Thermal power combustion furnace tail gas emission prediction method and system

Abstract

The invention discloses a thermal power combustion furnace tail gas emission prediction method and system. The method comprises the steps of obtaining thermal power combustion furnace tail gas data; performing discrete Fourier transform on the thermal power combustion furnace tail gas data to obtain dominant component data and stationary component data; inputting the stationary component data into a prediction backbone model for prediction to obtain a first prediction value, and inputting the dominant component data into an MLP model for prediction to obtain a second prediction value; and combining the first prediction value with the second prediction value to obtain a thermal power combustion furnace tail gas emission prediction value. According to the method, the problem of non-stationary data processing in thermal power combustion furnace tail gas emission monitoring is effectively solved, precise capture of complex dynamic trends and seasonal modes is realized, and the prediction precision is remarkably improved; due to the model independence and wide applicability of the method, the method can be flexibly applied to various prediction models and data sets, and the universality and real-time response capability of the system are enhanced.

Description

Technical Field

[0001] The present invention relates to the technical field of tail gas emission, and in particular to a method and system for predicting tail gas emission from a thermal power plant combustion furnace. Background Art

[0002] With the increasing awareness of environmental protection, the control and monitoring of tail gas emissions in thermal power plants has become particularly important. The flue gas produced during the combustion process contains a variety of pollutants, such as sulfur dioxide (SO 2 ), carbon monoxide (CO), and dust. These pollutants are not only harmful to the environment, but also affect human health. Therefore, accurately measuring and predicting flue gas emissions is an important step in achieving environmental protection and compliance.

[0003] In thermal power plants, the monitoring of flue gas emissions mainly involves the following variables: raw flue gas SO 2 Concentration, net flue gas SO 2 Concentration, net flue gas dust concentration, CO concentration, CO content of flue gas on side A and CO content of flue gas on side B, etc. These variables reflect the flue gas composition at different stages. Accurate measurement and analysis are essential for optimizing the combustion process and reducing emissions. Traditional time series prediction methods face the challenge of non-stationary data when dealing with these variables. Flue gas emission data often have dynamic trends and seasonal patterns, which makes prediction complicated. In order to solve these problems, reversible instance normalization methods have been proposed in recent years to mitigate the impact of trends through statistical means such as mean and variance. However, these methods mainly target basic trends and have limited processing capabilities for complex seasonal patterns. Summary of the invention

[0004] The purpose of this section is to summarize some aspects of embodiments of the present invention and briefly introduce some preferred embodiments. Some simplifications or omissions may be made in this section and the specification abstract and the invention title of this application to avoid blurring the purpose of this section, the specification abstract and the invention title, and such simplifications or omissions cannot be used to limit the scope of the present invention.

[0005] In view of the above existing problems, the present invention is proposed. Therefore, the present invention provides a method for predicting tail gas emissions from thermal power plant combustion furnaces to solve the problems of complex non-stationary data processing, insufficient trend and seasonal separation, poor model adaptability, and insufficient real-time and accuracy in monitoring tail gas emissions from thermal power plant combustion furnaces.

[0006] In order to solve the above technical problems, the present invention provides the following technical solutions:

[0007] In a first aspect, the present invention provides a method for predicting tail gas emissions from a thermal power plant combustion furnace, comprising:

[0008] Obtain exhaust gas data from thermal power plant combustion furnaces;

[0009] Performing discrete Fourier transform on the tail gas data of the thermal power plant combustion furnace to obtain dominant component data and stable component data;

[0010] Input the stationary component data into the prediction backbone model for prediction to obtain a first prediction value, and input the dominant component data into the MLP model for prediction to obtain a second prediction value;

[0011] The first predicted value and the second predicted value are combined to obtain a predicted value of tail gas emissions from a thermal power plant combustion furnace.

[0012] As a preferred solution of the method for predicting the exhaust gas emission of a thermal power plant combustion furnace described in the present invention, wherein: the exhaust gas data of the thermal power plant combustion furnace is subjected to discrete Fourier transform to obtain the dominant component data, including:

[0013] Convert the tail gas data of the thermal power plant combustion furnace into Fourier components in complex form, and select a frequency component set with the largest amplitude;

[0014] The frequency component set is restored to the time domain component to obtain the dominant component data, which is expressed as:

[0015]

[0016] in, is the dominant component data, K t is the frequency component set, Z t Convert the exhaust gas data of thermal power plant into Fourier components in complex form.

[0017] As a preferred solution of the method for predicting tail gas emissions from thermal power plant combustion furnaces described in the present invention, obtaining stable component data includes:

[0018] The dominant component data is subtracted from the Fourier component to obtain the stationary component data.

