Heating system module design parameter prediction method and device based on generative artificial intelligence

Abstract

The embodiment of the invention discloses a heating system module design parameter prediction method and device based on generative artificial intelligence, and relates to the field of heating system optimization design based on artificial intelligence. Heating system module design parameter analysis is performed at a server side by using temperature difference conditions, thermal load conditions and the like provided by a client side; the method comprises the steps of firstly obtaining original data, temperature difference conditions and thermal load conditions of heating system module design parameters, constructing a training sample database, training a condition diffusion model, and predicting the corresponding heating system module design parameters by using the trained condition diffusion model according to to-be-predicted temperature difference data and thermal load data provided by a client. And finally, returning the corresponding heating system composition module and the design parameters thereof to the client, and on this basis, assisting an engineer in designing the heating system.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of optimization design of heating system based on artificial intelligence, and particularly relates to a heating system module design parameter prediction method and device based on generative artificial intelligence. BACKGROUND

[0002] In order to improve the design efficiency and reduce the manual intervention, in recent years, the method based on machine learning is explored to be used for predicting the component structure of engineering system. This method can not only significantly reduce the work burden of the design personnel, but also provide support for the applications such as CAD integrated assembly, dynamic assembly optimization and part replacement prediction.

[0003] In the design method, the CAD designers usually adopt the bottom-up or top-down method. The bottom-up method starts from the design of single part, and gradually builds the whole, which is suitable for the design of small-scale or standardized products. The top-down method starts from the overall layout, defines the overall assembly structure first, and then refines to the design of parts, which is more advantageous in the design of large and complex products. However, no matter which method is used, how to learn from the prior relationship data of a large number of modular components and accurately predict the assembly structure of a certain system is still a key and challenging task.

[0004] Therefore, how to combine the machine learning technology with the CAD design process, realize the efficient prediction of the design parameters of the modular components of the heating system, and on the basis of obtaining the design parameters and modules, assist the engineers in drawing, optimize the design efficiency and assembly accuracy, has become an important direction of the current research. SUMMARY

[0005] The embodiments of the present application provide a heating system module design parameter prediction method and device based on generative artificial intelligence, which can predict the design parameters of the heating system module through the temperature difference condition and the heat load condition, so as to assist the engineers in improving the efficiency and assembly accuracy of the CAD design of the heating system.

[0006] To achieve the above object, the embodiments of the present application adopt the following technical solutions:

[0007] In a first aspect, the embodiments of the present application provide a method, which comprises:

[0008] The raw data uploaded by the client is accepted, and a training sample database is constructed, wherein the raw data includes temperature difference data and heat load data corresponding to the heating system; the raw data uploaded by the client further includes temperature difference data and heat load data corresponding to the heating system, wherein the module design parameters of the heating system in the raw data are obtained by using the temperature difference condition and the heat load condition of the heating system;

[0009] The conditional diffusion model is trained by using the sample database, and the conditional diffusion model is used to generate the corresponding relationship between the heating system module design parameters and the conditions, so as to construct a heating system module design parameter regression prediction model. The conditional diffusion model includes two opposite processes, namely a forward diffusion process and a reverse denoising process;

[0010] The trained conditional diffusion model is used for prediction analysis on the heating system module to be processed, wherein the prediction analysis object includes the heating system design parameters corresponding to the temperature difference data and the heat load data provided by the client;

[0011] According to the prediction analysis result, the heating system module is screened, and the screened heating system module and the corresponding design parameters are returned to the client.

[0012] Specifically, the data set in the training sample database includes: temperature difference data, heat load data of a heating system sample, and design parameters of a component module corresponding to the heating system sample; wherein the data set in the training sample database is represented as: (Q, K, a, b, c, d, e), Q is a temperature difference condition, K represents a heat load condition, a represents a water pump module design parameter, b represents a plate heat exchanger module design parameter, c represents a high-level water tank module design parameter, d represents a constant pressure expansion tank module design parameter, and e represents a constant pressure water supply pump module design parameter.

