CO2 concentration detection model training method and device and thermal power carbon emission monitoring system

Abstract

The invention discloses a CO2 concentration detection model training method and device and a thermal power carbon emission monitoring system. The method comprises the following steps: collecting CO2 with different concentrations and corresponding feature information as labeled sample data; training the generative adversarial network by using the labeled sample data to obtain a trained generator and a trained restorer; a potential representation of the pseudo data is generated by a generator according to the random noise, and then the potential representation of the pseudo data is converted into the pseudo data with the same dimension as the labeled sample data through a restorer; training a CO2 concentration detection model by using the pseudo data and the labeled sample data; according to the CO2 concentration detection model training method based on the generative adversarial network and semi-supervised learning, learning and establishment of the detection model can be realized with data as little as possible, the data acquisition cost is reduced, and the trained model is more accurate and can adapt to various detection conditions.

Description

Technical Field

[0001] The present invention relates to the technical field of industrial carbon emission monitoring, and in particular to a CO 2 Concentration detection model training method, device and thermal power carbon emission monitoring system. Background Art

[0002] With the continuous acceleration of the industrialization process, the impact of industrial production on the environment has become increasingly prominent, posing a threat to the ecological environment and human health. The large-scale emission of anthropogenic greenhouse gases has become one of the important reasons for the increase in extreme climate events. Among the many emission sources, carbon emissions from the power industry are the largest greenhouse gas emitters in my country. Thermal power units, especially coal-fired power units, are the largest source of carbon dioxide (CO 2 ) will become the main emission source and will become the focus of emission reduction. The statistics of carbon emissions is a key link. Therefore, in the field of industrial emission monitoring, CO 2 Real-time online monitoring technology is gradually becoming a research hotspot.

[0003] Optical gas detection methods are widely used due to their advantages such as high precision, high sensitivity and non-contact. Among them, the tunable diode laser absorption spectroscopy (TDLAS) technology uses the tunable characteristics of the laser to scan the absorption spectrum of the gas to be measured, obtain the harmonic signal of the concentration information of the gas to be measured, and then realize gas concentration detection. Since the laser intensity will be significantly attenuated after passing through the gas absorption area, the concentration of the gas to be measured can be detected by detecting the attenuation.

[0004] To achieve the above-mentioned gas concentration detection, the traditional method performs a series of calculations on the acquired harmonic signals to obtain the gas concentration. The calculation process is complicated, and the calculation results will be affected by many factors. The method based on machine learning provides another idea. By learning the relationship model between the light intensity change signal obtained by scanning the absorption spectrum of the gas to be measured and the concentration of the gas to be measured, accurate and fast concentration detection can be achieved. For example, in existing research, the patent application document with publication number CN109724941A is based on TDLAS technology. By extracting 5 statistical feature parameters from the difference signal between the absorbed laser and the original laser signal: mean, standard deviation, sum of squared deviations, coefficient of variation and maximum deviation, and using CO 2 Gas concentration is used as output to train RBF neural network and realize CO 2Accurate measurement and error analysis of gas concentration; however, the manual feature selection method used in this solution may lead to information loss, making the measurement model less applicable, and does not consider the problem of insufficient label data caused by insufficient sample gas concentration values ​​in actual model training. Although the patent application document with publication number CN112504970A uses a generative adversarial network to expand the data set, this solution also uses manual feature extraction, and does not consider the front-back correlation in the spectral signal in the model construction, which may cause the pseudo data generated by the generative adversarial network to be not realistic enough. Summary of the invention

[0005] The technical problem to be solved by the present invention is how to realize CO with as little data as possible. 2 The learning and establishment of concentration detection models can reduce data collection costs while improving CO 2 Generalization ability of concentration detection model and accuracy of carbon concentration detection.

[0006] The present invention solves the above technical problems by the following technical means: A CO 2 The concentration detection model training method includes: Collect different concentrations of CO 2 And the corresponding feature information as labeled sample data; Use labeled sample data to train the generative adversarial network to obtain trained generators and restorers; The generator generates the latent representation of the pseudo data according to random noise, and then the restorer converts the latent representation of the pseudo data into pseudo data with the same dimension as the labeled sample data. Using pseudo data and labeled sample data to analyze CO 2 The concentration detection model is trained.

[0007] Furthermore, the collection of different concentrations of CO 2 And the corresponding feature information as labeled sample data, including: Under different concentrations of standard gas, collect different concentrations of CO 2 And the corresponding feature information as labeled sample data; The characteristic information includes laser signal, gas temperature and gas flow rate.

[0008] Furthermore, the generative adversarial network includes a generator, a discriminator, an embedder and a restorer, the input of the generator is noise data, and the output is connected to the discriminator, the input of the embedder is labeled sample data, and the output is connected to the discriminator and the restorer respectively; The generator is used to generate latent representations of pseudo data based on noisy data; The embedder is used to generate a latent representation of the labeled sample data based on the labeled sample data; The discriminator is used to determine whether the output of the generator and the output of the embedder are labeled sample data or pseudo data; The restorer is used to convert the latent representation of the labeled sample data or the latent representation of the pseudo data into data with the same dimension as the labeled sample data.

[0009] Furthermore, the method of using labeled sample data to train the generative adversarial network to obtain a trained generator and restorer includes: Using labeled sample data as the input of the embedder, the embedder and restorer are trained and the training objective is the first optimization objective:

[0010] Where: and are the network parameters corresponding to the embedder and restorer respectively; To balance the reconstruction loss function And the supervised loss function Hyperparameters of Using the noise data as the input of the generator, the generator and discriminator are trained and the training objective is the second optimization objective:

[0011] Where: , are the network parameters corresponding to the generator and the discriminator respectively, is used to balance the unsupervised loss function And the supervised loss function Hyperparameters of When the iterative training reaches the preset first number of iterations, the trained generator and restorer are obtained.

