A furnace oxygen concentration prediction method based on a denoising diffusion model

Abstract

The application discloses a kind of based on denoising diffusion model's furnace oxygen concentration prediction method, belongs to energy process soft measurement modeling technical field.It includes the following steps: (1) obtains furnace flame image data;(2) furnace flame image data preprocessing and dataset division;(3) establish prediction model based on U-Net;(4) construct joint prediction model by fusing DDPM, and use flame image dataset to train model;The application proposes a kind of based on denoising diffusion model's furnace oxygen concentration prediction method, the process of U-Net output feature map is combined with the reverse denoising process of DDPM, improve the feature extraction capability of model, to improve the prediction accuracy of model using flame image data.

Description

Technical Field

[0001] The present invention relates to the field of soft-sensing modeling of energy processes, and in particular to a method for predicting oxygen concentration in a furnace based on a denoising diffusion model. Background Art

[0002] In industrial combustion systems, combustion efficiency can be determined by measuring the efficiency of the fuel's heat conversion. To reduce operating costs and meet environmental regulations, combustion efficiency and exhaust gas emissions should be controlled at appropriate levels. Gas analyzers are commonly used to measure the oxygen and nitrogen oxide levels in exhaust gases, but this method produces delays, and oxygen-based feedback controllers tend to overcompensate. Online flame images provide sufficient information to reflect the current combustion state. Therefore, building a soft sensing model based on deep learning technology that can predict the oxygen concentration in the furnace based on flame images has become a possible alternative to expensive hardware sensors.

[0003] A flame can be divided into the core, inner flame, and outer flame. These regions exhibit distinct properties depending on oxygen concentration. Therefore, semantic segmentation of flame images helps improve the model's ability to capture information from different regions within flame images. The U-Net (U-shape Network) is a convolutional neural network used for image segmentation. It utilizes a U-shaped network structure and skip connections, effectively capturing feature information at different levels, thereby improving image segmentation accuracy and detail preservation. The powerful performance of the U-Net has led to its widespread use in various fields. The Denoising Diffusion Probabilistic Model (DDPM) also utilizes a variant of the U-Net to achieve high-quality data generation. The DDPM model is a generative model that samples from standard Gaussian noise and gradually removes the noise through a reverse process to generate new data. This model can be divided into two steps: a forward process and a reverse process. In the forward process, Gaussian noise is gradually added to the original data until the data is completely converted to standard Gaussian noise. The reverse process gradually removes the noise until the data is restored. Based on this, a neural network is trained to predict the amount of noise that needs to be removed at each step in the reverse process. Considering that the process of the U-Net model outputting feature maps after semantic segmentation based on the original image can be regarded as the process of extracting effective features after removing useless information from the original image, which is similar to the DDPM inverse denoising process to a certain extent, the present invention proposes a soft measurement modeling method for furnace oxygen concentration prediction based on a denoising diffusion model. First, a soft measurement model is designed based on U-Net for oxygen concentration prediction. At the same time, the process of U-Net outputting feature maps is combined with the DDPM inverse denoising process to improve the feature extraction ability of the model, thereby improving the prediction accuracy of the model using flame image data. Summary of the Invention

[0004] In order to solve the problem of difficulty in predicting key quality variables using image data in industrial processes, the present invention proposes a furnace oxygen concentration prediction method based on a denoising diffusion model. The process of U-Net outputting feature maps is combined with the DDPM inverse denoising process to improve the feature extraction capability of the model, thereby improving the prediction accuracy of the model using flame image data.

[0005] The technical solution adopted by the present invention to solve its technical problem is:

[0006] A method for predicting furnace oxygen concentration based on a denoising diffusion model comprises the following steps:

[0007] (1) Obtain furnace flame image data as original sample data.

[0008] (2) Original sample data preprocessing and dataset division: the obtained original sample data is divided into a training set and a test set, in order to accelerate the convergence speed of the model and reduce the model training time, the training set data and the test set data are normalized.

[0009] (3) Constructing a flame prediction model based on U-Net.

[0010] (4) Constructing a joint prediction model incorporating DDPM: constructing a joint loss function based on a denoising diffusion model, and training the model using a flame image dataset.

[0011] (5) Model performance evaluation: in order to more objectively evaluate the method of the present application, the evaluation indexes root mean square error (RMSE) and mean absolute error (MAE) are introduced.