[0013] In S2, the process of training the conditional diffusion model includes: taking the temperature difference condition Q and the heat load condition K as the condition characteristics of the conditional diffusion model, taking the time embedding as the time step characteristics of the conditional diffusion model, and taking the heating system module design parameters as the model prediction target; in the forward diffusion process, noise is added to the module design parameter data of the heating system until the module design parameter data becomes pure Gaussian noise after being added with noise for multiple times, and the pure Gaussian noise is in the form of Gaussian distribution data; in the reverse denoising process, the added noise is removed for the pure Gaussian noise obtained in the forward diffusion process until the module design parameter data is restored.

[0014] Further, the conditional diffusion model adds control conditions on the basis of the diffusion model to control the denoising gradient of the model to generate the design parameters of the heating system module under the corresponding feature conditions. In the conditional diffusion model, the denoising gradient adopted is

[0015]

[0016] wherein ε θ (y t ,t,x) is the denoising gradient of the conditional diffusion model, ε θ (y t ,t) is the denoising gradient of the unconditional diffusion model, ω is a weight used to control the gradient between the two, and ε θ represents the prediction noise, y t represents the predicted value after the tth iteration, and x represents the condition label. Based on this, the conditional diffusion model can be used to obtain the design parameters of the heating system module under the corresponding conditions, and a regression prediction model of the design parameters of the heating system module is constructed. Specifically, a fully connected neural network is used to predict the noise of the conditional diffusion model, wherein the loss function used in the training stage is: y i represents the true value of the ith sample, represents the predicted value of the ith sample, and n is the number of samples.

[0017] Specifically, the process of adding noise at each step of the prediction forward process is divided into input feature alignment, diffusion, and dimension reduction links, and a fully connected neural network is used to establish a corresponding functional module for each link, including: the input feature alignment module is composed of two fully connected layers, the diffusion model denoising module is composed of three stacked diffusion modules, and the intermediate feature dimension reduction module is composed of three fully connected layers.

[0018] In the preferred scheme, the structure of the input feature alignment module is: (5, 64, 2), 512; the structure of the diffusion model denoising module is: 512, 512, feature fusion layer, 512, 512; and the structure of the intermediate feature dimension reduction module is: 256, 64, 5.

[0019] The forward diffusion process includes: x t represents the data after adding noise for the tth time, α t is a hyperparameter used to control the rate of diffusion (noise addition), and ε t represents the Gaussian noise added for the tth time.

[0020] The reverse denoising process includes: μ θ (x t ,t) represents the predicted value of the design parameters of the heating system module, is a hyperparameter for controlling the noise adding rate.

[0021] In a second aspect, the embodiments of the present application provide a device, comprising:

[0022] A data preprocessing module is configured to accept original data uploaded by a client and construct a training sample database, wherein the original data comprises temperature difference data and heat load data corresponding to a heating system;

[0023] A training module is configured to train a conditional diffusion model using the sample database, wherein the conditional diffusion model is used to generate a corresponding relationship between a heating system module design parameter and a condition, and the conditional diffusion model comprises a forward diffusion process and a reverse denoising process;

[0024] An execution module is configured to perform predictive analysis on a heating system module to be processed using the trained conditional diffusion model, wherein the object of the predictive analysis comprises a heating system design parameter corresponding to temperature difference data and heat load data provided by the client;

[0025] A sending module is configured to filter the heating system module according to the result of the predictive analysis, and return the filtered heating system module and the corresponding design parameter to the client.

[0026] The training module is specifically configured to take the temperature difference condition and the heat load condition as the condition feature of the conditional diffusion model, take the time embedding as the time step feature of the conditional diffusion model, and take the heating system module design parameter as the model prediction target. In the forward diffusion process, the training module adds noise to the module design parameter data of the heating system until the module design parameter data becomes pure Gaussian noise after being added with noise for multiple times, and the pure Gaussian noise is in the form of Gaussian distribution data. In the reverse denoising process, the training module removes the added noise for the pure Gaussian noise obtained in the forward diffusion process until the module design parameter data is restored.