[0012] Furthermore, the loss functions used in training the generative adversarial network include:

[0013]

[0014]

[0015] Where: The gas temperature and flow rate are defined as static characteristics, and Represent the vector of static features and the vector of temporal features respectively; and Represent the instantiated random vectors of static features and temporal features respectively, , ;by and Represent specific values ​​of static and temporal features respectively; is the reconstructed static feature, For reconstruction Moment time characteristics; Representing tuples From the distribution The expected sampling is Representing tuples From synthetic data distribution The expected sampling is represents the sequence length, Indicates the time, represents random noise; Represents the output of the generator for time features; is the true classification of static features, Classify and synthesize static features; for The true classification of moment-to-moment temporal features, for Classification and synthesis of moment-to-moment temporal features; is the true embedding of static features, for True embedding of moment-to-moment temporal features; is the two-norm; , , They represent the reconstruction loss function, unsupervised loss function and supervised loss function respectively.

[0016] Furthermore, the generator, discriminator, embedder and restorer all adopt a 3-layer gated recurrent network.

[0017] Furthermore, the CO 2 The concentration detection model includes a multi-scale enhanced feature extraction module, a prediction module and an auto-encoding module, wherein the input of the multi-scale enhanced feature extraction module is pseudo data and labeled sample data, and the output of the multi-scale enhanced feature extraction module is connected to the prediction module and the auto-encoding module respectively; The multi-scale enhanced feature extraction module is used to obtain multi-scale enhanced features corresponding to the input data based on the input data; The prediction module is used to predict CO based on multi-scale enhanced features. 2 concentration; The self-encoding module is used to obtain reconstructed data with the same dimension as the labeled sample data based on multi-scale enhanced features.

[0018] Furthermore, the multi-scale enhanced feature extraction module includes a multi-scale hybrid enhanced submodule and a feature extraction submodule, and the multi-scale hybrid enhanced submodule includes the same number of average pooling layers and multi-layer perceptrons, each average pooling layer is connected in sequence, each multi-layer perceptron is connected in sequence, and the first The output of the average pooling layer is The outputs of the multi-layer perceptrons are added together as the The mixed data obtained by adding the time features of the input data of the first average pooling layer and the output features of the first multi-layer perceptron is used as the input of the feature extraction submodule; The feature extraction submodule includes a Fourier transform layer, a first convolution layer, a batch normalization layer BN, an activation function ReLU, an inverse Fourier transform layer and a second convolution layer connected in sequence; The mixed data is used as the input of the Fourier transform layer, and the output of the Fourier transform layer is sequentially connected to the inverse Fourier transform layer through the first convolutional layer, the batch normalization layer BN, and the activation function ReLU. The mixed data and the output features of the inverse Fourier transform layer are added as the input of the second convolutional layer, and the output of the second convolutional layer is a multi-scale enhanced feature.

[0019] Furthermore, the use of pseudo data and labeled sample data to analyze CO 2 The concentration detection model is trained, including: The multi-scale enhanced feature extraction module and the prediction module are trained using labeled sample data, and the training objective is the third optimization objective:

[0020] Where: , are the network parameters corresponding to the multi-scale enhanced feature extraction module and the prediction module respectively; Indicates The true labels of labeled sample data, , For use in CO 2 The total amount of labeled sample data for concentration detection model training; Indicates that the prediction module is Output results of labeled sample data; is the two-norm; The self-encoding module is trained using pseudo data and the training objective is the fourth optimization objective:

[0021] Where: are the network parameters corresponding to the autoencoder module; Indicates Pseudo data; Represents the self-encoding module for The output result of the pseudo data is: , For use in CO 2 The total amount of pseudo data used to train the concentration detection model; When the iterative training reaches the preset second iteration number, the trained CO 2 Concentration detection model.

[0022] In addition, the present invention also proposes a CO 2 Concentration detection model training device, comprising: Data collection unit, used to collect different CO 2 The concentration information and the corresponding feature information are used as labeled sample data; A first training unit is used to train the generative adversarial network using labeled sample data to obtain a trained generator and restorer; A pseudo data generating unit, used for generating a potential representation of pseudo data according to random noise using a generator, and then converting the potential representation of the pseudo data into pseudo data with the same dimension as the labeled sample data through a restorer; The second training unit is used to train CO using pseudo data and labeled sample data. 2 The concentration detection model is trained.

[0023] In addition, the present invention also provides a computer-readable storage medium on which a computer program is stored. When the computer program is executed by a processor, the CO 2 Concentration detection model training method.

[0024] In addition, the present invention also proposes a thermal power carbon emission monitoring system, including a laser, an absorption cell, a photoelectric detector, a signal processor, an environmental detector and a CO 2 The concentration detector is arranged on the output laser light path of the laser and a gas path is provided on the absorption cell for the gas to be measured to enter the absorption cell. An environmental detector is arranged at the air inlet of the gas path. The output of the photoelectric detector is connected to the CO 2 The output of the concentration detector and the environmental detector is connected to the CO 2 Concentration detector; CO 2 The concentration detector is deployed in the CO 2 The CO obtained by training the concentration detection model 2 Concentration detection model.

[0025] Furthermore, the environmental detector is used to detect the temperature and flow rate of the gas to be measured.