[0012] Further, the process of (3) is:

[0013] Step 3.1, constructing a prediction model based on U-Net

[0014] U-Net is mainly composed of an encoder and a decoder containing a skip connection. Among them, the main function of the encoder is to map image data to latent vectors, and the main function of the decoder is to reconstruct image data from latent vectors, while the function of the skip connection is to improve the fineness of the image data recovered by the decoder. The present application proposes a prediction model based on U-Net, which first uses U-Net to perform semantic segmentation on flame pictures. The goal of this step is to remove noise and background while extracting features of flame images. Then, the model will use the feature map output by U-Net to predict the oxygen concentration label. Here, if X represents the flame picture data collected for soft measurement modeling, is the corresponding oxygen concentration label, where n represents the number of data, and the oxygen concentration predicted based on the model is Specific model loss function is:

[0015]

[0016] Further, the process of (4) is:

[0017] Step 4.1: Establishing a denoising diffusion model

[0018] DDPM is a latent variable model consisting of two Markov chains: a forward denoising process and a backward denoising process. Formally, in the forward denoising process, noise is added to the original data, while in the backward process, new data is generated from the noise through a series of denoising processes. Here, q(·) represents the forward process and p(·) represents the backward process. For the original data x0~q(x0), in a diffusion forward process consisting of T steps, the data x in step t is t is the data x at step (t-1) t-1 Add Gaussian noise on the basis of Represents Gaussian distribution, it can be expressed as:

[0019]

[0020] in, is the variance sequence used in each step, and its value is between 0 and 1. I represents the unit matrix. If the diffusion step number T is long enough, the final x T It completely becomes random noise. The entire diffusion process can be expressed in the form of a Markov chain as follows:

[0021]

[0022] where q(x 1:T |x0) represents the posterior probability of the forward diffusion process when the initial x0 is given. Furthermore, another α t =1-β t , The following are some of the heavy parameter techniques:

[0023]

[0024] The reverse process can be expressed as the joint probability distribution p θ (x 0:T ), the distribution q(xt|xt-1) is estimated through a neural network, and the inverse process can be expressed in the form of a Markov chain as:

[0025]

[0026] in p θ (x t-1 |x t ) is a parameterized Gaussian distribution with mean μ θ (x t ,t) and variance ∑ θ (x t ,t) is obtained by neural network training, θ represents the trainable parameters in the neural network. The training goal of the neural network is to make the predicted noise consistent with the real noise added in the forward process, and the target loss function is is:

[0027]

[0028] where t ∈ [1, T], ε is Gaussian noise, is the trained neural network, represents the expectation. The final inverse process generates data as:

[0029]

[0030] where z is the added random noise, and satisfies when t > 1, When t = 1, z = 0.

[0031] σ t is the variance coefficient, and in numerical value

[0032] Step 4.2: Constructing the denoising loss function

[0033] The DDPM inverse process has been obtained by step 4.2, that is, the noise is gradually removed from the noisy data until the data is restored, as shown in formula (8). Moving the term of formula (8) can obtain:

[0034]

[0035] Assuming that the feature map output by the U-Net module performs an additional denoising on the original data for the DDPM inverse process, and considering that when t = 1, z = 0, then construct according to formula (9):

[0036]

[0037] where x0 represents the known original flame picture data, and α0, σ0, t0 are unknown parameters, and u0 is the U-Net output feature map. Since the U-Net model is a neural network model that needs to be trained, in order to simplify the calculation, the above unknown parameters are regarded as trainable parameters of the U-Net model. F φ (·) represents the U-Net model, ψ θ (·) represents the noise prediction neural network of DDPM, and φ and θ are trainable parameters in the U-Net model and the noise prediction neural network of DDPM, respectively. The following can be constructed:

[0038] F φ (x0)-η·ψ θ (x0) = x0 (11)

[0039] where η represents the hyperparameter coefficient for balancing the two models, and the denoising loss function is obtained from formula (11)

[0040]

[0041] Step 4.3: Joint training

[0042] In order to improve the feature extraction capability of the model, the application proposes a furnace oxygen concentration prediction method based on a denoising diffusion model. The model contains a U-Net prediction model and a DDPM model. The U-Net prediction model first outputs a feature map from the flame picture, and then predicts the oxygen concentration based on the feature map. In step 4.2, the feature map output by the U-Net model is assumed to be an additional denoising process for the original data by the DDPM reverse process, and a denoising loss function is constructed. Finally, the DDPM model and the U-Net model are trained at the same time, and the noise prediction neural network in the DDPM is used to improve the feature extraction capability of the U-Net model. The joint loss function is composed of the DDPM model loss function, the U-Net model loss function and the denoising loss function. In order to balance these loss functions, a hyperparameter γ is introduced to balance the loss functions:

[0043]

[0044] Where γ>0, then train the model proposed in the application on the training set, and optimize it using the joint loss function. BRIEF DESCRIPTION OF DRAWINGS

[0045] Figure 1 is a flow chart of the method of the application;

[0046] Figure 2 is a comparison chart of the prediction results of the method of the application and other methods;

[0047] Figure 3 is a comparison chart of the absolute error of the method of the application and other methods.