[0027] The heating system module design parameter prediction method and device based on generative artificial intelligence provided by the embodiments of the present application use the temperature difference condition and the heat load condition provided by the client to perform heating system module design parameter analysis on the server side. First, the original data of the heating system module design parameter, the temperature difference condition and the heat load condition are obtained, a training sample database is constructed, a conditional diffusion model is trained, the corresponding heating system module design parameter is predicted using the trained conditional diffusion model according to the to-be-predicted temperature difference data and heat load data provided by the client, and finally the corresponding heating system composition module and its design parameter are returned to the client, thereby assisting engineers in the design work of the heating system. BRIEF DESCRIPTION OF DRAWINGS

[0028] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the following will briefly introduce the drawings needed in the embodiments. Obviously, the drawings described below only show some embodiments of the present application, and other drawings can also be obtained by those skilled in the art without any creative effort based on these drawings.

[0029] Figure 1 A schematic diagram of a hardware environment provided for the embodiments of the present application;

[0030] Figure 2 A method flowchart provided for the embodiments of the present application;

[0031] Figure 3 A schematic diagram of a heating system module composition provided for the embodiments of the present application;

[0032] Figure 4 A conditional diffusion model structure schematic diagram provided for the embodiments of the present application. DETAILED DESCRIPTION

[0033] In order for those skilled in the art to better understand the technical solutions of the present application, the following will further describe the present application in combination with the drawings and specific embodiments. In the following, the embodiments of the present application will be described in detail, and the examples of the embodiments are shown in the drawings, in which the same or similar reference signs represent the same or similar elements or elements having the same or similar functions throughout. The embodiments described below by referring to the drawings are exemplary and are only used to explain the present application, and cannot be interpreted as a limitation on the present application. Those skilled in the art can understand that, unless specifically stated, the singular forms "a", "an" and "the" used herein can also include the plural forms. It should be further understood that the phrase "comprising" used in the specification of the present application means that the features, integers, steps, operations, elements and / or components exist, but does not exclude the presence or addition of one or more other features, integers, steps, operations, elements, components and / or groups thereof. It should be understood that when we say an element is "connected" or "coupled" to another element, it can be directly connected or coupled to the other element, or there can be intermediate elements. In addition, "connected" or "coupled" used herein can include wireless connection or coupling. The phrase "and / or" used herein includes any one of the associated listed items and all combinations of the associated listed items. Those skilled in the art can understand that, unless otherwise defined, all terms (including technical terms and scientific terms) used herein have the same meaning as that generally understood by those skilled in the art. It should also be understood that terms such as those defined in a general dictionary should be understood to have meanings consistent with those in the context of the prior art, and should not be interpreted in an idealized or overly formal sense unless otherwise defined.

[0034] The embodiment of the present application provides a heating system module design parameter prediction method and device based on generative artificial intelligence. Figure 1 As shown in the client-server composed computing resource allocation system, wherein the client provides the server with original data and temperature difference conditions and heat load conditions, the data is constructed into a training sample database in the server, and a condition diffusion model is trained using the sample database, and the to-be-predicted temperature difference conditions and heat load conditions are input from the client, the conditions are input into the trained condition diffusion model in the server to calculate the heating system design parameters under the corresponding conditions, and the server feeds back the heating system module design parameter prediction information and the corresponding heating system module to the client. Figure 2 As shown in the client-server composed computing resource allocation system, wherein the client provides the server with original data and temperature difference conditions and heat load conditions, the data is constructed into a training sample database in the server, and a condition diffusion model is trained using the sample database, and the to-be-predicted temperature difference conditions and heat load conditions are input from the client, the conditions are input into the trained condition diffusion model in the server to calculate the heating system design parameters under the corresponding conditions, and the server feeds back the heating system module design parameter prediction information and the corresponding heating system module to the client.