[0026] Furthermore, the input end of the laser is connected to a wavelength controller, and the wavelength controller is used to control the laser to output laser light of a set wavelength.

[0027] The advantages of the present invention are: (1) Since the number of effective labeled samples in actual model training, such as in the thermal power carbon emission scenario, is relatively small, the present invention trains the generative adversarial network based on a small amount of labeled sample data. Since the labeled sample data is sequence data, it has a temporal correlation before and after. Therefore, the pseudo data generated by the trained generator and restorer is consistent with its temporal correlation and is closer to the real data distribution. Then, the pseudo data and the existing small amount of labeled sample data are used to form augmented data, and the augmented data is used to train the CO 2 The concentration detection model is semi-supervised learning; the CO concentration detection model based on the generative adversarial network and semi-supervised learning is 2 The concentration detection model training method can realize the learning and establishment of the detection model with as little data as possible, reduce the data collection cost, and improve the CO 2 The generalization ability of the concentration detection model, and the CO 2 The detection results of the concentration detection model are more accurate and can adapt to a variety of detection conditions.

[0028] (2) The present invention uses prior information to design the corresponding generative adversarial network architecture and the loss function used in generative adversarial network training, so that the trained generator and restorer can generate pseudo data with the same dimension and similar distribution as the real data, which is closer to the real data distribution, thereby improving CO 2 Accuracy of concentration detection model training.

[0029] (3) The CO designed by the present invention 2 The multi-scale enhanced feature extraction module is used in the concentration detection model to enhance the input features. Since the sequence data contains coarse-grained patterns and fine-grained patterns, they complement each other and reflect the complete information in the sequence. Therefore, the multi-scale enhanced feature extraction module obtains different granularity patterns through average pooling, and then mixes them and obtains enhanced features through time-frequency feature fusion for more detailed analysis, which is helpful for CO 2 Concentration detection model training.

[0030] Additional aspects and advantages of the present invention will be given in part in the following description and in part will be obvious from the following description, or will be learned through practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] Figure 1 A CO is proposed in one embodiment of the present invention. 2Schematic diagram of the flow of the concentration detection model training method; Figure 2 is a schematic diagram of the structure of a generative adversarial network in one embodiment of the present invention; Figure 3 In one embodiment of the present invention, CO 2 Schematic diagram of the structure of the concentration detection model; Figure 4 is a schematic diagram of the structure of a multi-scale enhanced feature extraction module in one embodiment of the present invention; Figure 5 A CO is proposed in one embodiment of the present invention. 2 A schematic diagram of the structure of a concentration detection model training device; Figure 6 It is a structural schematic diagram of a thermal power carbon emission monitoring system proposed in one embodiment of the present invention. DETAILED DESCRIPTION

[0032] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in combination with the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0033] like Figure 1 As shown, the first embodiment of the present invention proposes a CO 2 A concentration detection model training method, the method comprising the following steps: S10, collect different concentrations of CO 2 And the corresponding feature information as labeled sample data; Specifically, this embodiment can collect CO with different concentrations under different concentrations of standard gas through standard gas detection. 2 Corresponding characteristic information, the corresponding standard gas concentration is used as a label and the corresponding characteristic information to form a labeled sample, and the data set is composed of each labeled sample data, that is, the real sample data .

[0034] It should be noted that in order to conduct carbon emission detection of thermal power companies, for example, standard gas of known concentration can be detected to obtain corresponding characteristic information, thereby obtaining labeled sample data.

[0035] S20, using labeled sample data to train the generative adversarial network to obtain a trained generator and restorer; S30, using a generator to generate a potential representation of pseudo data according to random noise, and then converting the potential representation of the pseudo data into pseudo data with the same dimension as the labeled sample data through a restorer; It should be noted that considering that the number of effective labeled samples in the thermal power carbon emission scenario is relatively small, which leads to poor generalization ability of the trained model, this embodiment uses a generative adversarial network to expand the number of sample data to achieve the enhancement of labeled sample data, and in the training process of the generative adversarial network, the labeled sample data is used to train the generative adversarial network. Since the labeled sample data is sequence data, it has a time correlation before and after, and provides time series information when generating pseudo data for noise data, so that the pseudo data generated by the trained generator and restorer have the same dimension and similar distribution as the real sample data, which is closer to the real data distribution, thereby improving the CO2 emission in the thermal power carbon emission scenario. 2 Accuracy of concentration detection model training.

[0036] It should be understood that those skilled in the art may also refer to the data acquisition method proposed in this method to implement data enhancement for model training when detecting other harmful substances contained in industrial waste gas, wastewater, solid waste, etc.

[0037] S40, using pseudo data and labeled sample data to analyze CO 2 The concentration detection model is trained.

[0038] It should be noted that in this embodiment, pseudo data is used as unlabeled sample data to form an unlabeled data set. , in the dataset 80% of the data are randomly selected to form a labeled data set , dataset The remaining 20% ​​of the data forms the test set ; The unlabeled dataset With labeled dataset Together we can 2 Concentration detection model training, after training, on the test set On CO 2 The concentration detection model was tested and verified.

[0039] This embodiment introduces generative adversarial learning and semi-supervised learning, and uses as little real detection data as possible to learn CO 2 The relationship between gas concentration and the laser signal change caused by it is established 2 Concentration detection model: A generative adversarial method is used to synthesize pseudo data for training the concentration detection model, which increases the generalization ability of the model and improves the accuracy of the detection results when the sample gas concentration value is small and the detection data is small.

