on the basic data set and performs training; S50 inputs the target optical property data into the optical property data trained in S40 into the composition and thickness inverse design model of the passive radiation cooling film to obtain the initial design values ​​of the composition and thickness of the passive radiation cooling film; S60 performs optical simulation on the initial design values ​​of the composition and thickness of the passive radiative cooling film obtained in S50, and predicts the emissivity of the passive radiative cooling film in the atmospheric window band and the reflectivity in the solar band; S70 calculates the mean square error between the emissivity in the atmospheric window band and the reflectivity in the solar band obtained in S60 and the target optical characteristic data in S50; S80 determines whether the mean square error is less than 0.

1. If the mean square error is less than 0.1, the design of the passive radiation cooling film is completed. If the mean square error is greater than or equal to 0.1, steps S30 to S80 are repeated.

2. The method for inverse design of passive radiation cooling film based on machine learning according to claim 1, characterized in that: The passive radiation cooling film substrate is one or more of Al, Al2O3, Au, BaF2, Be, CaF2, Cr, Cu, Fe, Mn, KBr, SiO2, Si3N4, SiC, Ti, TiO2, ZnO, ZnS, PDMS, PE, PVA, and PVDF.

3. The method for inverse design of passive radiation cooling film based on machine learning according to claim 1, characterized in that: The optical characteristic data includes one or more of the refractive index, reflectivity, emissivity and extinction coefficient of the passive radiation cooling film substrate.

4. The method for inverse design of passive radiation cooling film based on machine learning according to claim 1, characterized in that: The basic data set includes the composition and thickness of the passive radiation cooling film and the optical property data of the passive radiation cooling film substrate.

5. The method for inverse design of passive radiation cooling film based on machine learning according to claim 1, characterized in that: The pretreatment method comprises: S21 handles missing values ​​or outliers in the data to ensure data quality; S22 uses the Robust standardization method to standardize the variables and compress the data into the range of [-1,1].

6. The method for inverse design of passive radiation cooling film based on machine learning according to claim 1, characterized in that: The K in the K-fold cross validation method is 10.

7. The method for inverse design of passive radiation cooling film based on machine learning according to claim 1, characterized in that: The specific steps of S30 include: S31 takes the composition and thickness of the passive radiation cooling film in the basic data set as input and the optical characteristic data as output, and constructs a prediction model from the composition and thickness of the passive radiation cooling film to the optical characteristic data; S32 uses the training set to train the prediction model of the passive radiation cooling film from the composition and thickness to the optical characteristic data constructed in S31; uses the test set to test the accuracy of the prediction model of the passive radiation cooling film from the composition and thickness to the optical characteristic data after the training; S33 selects mean square error as the loss function, selects Adam optimizer, and adaptively adjusts the learning rate; The training method selected by S34 is one of forward propagation, loss function calculation, back propagation and parameter update; the prediction model of the composition and thickness of the passive radiation cooling film to the optical property data is trained to an accuracy of not less than 0.

9.

8. The method for inverse design of passive radiation cooling film based on machine learning according to claim 1, characterized in that: The specific steps of S40 include: S41 takes the optical characteristic data in the basic data set as input and the composition and thickness of the passive radiation cooling film as output, and constructs a reverse design model from the optical characteristic data to the composition and thickness of the passive radiation cooling film; S42 uses the training set to train the inverse design model of the composition and thickness of the passive radiation cooling film from the optical characteristic data constructed in S41; and uses the test set to test the accuracy of the inverse design model of the composition and thickness of the passive radiation cooling film from the trained optical characteristic data; S43 selects mean square error as the loss function, selects Adam optimizer, and adaptively adjusts the learning rate; The training method selected by S44 is one of forward propagation, loss function calculation, back propagation and parameter update; the optical property data is trained into an inverse design model of the composition and thickness of the passive radiation cooling film to an accuracy of not less than 0.

9.

9. The method for inverse design of passive radiation cooling film based on machine learning according to claim 1, characterized in that: The target optical characteristic data are: the emissivity of the passive radiation cooling film in the atmospheric window band and the reflectivity in the solar band.

10. The passive radiation cooling film inverse design method based on machine learning according to claim 1, characterized in that: The optical simulation uses a transfer matrix method to simulate the optical response of the passive radiative cooling film to calculate the optical characteristic data of the passive radiative cooling film substrate at different wavelengths.

Images