[0035] S1, accepting original data uploaded by the client, wherein the original data comprises heating system module design parameters;

[0036] S2, constructing a training sample database;

[0037] S3, training a condition diffusion model using the sample database;

[0038] S4, performing prediction analysis on the to-be-processed heating system module using the trained condition diffusion model, wherein the object of the prediction analysis is the heating system design parameters corresponding to the temperature difference data and the heat load data provided by the client;

[0039] S5, constructing a heating system module database.

[0040] S6, screening the heating system module by means of the prediction result and returning the corresponding module and the design parameters thereof to the client to assist engineers in drawing.

[0041] The original data uploaded by the client further comprises temperature difference conditions and heat load conditions, wherein the heating system module design parameters in the original data are obtained using the temperature difference conditions and the heat load conditions; the data set in the training sample database comprises temperature difference conditions, heat load conditions and corresponding component module design parameters of the heating system. The module composition of the heating system is as shown in the heating system module design parameters comprise water pump module design parameters, plate heat exchanger design parameters, high-level water tank module design parameters, constant pressure expansion tank module design parameters and constant pressure water supply pump module design parameters. The structure of the condition diffusion model is as shown in Figure 3 Figure 4 ​The heating system module design parameters are shown. Specifically, the heating system module design parameters refer to the design parameters of each module in the system corresponding to a certain temperature difference condition and heat load condition. The original data is constructed as model input, that is, the original data is represented as: (Q, K, a, b, c, d, e), wherein Q is the temperature difference, K represents the heat load, a represents the water pump module design parameter, b represents the plate heat exchanger design parameter, c represents the high-level water tank module design parameter, d represents the constant pressure expansion tank module design parameter, and e represents the constant pressure water supply pump module design parameter.

[0042] In the process of training the conditional diffusion model by using the sample database, the following steps are included:

[0043] The model input is taken as the feature condition of the conditional diffusion model, and the heating system module design parameter is taken as the prediction target of the conditional diffusion model. For example, the temperature difference condition and the heat load condition are taken as the feature condition of the conditional diffusion model, and the corresponding module design parameters of the heating system are taken as the prediction target of the model, to obtain the data for model training. In the training stage, the temperature difference data and the heat load data are taken as the condition feature of the conditional diffusion model, the time embedding (artificial setting) is taken as the time step feature of the conditional diffusion model, and the heating system module design parameter is taken as the model prediction target (i.e. supervision information). For example, in the training stage, the temperature difference condition and the heat load condition are taken as the condition feature of the conditional diffusion model, represented as: (Q, K, a, b, c, d, e), and the time embedding is taken as the time step feature of the conditional diffusion model, and the heating system module design parameters a, b, c, d, e are taken as the model prediction target (also called supervision information)

[0044] The conditional diffusion model includes two opposite processes, namely the forward diffusion process and the reverse denoising process. In the forward diffusion process, noise is continuously added to the heating system module design parameter data, and the reparameterization technique is used in this process. The data x t can be obtained by recursion , wherein α t is a hyperparameter for controlling the rate of diffusion (adding noise); ε t is a Gaussian noise, and until the heating system module design parameter data becomes pure Gaussian noise after being added with noise for multiple times.

[0045] In the reverse denoising process, the pure Gaussian noise obtained in the forward diffusion process is continuously removed to restore it to the heating system module design parameter data; this process can be obtained by , wherein the variance is only a constant value related to the hyperparameter α t , and ε θTo predict the noise, the mean μ θ (x t , t) is taken as the predicted value of the heating system module design parameter.

[0046] The conditional diffusion model adds a feature condition to the diffusion model to control the denoising gradient of the model and generate the heating system module design parameter under the corresponding feature condition. In the conditional diffusion model, the conditional diffusion model denoising gradient is: where ε θ (y t , t, x) is the conditional diffusion model denoising gradient, ε θ (y t , t) is the unconditional diffusion model gradient, and ω is used to control the weight between the two gradients. Based on this, the conditional diffusion model can be used to obtain the heating system module design parameter under the corresponding condition, and a regression prediction model of the heating system module design parameter is constructed.