[0040] As a further preferred technical solution, step S10: collecting CO at different concentrations 2 and the corresponding feature information as labeled sample data, specifically including the following steps: Under different concentrations of standard gas, collect different concentrations of CO 2 And the corresponding feature information as labeled sample data; The characteristic information includes laser signal, gas temperature and gas flow rate.

[0041] It should be noted that in this embodiment, different concentrations of CO 2 The corresponding laser signal, gas temperature and gas flow data are used as features , the corresponding standard gas concentration is used as the label , forming labeled sample data , and by Labeled sample data Composition of the dataset .

[0042] Since in practical applications, CO 2 The concentration of CO is also related to environmental factors. In addition to selecting the laser signal, this embodiment also selects gas temperature and flow rate as factors related to CO 2 The concentration-related feature information is used to construct labeled sample data, and the labeled sample data is used for data enhancement, thereby further ensuring that the synthesized pseudo data is closer to the real data distribution.

[0043] As a further preferred technical solution, Figure 2 As shown, the generative adversarial network includes a generator, a discriminator, an embedder and a restorer. The input of the generator is noise data, and the output is connected to the discriminator. The input of the embedder is labeled sample data, and the output is connected to the discriminator and the restorer respectively. The generator is used to generate latent representations of pseudo data based on noisy data; The embedder is used to generate a latent representation of the labeled sample data based on the labeled sample data; The discriminator is used to determine whether the output of the generator and the output of the embedder are labeled sample data or pseudo data; The restorer is used to convert the latent representation of the labeled sample data or the latent representation of the pseudo data into data with the same dimension as the labeled sample data.

[0044] Specifically, since the input labeled sample data is sequence data, there is a temporal correlation between the previous and the next data. In order to generate pseudo data that conforms to the temporal correlation, this embodiment uses prior information to design a corresponding generative adversarial network architecture and loss function. The generative adversarial network architecture includes a generator , Discriminator , Embedder and restorer , the corresponding generator , Discriminator , Embedder and restorer The network parameters are used one by one , , and It means that the generator , Discriminator , Embedder and restorer The network uses a 3-layer gated recurrent network, where: Generator The input is random noise and the output is the potential representation of pseudo data; Discriminator For the generator Embedder The output of the generator is used to determine whether it is real data or fake data. The generator and the discriminator achieve the effect of confrontation through alternating training, continuously improving their respective generation and identification capabilities, and finally making the confrontation between the generator and the discriminator reach a Nash equilibrium; Embedder The input is the real labeled sample data, and the output is the potential representation of the labeled sample data; Restorer The input is the potential representation of the real labeled sample data, and the output is the reconstructed data with the same dimension as the labeled sample data.

[0045] After the training of the generative adversarial network is completed, the generator is used to generate a latent representation of the pseudo data based on random noise, and then the restorer is used to convert the latent representation of the pseudo data into pseudo data with the same dimension and similar distribution as the labeled sample data.

[0046] As a further preferred technical solution, the step S20: using labeled sample data to train the generative adversarial network to obtain a trained generator and restorer specifically includes the following steps: S21. Unlabeled dataset using pseudo data With labeled dataset Build a training dataset; S22. Use the labeled sample data as the input of the embedder, train the embedder and the restorer, and the training target is the first optimization target:

[0047] Where: and are the network parameters corresponding to the embedder and restorer respectively; To balance the reconstruction loss function And the supervised loss function Hyperparameters of S23. Use the noise data as the input of the generator, train the generator and the discriminator, and the training target is the second optimization target:

[0048] Where: , are the network parameters corresponding to the generator and the discriminator respectively, is used to balance the unsupervised loss function And the supervised loss function Hyperparameters of S24, repeat the above steps S22 to S23, iteratively train the generative adversarial network until the preset first iteration number is reached, and obtain the trained generator , Discriminator , Embedder and restorer .

[0049] It should be noted that after the training is completed, the trained generator is used and restorer Get synthetic data, i.e. pseudo data, to complete the data set The specific process of obtaining synthetic data is as follows: random noise is input into the trained generator , get its potential representation, and then enter the trained restorer Reconstruct synthetic data with the same dimension and similar distribution as the real data based on the latent representation.

[0050] As a further preferred technical solution, the loss function used when training the generative adversarial network includes:

[0051]

[0052]

[0053] Where: The gas temperature and flow rate are defined as static characteristics, and Represent the vector of static features and the vector of temporal features respectively; and Represent the instantiated random vectors of static features and temporal features respectively, , ;by and Represent specific values ​​of static and temporal features respectively; is the reconstructed static feature, For reconstruction Moment time characteristics; Representing tuples From the distribution The expected sampling is Representing tuples From synthetic data distribution The expected sampling is represents the sequence length, Indicates the time, represents random noise; Represents the output of the generator for time features; is the true classification of static features, Classify and synthesize static features; for The true classification of moment-to-moment temporal features, for Classification and synthesis of moment-to-moment temporal features; is the true embedding of static features, for True embedding of moment-to-moment temporal features; is the two-norm; , , They represent the reconstruction loss function, unsupervised loss function and supervised loss function respectively.

[0054] It should be noted that, since the carbon concentration in the flue gas emitted by thermal power plants is related to the flue gas flow rate and temperature, this embodiment selects the temperature and flow rate of the gas as static features, which are used together to train the concentration detection model to improve the accuracy of the model training results.