[0019] As a preferred solution of the method for predicting tail gas emissions from a thermal power plant combustion furnace according to the present invention, obtaining the first predicted value includes:

[0020] The prediction backbone model receives the stationary component data as input and passes it through multiple fully connected layers;

[0021] In each fully connected layer, the stationary component data is weighted and summed with the neurons in the previous layer using the learned weights, and a bias term is added to extract stationary features;

[0022] The stationary features are weighted and transformed by a nonlinear activation function to obtain enhanced stationary features;

[0023] The enhanced stationary features are transmitted to the output layer, and the output layer obtains the first prediction value by using a linear activation function calculation based on the learned weights and biases.

[0024] As a preferred solution of the method for predicting tail gas emissions from thermal power plant combustion furnaces described in the present invention, obtaining the second predicted value includes:

[0025] The MLP model is used to predict the future value of the dominant component data to obtain a second predicted value, which is expressed as:

[0026]

[0027] Among them, W 1 , W 2 , W 3 are all learnable parameters, X non,t is the dominant component data, X t The exhaust gas data of thermal power plant is converted into Fourier components in complex form, ReLU is the activation function, is the second predicted value.

[0028] As a preferred embodiment of the method for predicting the tail gas emission of a thermal power plant combustion furnace described in the present invention, the backbone prediction model and the MLP model include:

[0029] The prior guidance loss function and the overall loss function are introduced to optimize the backbone prediction model and the MLP model, which are expressed as:

[0030]

[0031] Among them, L nonstat,φ is the overall loss function, L forecast,θ,φ is the prior guidance loss function, is the second predicted value, Y non,t is the offset prediction value, It is the predicted value of tail gas emission from thermal power combustion furnace.

[0032] As a preferred solution of the method for predicting tail gas emissions from thermal power plant combustion furnaces described in the present invention, it also includes:

[0033] Taking the prior guidance loss function and the overall loss function as objective functions;

[0034] Determine the parameter space of the backbone prediction model and the MLP model and initialize the parameters;

[0035] The objective function is modeled using a Bayesian algorithm to establish a probability model, and the parameter combination is evaluated using the probability model to obtain the optimal parameter combination.

[0036] In a second aspect, the present invention provides a system for predicting tail gas emissions from thermal power plant combustion furnaces, comprising:

[0037] A data acquisition module, used to obtain tail gas data of thermal power combustion furnaces;

[0038] A classification module, used for performing discrete Fourier transform on the tail gas data of the thermal power combustion furnace to obtain dominant component data and stable component data;

[0039] A classification prediction module, used for inputting the stationary component data into a prediction backbone model for prediction to obtain a first prediction value, and inputting the dominant component data into an MLP model for prediction to obtain a second prediction value;

[0040] The result prediction module is used to combine the first prediction value and the second prediction value to obtain a prediction value of tail gas emissions from a thermal power plant combustion furnace.

[0041] In a third aspect, the present invention provides a computing device, comprising:

[0042] Memory and processor;

[0043] The memory is used to store computer executable instructions, and the processor is used to execute the computer executable instructions. When the computer executable instructions are executed by the processor, the steps of the method for predicting the exhaust gas emissions of thermal power combustion furnaces are implemented.

[0044] In a fourth aspect, the present invention provides a computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, implement the steps of the method for predicting exhaust gas emissions from a thermal power plant combustion furnace.

[0045] Compared with the prior art, the present invention has the following beneficial effects: the present invention introduces frequency adaptive normalization, which effectively solves the problem of non-stationary data processing in the exhaust emission monitoring of thermal power combustion furnaces, realizes the accurate capture of complex dynamic trends and seasonal patterns, and significantly improves the prediction accuracy; the model independence and wide applicability of the inventive method enable it to be flexibly applied to a variety of prediction models and data sets, enhances the universality and real-time response capability of the system, provides thermal power plants with a more accurate and efficient exhaust emission measurement solution, and strongly supports environmental management and decision-making optimization. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for describing the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative labor. Among them:

[0047] Figure 1 It is a schematic diagram of the overall process of a method for predicting tail gas emissions from a thermal power plant combustion furnace according to an embodiment of the present invention;

[0048] Figure 2 A fast Fourier transform schematic diagram of a method for predicting tail gas emissions from a thermal power plant combustion furnace according to an embodiment of the present invention;

[0049] Figure 3 The figure is a schematic diagram of the overall structure of a method for predicting tail gas emissions from a thermal power plant combustion furnace according to an embodiment of the present invention. DETAILED DESCRIPTION

[0050] In order to make the above-mentioned purposes, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the drawings of the specification. Obviously, the described embodiments are part of the embodiments of the present invention, but not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary persons in the art without creative work should fall within the scope of protection of the present invention.