[0048] By using the above-mentioned technology, compared with the prior art, the application has the following beneficial effects:

[0049] The application proposes a furnace oxygen concentration prediction method based on a denoising diffusion model, which combines the process of outputting a feature map from a U-Net with the reverse denoising process of a DDPM, improves the feature extraction capability of the model, and thus improves the prediction accuracy of the model using flame image data. DETAILED DESCRIPTION

[0050] In order to make the purpose, technical scheme and advantages of the application clearer and more apparent, the application will be further described in detail below in combination with the drawings and examples. It should be understood that the specific examples described herein are only used to explain the application and do not limit the application.

[0051] On the contrary, the present invention covers any alternatives, modifications, equivalents, and solutions that fall within the spirit and scope of the present invention as defined by the claims. Furthermore, to facilitate a better understanding of the present invention, certain specific details are described in detail below in the detailed description of the present invention. Those skilled in the art will be able to fully understand the present invention without these details.

[0052] Reference Figures 1 to 3 A soft-sensor modeling method for predicting furnace oxygen concentration based on a denoising diffusion model is provided, the method comprising the following steps:

[0053] (1) Obtaining furnace flame image data: A combustion experiment was conducted using an experimental furnace with a burner model of 5514-6 and industrial heavy oil as fuel. The filtered exhaust gas sampled from the chimney was sent to the exhaust gas analyzer to measure the exhaust temperature, oxygen, carbon dioxide, and carbon monoxide concentrations of the exhaust gas. A charge-coupled device (CCD) camera was used to capture the image of the flame in the furnace through the observation window of the experimental furnace. The output signal of the CCD camera was sent to the on-site computer via the IEEE-1394a interface. During operation, the oxygen concentration in the furnace and the flame image were automatically recorded;

[0054] (2) Preprocessing of original sample data and data set division: The original sample data obtained is divided into a training set and a test set. In order to speed up the model convergence and reduce the model training time, the training set data and the test set data are normalized;

[0055] (3) Establish and train a flame soft measurement model based on U-Net;

[0056] (4) Constructing a joint prediction model incorporating DDPM: constructing a joint loss function based on the denoising diffusion model and training the model using a flame image dataset;

[0057] (5) Model performance evaluation: In order to evaluate the method of this embodiment more objectively, the evaluation indicators Root Mean Square Error (RMSE) and Maximum Absolute Error (MAE) are introduced.

[0058] Specifically, the process of step (2) is:

[0059] Step 2.1: In order to speed up the training of the model, the image data is normalized to the maximum and minimum values, and the normalized data set is divided into a training set and a test set. The training set consists of 110 groups of data, the validation set consists of 44 groups of data, and the test set consists of 44 groups.

[0060] Specifically, the process of step (3) is:

[0061] Step 3.1: Build a prediction model based on U-Net

[0062] U-Net is mainly composed of an encoder and a decoder with skip connections. The main function of the encoder is to map image data into latent vectors, the main function of the decoder is to reconstruct image data from latent vectors, and the function of the skip connection is to improve the fineness of the image data restored by the decoder. This embodiment proposes a prediction model based on U-Net. The model first uses U-Net to perform semantic segmentation on the flame image. The goal of this step is to remove noise and background while extracting the features of the flame image. Afterwards, the model will use the feature map output by U-Net to predict the oxygen concentration label. Here, if X represents the flame image data collected for soft measurement modeling, is the corresponding oxygen concentration label, where n represents the number of data. The oxygen concentration predicted by the model based on the flame image is The loss function of the model at this time is:

[0063]

[0064] Specifically, the process of step (4) is:

[0065] Step 4.1: Build a denoising diffusion model

[0066] DDPM is a latent variable model consisting of two Markov chains: a forward denoising process and a backward denoising process. Formally, in the forward denoising process, noise is added to the original data, while in the backward process, new data is generated from the noise through a series of denoising processes. Here, q(·) represents the forward process and p(·) represents the backward process. For the original data x0~q(x0), in a diffusion forward process consisting of T steps, the data x in step t is t is the data x at step (t-1) t-1 Add Gaussian noise on the basis of Represents Gaussian distribution, it can be expressed as:

[0067]