[0055] In the process of constructing the generative adversarial network, this embodiment fully considers the static features and time features existing in the data, as well as the time correlation before and after, and calculates the loss of the static features and time features in the loss function, so that the data synthesized by the model is more consistent with the characteristics of real data; the advantages of each loss function set in this embodiment are: (1) It promotes the generator to generate samples that are as close to the real data distribution as possible. (2) The discriminator improves its ability to distinguish between real data and fake data by optimizing the loss function. (3) Optimizing the loss function ensures that there is a balance between the generator and the discriminator, avoiding over-optimization of one of them and causing unstable training.

[0056] As a further preferred technical solution, the CO 2The concentration detection model includes a multi-scale enhanced feature extraction module, a prediction module and an auto-encoding module, wherein the input of the multi-scale enhanced feature extraction module is pseudo data and labeled sample data, and the output of the multi-scale enhanced feature extraction module is connected to the prediction module and the auto-encoding module respectively; The multi-scale enhanced feature extraction module is used to obtain multi-scale enhanced features corresponding to the input data based on the input data; The prediction module is used to predict CO based on multi-scale enhanced features. 2 concentration; The self-encoding module is used to obtain reconstructed data with the same dimension as the labeled sample data based on multi-scale enhanced features.

[0057] Specifically, Figure 3 As shown, the CO 2 The concentration detection model includes a multi-scale enhanced feature extraction module , prediction module and the autoencoder module , the network parameters corresponding to each module are respectively , and Indicates; among them, The structure includes a multi-scale hybrid enhancement submodule and a feature extraction submodule; and The structure is a fully connected layer. What needs to be explained here is that Figure 3 Indicates CO 2 The process and structure of the concentration detection model training. During training, the multi-scale enhanced feature extraction module is used in two iterative steps respectively. It can be considered that CO 2 The concentration detection model contains two multi-scale enhanced feature extraction modules with the same structure, but the parameters of the two modules are different.

[0058] The multi-scale hybrid enhancement submodule is used to perform multi-scale hybrid enhancement on the input data to obtain hybrid data with the same dimension as the input data; specifically, the multi-scale hybrid enhancement submodule includes the same number of average pooling layers and multi-layer perceptrons, each average pooling layer is connected in sequence, each multi-layer perceptron is connected in sequence, and the first The output of the average pooling layer is The outputs of the multi-layer perceptrons are added together as the The mixed data obtained by adding the time features of the input data of the first average pooling layer and the output features of the first multi-layer perceptron is used as the input of the feature extraction submodule.

[0059] Specifically, the multi-scale hybrid enhancement submodule performs multi-scale hybrid enhancement on the input data to obtain hybrid data with the same dimension as the input data. Assume that the temporal characteristics of the input data of the multi-scale hybrid enhancement submodule are , the specific calculation process is as follows:

[0060]

[0061] Where: represents the average pooling operation, For the The output of the sub-average pooling, represents a multi-layer perceptron, , Output intermediate variables for features, That is, the obtained mixed data has the same dimension as the input data.

[0062] Furthermore, the feature extraction submodule is used to extract features from the mixed data to obtain multi-scale enhanced features; specifically, the feature extraction submodule includes a Fourier transform layer PPT, a first convolution layer Conv, a batch normalization layer BN, an activation function ReLU, an inverse Fourier transform layer Inv FFT and a second convolution layer Conv connected in sequence; The mixed data is used as the input of the Fourier transform layer PPT, and the output of the Fourier transform layer PPT is sequentially connected to the inverse Fourier transform layer Inv FFT through the first convolution layer Conv, the batch normalization layer BN, and the activation function ReLU. The mixed data and the output features of the inverse Fourier transform layer Inv FFT are added as the input of the second convolution layer Conv, and the output of the second convolution layer Conv is a multi-scale enhanced feature.

[0063] It should be noted that the structure of the multi-scale enhanced feature extraction module is as follows: Figure 4 As shown in the figure, its output has the same dimension as the input, which plays the role of feature enhancement of the input. Since the sequence contains coarse-grained patterns and fine-grained patterns, they complement each other and jointly reflect the complete information in the sequence. Therefore, this module obtains different granularity patterns through average pooling, and then mixes them to obtain enhanced features for more detailed analysis.

[0064] This embodiment designs a concentration detection model structure based on multi-scale enhanced feature extraction, which can capture different granularity features in the data, obtain richer feature information, and enhance the robustness of the model. It can adapt to various detection conditions with small computational complexity, thereby achieving more efficient, accurate and low-cost online real-time monitoring of carbon emissions.

[0065] As a further preferred technical solution, the step S40: using pseudo data and labeled sample data to analyze CO 2 The concentration detection model is trained, which includes the following steps: S41, using labeled sample data to train the multi-scale enhanced feature extraction module and the prediction module, and the training target is the third optimization target:

[0066] Where: , are the network parameters corresponding to the multi-scale enhanced feature extraction module and the prediction module respectively; Indicates The true labels of labeled sample data, , For use in CO 2 The total amount of labeled sample data for concentration detection model training; Indicates that the prediction module is Output results of labeled sample data; is the two-norm; S42, using pseudo data to train the autoencoder module and the training target is the fourth optimization target:

[0067] Where: are the network parameters corresponding to the autoencoder module; Indicates Pseudo data; Represents the self-encoding module for The output result of the pseudo data is: , For use in CO 2 The total amount of pseudo data used to train the concentration detection model; S43, repeat the above steps S41-S42 to perform iterative training until the preset second iteration number is reached to obtain the trained CO 2 Concentration detection model.

[0068] It should be noted that after the iterative training is completed, the test set is used For the CO 2 The concentration detection model is tested for performance; if the performance meets the set requirements, the CO 2 Multi-scale enhanced feature extraction module in concentration detection model and prediction module It can be used for CO 2 Concentration detection; if the performance does not meet the set requirements, adjust the network hyperparameters and re-run CO2 Concentration detection model training.