[0051] In the following description, many specific details are set forth to facilitate a full understanding of the present invention, but the present invention may also be implemented in other ways different from those described herein, and those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0052] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The term "in one embodiment" that appears in different places in this specification does not necessarily refer to the same embodiment, nor does it refer to a separate or selective embodiment that is mutually exclusive with other embodiments.

[0053] The present invention is described in detail with reference to schematic diagrams. When describing the embodiments of the present invention, for the sake of convenience, the cross-sectional diagrams showing the device structure will not be partially enlarged according to the general scale, and the schematic diagrams are only examples, which should not limit the scope of protection of the present invention. In addition, in actual production, the three-dimensional dimensions of length, width and depth should be included.

[0054] At the same time, in the description of the present invention, it should be noted that the orientations or positional relationships indicated by the terms "upper, lower, inner and outer" are based on the orientations or positional relationships shown in the drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the system or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as limiting the present invention. In addition, the terms "first, second or third" are only used for descriptive purposes and cannot be understood as indicating or implying relative importance.

[0055] In the present invention, unless otherwise clearly specified and limited, the terms "install, connect, connect" should be understood in a broad sense, for example: it can be a fixed connection, a detachable connection or an integral connection; it can also be a mechanical connection, an electrical connection or a direct connection, or it can be indirectly connected through an intermediate medium, or it can be the internal communication of two components. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0056] Example 1

[0057] Reference Figure 1-3 , which is an embodiment of the present invention, provides a method for predicting tail gas emissions from a thermal power plant combustion furnace, comprising:

[0058] S100: Obtaining tail gas data of thermal power plant combustion furnace;

[0059] In the embodiment of the present application, the tail gas data of the thermal power plant combustion furnace includes the raw flue gas SO 2 Concentration, net flue gas SO 2 concentration, net flue gas dust concentration, CO concentration, CO content of flue gas on the A side and CO content of flue gas on the B side, etc.;

[0060] After obtaining the exhaust gas data of thermal power plant combustion furnaces, the data is preprocessed. The preprocessing includes: cleaning the data, eliminating abnormal, missing and erroneous records, ensuring the rationality of the data, and processing missing values ​​through interpolation and other methods; transforming the data distribution, such as logarithmic or square root transformation, to reduce skewness, and standardizing or normalizing to eliminate dimensional differences; laying a solid foundation for subsequent accurate analysis and prediction.

[0061] S102: Performing discrete Fourier transform on the tail gas data of the thermal power plant combustion furnace to obtain dominant component data and stable component data;

[0062] Preferably, the tail gas data of the thermal power plant combustion furnace is converted into Fourier components in complex form, and the frequency component set with the largest amplitude is selected;

[0063] It should be noted that frequency adaptive normalization removes the first K dominant components for each input instance in the frequency domain, so that the prediction backbone network can focus on the stationary aspects of the input. This process is called frequency residual learning. At the input X at time point t, t is multivariate data with D dimensions, each dimension may have a different frequency pattern; therefore, frequency residual learning is applied to each dimension in a channel-independent setting;

[0064] In the embodiment of the present application, the frequency residual learning is performed by analyzing the tail gas data X of the thermal power plant combustion furnace. t It is realized by performing one-dimensional discrete Fourier transform, which is specifically expressed as:

[0065] Z t =DFT(X t )

[0066] K t =TopK(Amp(Z t ))

[0067] Among them, X t is the exhaust gas data of thermal power plant combustion furnace, Z t The data obtained by Fourier transforming the exhaust gas data of thermal power plant combustion furnace. K t is the frequency component set;

[0068] It should be noted that TopK(·) selects the top K frequency component sets with the largest magnitude, which is calculated by the Amp(·) function;

[0069] Restore the frequency component set to the time domain component and obtain the dominant component data, which is expressed as:

[0070]

[0071] in, is the dominant component data, K t is the frequency component set, Z t Convert the exhaust gas data of thermal power plant into Fourier components in complex form;

[0072] It should be noted that Filter is from Z t K t Frequency operation, in order to reduce the impact of non-stationary signals, frequency residual learning restores the first K components to time domain components In the implementation, IDFT(·) is used;

[0073] Preferably, the dominant component data is subtracted from the Fourier component to obtain the stationary component data;

[0074] Specifically, the formula for obtaining stable component data is expressed as:

[0075]

[0076] Among them, X res For stationary component data, is the dominant component data, X t It is the exhaust gas data of thermal power combustion furnace;

[0077] It should be noted that DFT(·) and IDFT(·) can be performed using Fast Fourier Transform FFT, with a computational complexity of O(LlogL); while TopK and Filter operations both have a complexity of O(L+K). It should be noted that all these operations are GPU-friendly and can be fully parallelized; therefore, the impact of applying these operations to each dimension independently can be largely mitigated. After the normalization step, the normalized sequence can be more stationary and have a more consistent distribution of covariates, a process such as Figure 2 shown.