[0047] Further, a fully connected neural network is used as a tool for predicting the noise of the conditional diffusion model, wherein the loss function for the training stage is: where y i represents the real noise, represents the predicted noise, and n is the number of samples.

[0048] In this embodiment, noise prediction is realized based on a fully connected neural network to predict the noise added at each step of the forward process, including feature alignment, diffusion, dimensionality reduction, etc. The structure of the noise prediction of the conditional diffusion model includes a feature alignment module, a diffusion module, and a dimensionality reduction module (all three modules use a fully connected neural network).

[0049] In a preferred scheme of this embodiment, the feature alignment module is composed of two fully connected layers, the diffusion module is composed of three stacked diffusion modules, and the dimensionality reduction module is composed of three fully connected layers. The structure of the input feature alignment module is (2, 64, 5), 512. Wherein 2 represents a 2-dimensional feature condition including temperature difference and heat load (Q, K), wherein the feature is less than five-dimensional design parameter data a, b, c, d, e as the prediction target compared with the model input. 64 represents the dimension of the time embedding (artificially set), and 5 represents the heating system module design parameter data a, b, c, d, e corresponding to the feature condition. The feature alignment module increases the three input features with different dimensions to 512 dimensions through a fully connected neural network;

[0050] The structure of the diffusion model denoising module is: 512, 512, feature fusion layer, 512, 512. The module linearly transforms the aligned features through two 512-dimensional fully connected layers, then performs feature fusion through the feature fusion layer, and then linearly transforms through two 512-dimensional fully connected layers. The diffusion module can be stacked multiple times.

[0051] The structure of the dimension reduction module is: 256, 64, 5. The output features of the foregoing diffusion module (several) are dimensionally reduced to obtain predicted noise. The target dimension of the dimension reduction is consistent with the predicted module design parameter dimension, both being 5-dimensional. By using the reverse denoising process , the noise can be gradually removed from the Gaussian noise to generate the heating system module design parameters under the corresponding feature conditions.

[0052] For example, the conditional diffusion model structure designed in the embodiment is as shown in Figure 4 . The temperature difference data and the heat load data are taken as the feature conditions C of the diffusion model, and the heating system design parameters a, b, c, d, and e are taken as the prediction targets to obtain a data matrix for model training.

[0053] The conditional diffusion model includes two opposite processes, forward diffusion and reverse denoising. The forward diffusion continuously adds noise to the original heating system module design parameter data, and after several steps of noise addition, the data becomes pure Gaussian noise. The reverse denoising continuously removes the added noise from the data with Gaussian distribution to restore the original heating system module design parameter data. The model obtains the data distribution characteristics of the heating system module design parameters hidden in the noise addition / denoising process through training. The noise prediction module is based on a fully connected neural network and includes a feature alignment module, a diffusion module, and a dimension reduction module. The noise prediction module is used to predict the noise added at each step of the forward process.

[0054] The input of the conditional diffusion model is: the t-step prediction value y t , the time step t, and the two-dimensional feature condition C. Since the dimensions of the three inputs are different, the feature dimensions need to be aligned through a fully connected layer; the diffusion module is used to enhance the ability to capture the time step and the feature condition; the dimension reduction module is used to reduce the intermediate features to predict five-dimensional noise ε t , and then generate the t-1-step prediction value y t-1 through the diffusion model denoising. Then the next iteration is performed until y0 is generated, that is, the model predicts the heating system module design parameter data according to the feature conditions.