[0069] In addition, if Figure 5 As shown, the second embodiment of the present invention proposes a CO 2 A concentration detection model training device, the device comprising: The data collection unit 10 is used to collect different CO 2 The concentration information and the corresponding feature information are used as labeled sample data; A first training unit 20 is used to train the generative adversarial network using labeled sample data to obtain a trained generator and restorer; A pseudo data generating unit 30, configured to generate a potential representation of pseudo data according to random noise using a generator, and then convert the potential representation of the pseudo data into pseudo data having the same dimension as the labeled sample data through a restorer; The second training unit 40 is used to train CO using pseudo data and labeled sample data. 2 The concentration detection model is trained.

[0070] As a further preferred technical solution, the data collection unit 10 is specifically used for: Under different concentrations of standard gas, collect different concentrations of CO 2 And the corresponding feature information as labeled sample data; The characteristic information includes laser signal, gas temperature and gas flow rate.

[0071] As a further preferred technical solution, the first training unit 20 includes: The first training subunit is used to take the labeled sample data as the input of the embedder, train the embedder and the restorer, and the training target is the first optimization target:

[0072] Where: and are the network parameters corresponding to the embedder and restorer respectively; To balance the reconstruction loss function And the supervised loss function Hyperparameters of The second training subunit is used to take the noise data as the input of the generator, train the generator and the discriminator, and the training objective is the second optimization objective:

[0073] Where: , are the network parameters corresponding to the generator and the discriminator respectively, is used to balance the unsupervised loss function And the supervised loss function Hyperparameters of The first iterative training subunit is used to perform iterative training until a preset first number of iterations is reached to obtain a trained generator and restorer.

[0074] As a further preferred technical solution, the second training unit 40 includes: The third training subunit is used to train the multi-scale enhanced feature extraction module and the prediction module with labeled sample data, and the training target is a third optimization target:

[0075] Where: , are the network parameters corresponding to the multi-scale enhanced feature extraction module and the prediction module respectively; Indicates The true labels of labeled sample data, , For use in CO 2 The total amount of labeled sample data for concentration detection model training; Indicates that the prediction module is Output results of labeled sample data; is the two-norm; The fourth training subunit is used to train the autoencoder module using pseudo data and the training target is a fourth optimization target:

[0076] Where: are the network parameters corresponding to the autoencoder module; Indicates Pseudo data; Represents the self-encoding module for The output result of the pseudo data is: , For use in CO 2 The total amount of pseudo data used to train the concentration detection model; The second iterative training subunit is used to perform iterative training until the preset second iteration number is reached to obtain the trained CO 2 Concentration detection model.

[0077] It should be noted that the CO 2 Other embodiments or specific implementation methods of the concentration detection model training device can refer to the above-mentioned method embodiments and will not be described in detail here.

[0078] In addition, if Figure 6 As shown, the third embodiment of the present invention also proposes a thermal power carbon emission monitoring system, including a laser, an absorption cell, a photodetector, a signal processor, an environmental detector and a CO 2 The concentration detector is arranged on the output laser light path of the laser and a gas path is provided on the absorption cell for the gas to be measured to enter the absorption cell. An environmental detector is arranged at the air inlet of the gas path. The output of the photoelectric detector is connected to the CO 2 The output of the concentration detector and the environmental detector is connected to the CO 2 Concentration detector; CO 2 The CO concentration detector is configured by the first embodiment described above. 2 The CO concentration detection model training method is used to monitor carbon emissions from thermal power plants. 2 Concentration detection model.

[0079] As a further preferred technical solution, the environmental detector is used to detect the temperature and flow rate of the gas to be measured.

[0080] It should be noted that, since changes in the flue gas flow and temperature of carbon emissions from thermal power plants may cause changes in the CO2 concentration detection results, this embodiment arranges, for example, temperature sensors and flow sensors in the flue gas path to detect the temperature and flow of the flue gas.

[0081] As a further preferred technical solution, the input end of the laser is connected to a wavelength controller, and the wavelength controller is used to control the laser to output laser light of a set wavelength.

[0082] Specifically, the laser used in this embodiment is a tunable semiconductor laser that can emit infrared light of a specific wavelength, which can be absorbed by carbon dioxide molecules. The gas to be measured enters the absorption cell from the gas path, and the laser emitted by the laser passes through the absorption cell, where the carbon dioxide molecules absorb light of a specific wavelength, causing the laser light intensity to attenuate. The photodetector is used to measure the change in light intensity after passing through the absorption cell. Carbon dioxide molecules will absorb light energy of a specific wavelength, resulting in a decrease in the light intensity passing through the detector. The photodetector can convert these changing light signals into electrical signals. The signal processor processes the signal output by the photodetector, such as filtering, amplifying, and converting to extract and analyze information related to carbon dioxide absorption. The wavelength controller accurately controls the output wavelength of the laser by adjusting the current and temperature of the laser, so that it is locked near the specific absorption line of carbon dioxide to improve the detection sensitivity. The environmental detector obtains the gas flow and temperature of the measured gas sample through detection; CO 2 Set up CO concentration detector 2 The concentration detection model is used to detect the concentration of the gas sample according to the laser signal of the gas sample to be measured and the gas flow rate and temperature of the gas sample to be measured.2 Concentration detection model for CO 2 Concentration detection to achieve online real-time monitoring of carbon emissions; the CO 2 The concentration detection model is based on the generative adversarial network and semi-supervised learning proposed in the first embodiment, and is trained using real data and augmented data.