[0068] in, is the variance sequence used in each step, and its value is between 0 and 1. I represents the unit matrix. If the diffusion step number T is long enough, the final x T It completely becomes random noise. The entire diffusion process can be expressed in the form of a Markov chain as follows:

[0069]

[0070] where q(x1: T |x0) represents the posterior probability of the forward diffusion process given the initial x0. t =1-β t , The following are some of the heavy parameter techniques:

[0071]

[0072] The reverse process can be expressed as the joint probability distribution p θ (x 0:T ), the distribution q(xt|xt-1) is estimated through a neural network, and the inverse process can be expressed in the form of a Markov chain as:

[0073]

[0074] in p θ (x t-1 |x t ) is a parameterized Gaussian distribution with mean μ θ (x t ,t) and variance ∑ θ (x t ,t) is obtained by neural network training, θ represents the trainable parameters in the neural network. The training goal of the neural network is to make the predicted noise consistent with the real noise added in the forward process, and the target loss function is for:

[0075]

[0076] Where t∈[1,T], ε is Gaussian noise, is the trained neural network, Represents expectation. The final reverse process generates data as follows:

[0077]

[0078] Where z is the added random noise, and when t>1, When t=1, z=0.

[0079] σ t is the coefficient of variance, which is numerically

[0080] Step 4.2: Construct denoising loss function

[0081] The DDPM inverse process has been obtained through step 4.2, that is, the noise of the noisy data is gradually removed until the data is restored. Specifically, as shown in formula (21), by shifting the terms in formula (21), we can get:

[0082]

[0083] Assuming that the feature map output by the U-Net module is one additional denoising performed by the DDPM reverse process on the original data, formula (22) is constructed as follows:

[0084]

[0085] Where x0represents the known original flame picture data, and a0, σ0, t0are unknown parameters, and u0is the U-Net output feature map. Since the U-Net model is a neural network model that needs to be trained, the above unknown parameters are regarded as trainable parameters of the U-Net model to simplify the calculation. F φ (·) represents the U-Net model, and ψ θ (·) represents the noise prediction neural network of DDPM, and φ and θ are trainable parameters in the U-Net model and the noise prediction neural network of DDPM, respectively. The following can be constructed:

[0086] F φ (x0)-η·ψ θ (x0)=x0 (24)

[0087] Where η represents the hyperparameter coefficient for balancing the two models. Based on formula (11), the denoising loss function is obtained

[0088]

[0089] Step 4.3: Joint training

[0090] In order to improve the feature extraction capability of the model, the embodiment proposes a furnace oxygen concentration prediction method based on a denoising diffusion model. The model contains a U-Net prediction model and a DDPM model. The U-Net prediction model first outputs a feature map from the flame picture, and then predicts the oxygen concentration based on the feature map. In step 4.2, it is assumed that the feature map output by the U-Net module is one additional denoising performed by the DDPM reverse process on the original data, and a denoising loss function is constructed. Finally, the DDPM model and the U-Net model are trained at the same time, the noise prediction neural network in DDPM is used to improve the feature extraction capability of the U-Net model, and the joint loss function is composed of the DDPM model loss function, the U-Net model loss function and the denoising loss function. In order to balance these loss functions, a hyperparameter γ is introduced to balance the loss functions:

[0091]

[0092] where γ>0, then train the model proposed in the present application on the training set and optimize it using the joint loss function.

[0093] Then train the model proposed in the present embodiment on the training set, select the hyperparameters on the validation set, and finally test the model effect on the test set. Finally, the hyperparameters γ and η are selected as 0.006 and -1500, respectively.

[0094] Specifically, the process of step (5) is as follows:

[0095] Step 5.1: Comparative evaluation of model effect

[0096] In order to compare the model effect, Support Vector Regression (SVR) and Residual Network (Res-Net) are used for comparison with the method proposed in the present embodiment (Proposed). In order to verify whether the joint loss function proposed in the present embodiment works, a flame oxygen concentration prediction model based on U-Net (U-Net Soft Sensor, USS) is used for separate experiments, and the final comparison results are shown in Table 1:

[0097] Table 1 Comparison of model performance

[0098]

[0099] The results show that the indicators of the method proposed in the present application are all less than those of the SVR and Res-Net methods, indicating that the method has better ability to predict oxygen concentration using flame pictures. The results also show that using the joint loss function proposed in the present embodiment can further improve the model performance. From the results, the RMSE value is reduced by 21.83%, and the MAE value is reduced by 17.94%, proving the effectiveness and superiority of the method proposed in the present embodiment.