[0055] In the preferred embodiment, the feature alignment module structure has two fully connected layers, namely (2, 64, 5), 512, where 2 represents the 2D feature conditions including the temperature difference condition and the heat load condition (Q, K), 64 represents the dimension of time embedding (artificially set), and 5 represents the design parameters a, b, c, d, and e of the heating system module corresponding to the feature condition. The input feature alignment module upgrades the input features of three different dimensions to 512 dimensions through a fully connected neural network; the diffusion module structure is 512, 512, feature fusion layer, 512, 512. The module linearly transforms the aligned features through two 512-dimensional fully connected layers, then fuses the features through a feature fusion layer, and then linearly transforms the features through two 512-dimensional fully connected layers. The diffusion module can be stacked multiple times.

[0056] The dimensionality reduction module structure is: 256, 64, 5. This module reduces the dimensionality of the final output features of the aforementioned diffusion module (several). The target dimension of the dimensionality reduction is consistent with the dimension of the predicted heating system module design parameters, both of which are 5 dimensions. The inverse denoising process can be used to gradually remove the noise from the Gaussian noise and generate the heating system module design parameter data under the corresponding characteristic conditions.

[0057] In the training phase: Each round of iteration randomly selects a step size t to add noise to obtain the noisy data y t , the characteristic condition C and the noise step length t-1 are used as the input of the model to obtain the t-step prediction noise ε t , train the model to approximate the real noise ε. The training loss function is: where y i represents the real noise, Represents the prediction noise. In the prediction stage, the model gradually removes Gaussian noise through the prediction noise to achieve the prediction of the design parameters of the heating system module.

[0058] An embodiment of the present invention provides a method and device for predicting design parameters of a heating system module based on generative artificial intelligence. The method uses the heating system module design parameter data, temperature difference conditions, and heat load conditions provided by the client to perform predictive modeling and analysis of the design parameters of the heating system module on the server side. The method first obtains the original data of the module design parameters, temperature difference conditions, and heat load conditions, constructs a sample database, and trains a conditional diffusion model. According to the temperature difference conditions and heat load conditions to be predicted provided by the client, the trained conditional diffusion model is used to predict the corresponding heating system module design parameters. Finally, the heating system module design parameters and the corresponding heating system module under these conditions are returned to the client.

[0059] The main advantage of the embodiment is that the training sample database is built quickly, the condition diffusion model is trained through the sample database, the time period is short, and the implementation is easy; in the prediction stage, a large amount of workload is reduced by relatively artificial calculation and design, the design parameters of the heating system module corresponding to the temperature difference condition and the heat load condition can be obtained efficiently and accurately, and the corresponding module is found based on the design parameters obtained by prediction, the corresponding module and the parameters are provided to the engineers, and the engineers are assisted to complete the design work of the heating system.

[0060] Each of the embodiments in the specification is described in a progressive manner, and the same or similar parts between the embodiments can be referred to each other. Each embodiment focuses on the difference from other embodiments. In particular, for the device embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant part can be referred to the part of the method embodiment. The above is only a specific implementation of the present application, but the protection scope of the present application is not limited to this. Any person skilled in the art can easily think of changes or replacements within the technical range disclosed by the present application, which should be covered within the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

Claims

1. A generative artificial intelligence based heating system module design parameter prediction method, characterized by, The application comprises the following steps: accepting raw data uploaded by a client and constructing a training sample database, wherein the raw data comprises temperature difference data and heat load data corresponding to a heating system; training a conditional diffusion model using the sample database, wherein the conditional diffusion model is used to generate a corresponding relationship between heating system module design parameters and conditions, and the conditional diffusion model comprises a forward diffusion process and a reverse denoising process; performing predictive analysis on a heating system module to be processed using the trained conditional diffusion model, wherein the object of the predictive analysis comprises heating system design parameters corresponding to temperature difference data and heat load data provided by the client; selecting a heating system module according to the result of the predictive analysis and returning the selected heating system module and corresponding design parameters to the client.

2. The method of claim 1, wherein, The data set in the training sample database comprises temperature difference data, heat load data and design parameters of constituent modules of a corresponding heating system sample; wherein the data set in the training sample database is represented as (Q, K, a, b, c, d, e), Q is a temperature difference condition, K represents a heat load condition, a represents water pump module design parameters, b represents plate heat exchanger module design parameters, c represents high-level water tank module design parameters, d represents constant-pressure expansion tank module design parameters, and e represents constant-pressure water supply pump module design parameters.