[0083] In actual use, the gas to be measured is introduced into the online real-time monitoring system for carbon emissions designed in this embodiment through the air inlet, and the spectral signal of the gas is processed and then enters the CO 2 Concentration detection module, in which CO 2 The concentration detection model will output its corresponding concentration, specifically, in CO 2 In the concentration detection model, the signal is input into the multi-scale enhanced feature extraction module to obtain the feature representation, and then input into the prediction module to obtain the corresponding concentration value according to its feature representation.

[0084] As a further preferred technical solution, this embodiment can also use online data to 2 The network parameters of the concentration detection module are tuned to 2 The concentration detection module is updated online.

[0085] In addition, the fourth embodiment of the present invention further provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the CO described in the first embodiment above is realized. 2 Concentration detection model training method.

[0086] It should be noted that the logic and / or steps represented in the flowchart or described in other ways herein, for example, can be considered as a sequenced list of executable instructions for implementing logical functions, and can be specifically implemented in any computer-readable medium for use by an instruction execution system, device or apparatus (such as a computer-based system, a system including a processor, or other system that can fetch instructions from an instruction execution system, device or apparatus and execute instructions), or in combination with these instruction execution systems, devices or apparatuses. For the purpose of this specification, "computer-readable medium" can be any device that can contain, store, communicate, propagate or transmit a program for use by an instruction execution system, device or apparatus, or in combination with these instruction execution systems, devices or apparatuses. More specific examples (non-exhaustive list) of computer-readable media include the following: an electrical connection portion with one or more wirings (electronic device), a portable computer disk box (magnetic device), a random access memory (RAM), a read-only memory (ROM), an erasable and programmable read-only memory (EPROM or flash memory), an optical fiber device, and a portable compact disk read-only memory (CDROM). In addition, the computer-readable medium may even be paper or other suitable medium on which the program is printed, since the program may be obtained electronically, for example, by optically scanning the paper or other medium and then editing, interpreting or processing in other suitable ways if necessary, and then stored in a computer memory.

[0087] It should be understood that the various parts of the present invention can be implemented by hardware, software, firmware or a combination thereof. In the above-mentioned embodiments, multiple steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented by hardware, as in another embodiment, it can be implemented by any one of the following technologies known in the art or their combination: a discrete logic circuit having a logic gate circuit for implementing a logic function for a data signal, a dedicated integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.

[0088] In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "examples", "specific examples", or "some examples" means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representation of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner.

[0089] In addition, the terms "first" and "second" are used for descriptive purposes only and should not be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Therefore, the features defined as "first" and "second" may explicitly or implicitly include at least one of the features. In the description of the present invention, the meaning of "plurality" is at least two, such as two, three, etc., unless otherwise clearly and specifically defined.

[0090] Although the embodiments of the present invention have been shown and described above, it is to be understood that the above embodiments are exemplary and are not to be construed as limitations of the present invention. A person skilled in the art may change, modify, replace and vary the above embodiments within the scope of the present invention.

Claims

1. A CO2 concentration detection model training method, characterized in that: include: Collect different concentrations of CO2 and corresponding feature information as labeled sample data; Use labeled sample data to train the generative adversarial network to obtain trained generators and restorers; The generator generates the latent representation of the pseudo data according to random noise, and then the restorer converts the latent representation of the pseudo data into pseudo data with the same dimension as the labeled sample data. The CO2 concentration detection model is trained using pseudo data and labeled sample data.

2. The CO2 concentration detection model training method according to claim 1, characterized in that: The collecting of different concentrations of CO2 and corresponding characteristic information as labeled sample data includes: Under different concentrations of standard gas, different concentrations of CO2 and corresponding characteristic information are collected as labeled sample data; The characteristic information includes laser signal, gas temperature and gas flow rate.

3. The CO2 concentration detection model training method according to claim 1, characterized in that: The generative adversarial network includes a generator, a discriminator, an embedder and a restorer. The input of the generator is noise data, and the output is connected to the discriminator. The input of the embedder is labeled sample data, and the output is connected to the discriminator and the restorer respectively. The generator is used to generate latent representations of pseudo data based on noisy data; The embedder is used to generate a latent representation of the labeled sample data based on the labeled sample data; The discriminator is used to determine whether the output of the generator and the output of the embedder are labeled sample data or pseudo data; The restorer is used to convert the latent representation of the labeled sample data or the latent representation of the pseudo data into data with the same dimension as the labeled sample data.

4. The CO2 concentration detection model training method according to claim 2, characterized in that: The method of using labeled sample data to train a generative adversarial network to obtain a trained generator and restorer includes: Using labeled sample data as the input of the embedder, the embedder and restorer are trained and the training objective is the first optimization objective: Where: and are the network parameters corresponding to the embedder and restorer respectively; To balance the reconstruction loss function And the supervised loss function Hyperparameters of Using the noise data as the input of the generator, the generator and discriminator are trained and the training objective is the second optimization objective: Where: , are the network parameters corresponding to the generator and the discriminator respectively, is used to balance the unsupervised loss function And the supervised loss function Hyperparameters of When the iterative training reaches the preset first number of iterations, the trained generator and restorer are obtained.