[0100] The method proposed in the present embodiment is a furnace oxygen concentration prediction method based on a denoising diffusion model. The process of outputting feature maps of U-Net is combined with the inverse denoising process of DDPM to improve the feature extraction capability of the model, thereby improving the prediction accuracy of the model using flame image data.

[0101] The content described in the embodiments of the present specification is only a list of implementation forms of the inventive concept, and the protection scope of the present application should not be considered as limited to the specific forms described in the embodiments, and the protection scope of the present application also extends to equivalent technical means that can be thought of by those skilled in the art according to the inventive concept.

Claims

1. A method for predicting furnace oxygen concentration based on a denoised diffusion model, the method comprising the following steps: (1) Obtaining furnace flame image data as original sample data; (2) Preprocessing of original sample data and data set division: Divide the acquired original sample data into training set and test set; in order to speed up the model convergence and reduce the model training time, normalize the training set data and the test set data; (3) Establish a prediction model based on U-Net for oxygen concentration prediction; (4) Constructing a joint prediction model incorporating DDPM: Constructing a joint loss function based on the denoising diffusion model and training the model using the flame image dataset; in step (4), the process of constructing the denoising loss function is as follows: 1) Generate data formula for the final reverse process: Where z is the added random noise, and when t>1, When t = 1, z = 0; σ t is the coefficient of variance, which is numerically By moving the items, we can get: Assume that the feature map output by the U-Net module is DDPM. The inverse process performs an additional denoising on the original data, and considers that when t = 1, z = 0, and constructs it according to formula (9): Where x0 represents the known original flame image data, α0, σ0 and t0 are unknown parameters, and u0 is the U-Net output feature map; 2) Treat the unknown parameters as trainable parameters of the U-Net model and use F φ (·) represents the U-Net model, ψ θ (·) represents the noise prediction neural network of DDPM, φ and θ are the trainable parameters in the U-Net model and the noise prediction neural network of DDPM, respectively, which can be constructed as follows: F φ (x0)-η·ψ θ (x0) = x0 (11) where η represents the hyperparameter coefficient used to balance the two models; the denoising loss function is obtained based on formula (11) In step (4), the model training adopts joint training, and the specific process is as follows: The DDPM model and the U-Net model are trained simultaneously, and the noise prediction neural network in DDPM is used to improve the feature extraction ability of the U-Net model. The joint loss function is composed of the DDPM model loss function. U-Net model loss function And the denoising loss function At the same time, the hyperparameter γ is introduced to balance the loss function: Where γ>0, then the proposed model is trained on the training set and optimized using the joint loss function; (5) Model performance evaluation.

2. The method for predicting furnace oxygen concentration based on a denoised diffusion model according to claim 1, wherein: The process of (3) is as follows: The U-Net model includes an encoder and decoder with skip connections. The model first uses the U-Net to perform semantic segmentation on the flame image to remove noise and background while extracting flame image features. The model then uses the feature map output by the U-Net to predict the oxygen concentration label. Let X represent the flame image data used for soft sensor modeling, is the corresponding oxygen concentration label, where n represents the number of data. The oxygen concentration predicted based on the model is: Specific model loss function for:

3. A method for predicting furnace oxygen concentration based on a denoised diffusion model according to claim 1 or claim 2, characterized in that: In (4), the process of establishing the denoising diffusion model is as follows: 1) Let q(·) denote the forward process and p(·) denote the reverse process. For the original data x0~q(x0), in the diffusion forward process with T steps, the data x in step t is t is the data x at step (t-1) t-1 Add Gaussian noise to the basis of Represents Gaussian distribution, it can be expressed as: in, is the variance sequence used in each step, and its value is between 0 and 1. I represents the unit matrix. For the entire diffusion process, it is expressed in the form of a Markov chain as follows: where q(x 1：T |x0) represents the posterior probability of the forward diffusion process given the initial x0; 2) Let α t =1-β t , The following are some of the techniques for re-parameterization: Estimate the distribution q(x t |x t-1 ), the reverse process is expressed in the form of a Markov chain as: in p θ (x t-1 |x t ) is a parameterized Gaussian distribution with mean μ θ (x t ,t) and variance ∑ θ (x t ,t) is obtained by neural network training, θ represents the trainable parameters in the neural network; the training goal of the neural network is to make the predicted noise consistent with the real noise added in the forward process, and the target loss function is for: Where t∈[1,T], ε is Gaussian noise, is the trained neural network, represents the expectation; the final reverse process generates data as follows: Where z is the added random noise, and when t>1, When t = 1, z = 0; σ t is the coefficient of variance, which is numerically

Images