3. The method according to claim 1 or 2, characterized in that, The process of training the conditional diffusion model comprises the following steps: taking the temperature difference condition Q and the heat load condition K as the condition characteristics of the conditional diffusion model, taking time embedding as the time step characteristics of the conditional diffusion model, and taking heating system module design parameters as the model prediction target; in the forward diffusion process, noise is added to the module design parameter data of the heating system until the module design parameter data becomes pure Gaussian noise after multiple noise addition, and the pure Gaussian noise is in the form of Gaussian distribution data; in the reverse denoising process, the added noise is removed for the pure Gaussian noise obtained in the forward diffusion process until the module design parameter data is restored.

4. The method of claim 3, wherein, The application further comprises the following steps: In the conditional diffusion model, the denoising gradient used is and where ε θ (y t ,t,x) is the conditional diffusion model denoising gradient, ε θ (y t ,t) is the unconditional diffusion model denoising gradient, ω is the weight used to control the weight between the two gradients, ε θ represents the prediction noise, y t represents the predicted value after the tth iteration, and x represents the conditional label.

5. The method of claim 3, wherein, a process of predicting the noise added at each step of the forward process, which comprises input feature alignment, diffusion and dimension reduction links, and each link utilizes a fully connected neural network to establish a corresponding functional module, including: an input feature alignment module composed of two fully connected layers, a diffusion model denoising module composed of three stacked diffusion modules, and an intermediate feature dimension reduction module composed of three fully connected layers.

6. The method of claim 5, wherein, The structure of the input feature alignment module is (5, 64, 2), 512; the structure of the diffusion model denoising module is 512, 512, a feature fusion layer, 512, 512; the structure of the intermediate feature dimension reduction module is 256, 64, 5.

7. The method of claim 3, wherein, The forward diffusion process comprises the following steps: x t denotes the data after the t-th addition of noise, a t is a hyperparameter used to control the rate of diffusion (additive noise), e t denotes the t-th added Gaussian noise; The reverse denoising process comprises the following steps: μ θ (x t ,t) represents a predicted value of a heating system module design parameter, is a hyperparameter, ε θ represents a prediction noise, used to control the rate of noise addition. 8.A device for predicting a module design parameter of a heating system based on generative artificial intelligence, characterized by, The application comprises the following steps: a data preprocessing module for accepting raw data uploaded by a client and constructing a training sample database, wherein the raw data comprises temperature difference data and heat load data corresponding to a heating system; The training module is configured to train a conditional diffusion model by using the sample database, the conditional diffusion model is configured to generate a correspondence between a heating system module design parameter and a condition, and the conditional diffusion model comprises a forward diffusion process and a reverse denoising process. The execution module is configured to perform predictive analysis on the heating system module to be processed by using the trained conditional diffusion model, wherein the object of the predictive analysis comprises a heating system design parameter corresponding to the temperature difference data and the heat load data provided by the client. The sending module is configured to screen the heating system module according to the result of the predictive analysis, and return the screened heating system module and the corresponding design parameter to the client.

9. The apparatus of claim 8, wherein, The training module is specifically configured to take the temperature difference condition and the heat load condition as a condition feature of the conditional diffusion model, take time embedding as a time step feature of the conditional diffusion model, and take the heating system module design parameter as a model prediction target.

10. The apparatus of claim 9, wherein, In the forward diffusion process, the training module adds noise to the module design parameter data of the heating system until the module design parameter data becomes pure Gaussian noise after being added with noise for multiple times, and the pure Gaussian noise is in the form of Gaussian distribution data. In the reverse denoising process, the training module continuously removes the added noise for the pure Gaussian noise obtained in the forward diffusion process until the module design parameter data is restored.

Images