5. The CO2 concentration detection model training method according to any one of claims 1 to 4, characterized in that: The loss functions used when training the generative adversarial network include: Where: The gas temperature and flow rate are defined as static characteristics, and Represent the vector of static features and the vector of temporal features respectively; and Represent the instantiated random vectors of static features and temporal features respectively, , ;by and Represent specific values ​​of static and temporal features respectively; is the reconstructed static feature, For reconstruction Moment time characteristics Representing tuples From the distribution The expected sampling is Representing tuples From synthetic data distribution The expected sampling is represents the sequence length, Indicates the time, represents random noise; Represents the output of the generator for time features; is the true classification of static features, Classify and synthesize static features; for The true classification of moment-to-moment temporal features, for Classification and synthesis of moment-to-moment temporal features; is the true embedding of static features, for True embedding of moment-to-moment temporal features; is the two-norm; , , They represent the reconstruction loss function, unsupervised loss function and supervised loss function respectively.

6. The CO2 concentration detection model training method according to claim 3, characterized in that: The generator, discriminator, embedder and restorer all adopt a 3-layer gated recurrent network.

7. The CO2 concentration detection model training method according to claim 1, characterized in that: The CO2 concentration detection model includes a multi-scale enhanced feature extraction module, a prediction module and an autoencoder module, the input of the multi-scale enhanced feature extraction module is pseudo data and labeled sample data, and the output of the multi-scale enhanced feature extraction module is connected to the prediction module and the autoencoder module respectively; The multi-scale enhanced feature extraction module is used to obtain multi-scale enhanced features corresponding to the input data based on the input data; The prediction module is used to predict CO2 concentration based on multi-scale enhanced features; The self-encoding module is used to obtain reconstructed data with the same dimension as the labeled sample data based on the multi-scale enhanced features.

8. The CO2 concentration detection model training method according to claim 7, characterized in that: The multi-scale enhancement feature extraction module includes a multi-scale hybrid enhancement submodule and a feature extraction submodule connected in sequence; the multi-scale hybrid enhancement submodule includes an equal number of average pooling layers and multi-layer perceptrons, each average pooling layer is connected in sequence, each multi-layer perceptron is connected in sequence, and the first The output of the average pooling layer is The outputs of the multi-layer perceptrons are added together as the The mixed data obtained by adding the time features of the input data of the first average pooling layer and the output features of the first multi-layer perceptron is used as the input of the feature extraction submodule; the feature extraction submodule includes a Fourier transform layer, a first convolution layer, a batch normalization layer BN, an activation function ReLU, an inverse Fourier transform layer and a second convolution layer connected in sequence; The multi-scale mixed enhancement submodule is used to perform multi-scale mixed enhancement on the input data to obtain mixed data with the same dimension as the input data; The feature extraction submodule is used to extract features from mixed data to obtain multi-scale enhanced features.

9. The CO2 concentration detection model training method according to claim 8, characterized in that: It includes the same number of average pooling layers and multi-layer perceptrons, each average pooling layer is connected in sequence, and the first The output of the average pooling layer is The outputs of the multi-layer perceptrons are concatenated as the The multi-layer perceptron input, the multi-scale hybrid enhancement submodule is used to perform the following calculation on a certain input data: Where: represents the average pooling operation, For the The output of the sub-average pooling, represents a multi-layer perceptron, , , Output intermediate variables for features, It is the mixed data output by the multi-scale mixed enhancement submodule.

10. The CO2 concentration detection model training method according to claim 7, characterized in that: The method of training the CO2 concentration detection model using pseudo data and labeled sample data includes: The multi-scale enhanced feature extraction module and the prediction module are trained using labeled sample data, and the training objective is the third optimization objective: Where: , are the network parameters corresponding to the multi-scale enhanced feature extraction module and the prediction module respectively; Indicates The true labels of labeled sample data, , is the total amount of labeled sample data used to train the CO2 concentration detection model; Indicates that the prediction module is Output results of labeled sample data; is the two-norm; The self-encoding module is trained using pseudo data and the training objective is the fourth optimization objective: Where: are the network parameters corresponding to the autoencoder module; Indicates Pseudo data; Represents the self-encoding module for The output result of the pseudo data is: , is the total amount of pseudo data used to train the CO2 concentration detection model; When the iterative training reaches the preset second iteration number, a trained CO2 concentration detection model is obtained.

11. A CO2 concentration detection model training device, characterized in that: include: A data collection unit, used to collect different CO2 concentration information and corresponding feature information as labeled sample data; A first training unit is used to train the generative adversarial network using labeled sample data to obtain a trained generator and restorer; A pseudo data generating unit, used for generating a potential representation of pseudo data according to random noise using a generator, and then converting the potential representation of the pseudo data into pseudo data with the same dimension as the labeled sample data through a restorer; The second training unit is used to train the CO2 concentration detection model using pseudo data and labeled sample data.

12. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the method according to any one of claims 1 to 10 is implemented.

13. A thermal power carbon emission monitoring system, characterized in that: It includes a laser, an absorption cell, a photoelectric detector, a signal processor, an environmental detector and a CO2 concentration detector. The absorption cell is arranged on the output laser light path of the laser and a gas path is provided on the absorption cell for the gas to be measured to enter the absorption cell. The environmental detector is arranged at the air inlet of the gas path. The output of the photoelectric detector is connected to the CO2 concentration detector via the signal processor, and the output of the environmental detector is connected to the CO2 concentration detector. The CO2 concentration detector deploys a CO2 concentration detection model trained by the CO2 concentration detection model training method described in any one of claims 1-10.

14. The thermal power carbon emission monitoring system according to claim 13, characterized in that: The environmental detector is used to detect the temperature and flow rate of the gas to be tested.

15. The thermal power carbon emission monitoring system according to claim 13, characterized in that: The input end of the laser is connected to a wavelength controller, and the wavelength controller is used to control the laser to output laser light of a set wavelength.

Images