[0078] S104: inputting the stationary component data into the prediction backbone model for prediction to obtain a first prediction value, and inputting the dominant component data into the MLP model for prediction to obtain a second prediction value;

[0079] Preferably, the prediction backbone model receives stationary component data as input and passes it through multiple fully connected layers;

[0080] Preferably, in each fully connected layer, the stationary component data is weighted summed with the neurons of the previous layer by the learned weights, and a bias term is added to extract stationary features;

[0081] Preferably, the stationary features are weighted and transformed through a nonlinear activation function to obtain enhanced stationary features;

[0082] Preferably, the enhanced stationary features are transmitted to the output layer, and the output layer obtains the first prediction value by calculating the linear activation function based on the learned weights and biases;

[0083] In the embodiment of the present application, the prediction backbone model receives the transformed data as input and only predicts the stable part of the output, which is expressed as:

[0084]

[0085] Among them, g θ is the prediction backbone model, X res is the stable component data;

[0086] In another possible embodiment, the prediction backbone model may adopt a convolutional neural network, a long short-term memory network, or a hybrid model of a convolutional neural network combined with a long short-term memory network;

[0087] Taking the hybrid model as an example, the stationary component data is organized in the form of time series and fed into the convolutional neural network as input. In the convolutional neural network, high-level local features can be automatically extracted from the original data through multiple convolutional layers, activation layers, and pooling layers. These features are very critical for subsequent prediction tasks. Subsequently, the feature map output by the convolutional neural network is flattened into a one-dimensional vector and used as the input of the long short-term memory network. The long short-term memory network layer can effectively capture the temporal dependency and long-term memory in the data through its internal forget gate, input gate, and output gate mechanism.

[0088] After the LSTM network layer, one or two fully connected layers (Dense layers) can be added to further integrate the features of the LSTM network output and prepare the final prediction output. The first prediction value is calculated at the output layer through a linear activation function. This prediction value represents the future trend or state of the stationary component data.

[0089] Preferably, the MLP model is used to predict the future value of the dominant component data to obtain a second predicted value, which is expressed as:

[0090]

[0091] Among them, W 1 , W 2 , W 3 are all learnable parameters, X non,t is the dominant component data, X t The exhaust gas data of thermal power plant is converted into Fourier components in complex form, ReLU is the activation function, is the second predicted value;

[0092] It should be noted that for reverse instance normalization, a second prediction value needs to be estimated in the output. Since the input and its corresponding output are quite close, RevIN directly adds the dominant component data back by assuming that the second prediction value has exactly the same trend as the dominant component data; however, this assumption is difficult to hold because the non-stationary information between the input and output may change. In addition, although SAN and Dish-TS predict statistics to resolve the difference between the input and output, these statistics can only represent the most significant trend patterns. Therefore, the present invention does not predict statistics, but uses a simple MLP model q φ Directly predict the future values ​​of the first K frequency components of D dimensions;

[0093] It should also be noted that by inputting the stationary component data and the dominant component data into the prediction backbone model and the MLP model respectively, this method achieves accurate prediction of the exhaust gas data of thermal power plant combustion furnaces; the prediction backbone model focuses on capturing and predicting the future trend of the stationary component, and through multi-level feature extraction and transformation, it effectively enhances the model's ability to identify stationary characteristics; and for the dominant component, the MLP model directly predicts its future value, avoiding the excessive reliance on statistical data in traditional methods, so that it can more flexibly respond to non-stationary changes between input and output. This dual-model parallel prediction strategy not only improves the accuracy and robustness of the prediction, but also reduces the error caused by insufficient assumptions of a single model, providing strong technical support for the processing and analysis of complex time series data such as exhaust gas from thermal power plant combustion furnaces.

[0094] Preferably, the prior guidance loss function and the overall loss function are introduced to optimize the backbone prediction model and the MLP model, which can be expressed as:

[0095]

[0096] Among them, L nonstat,φ is the overall loss function, L forecast,θ,φ is the prior guidance loss function, is the second predicted value, Y non,t is the offset prediction value, It is the predicted value of tail gas emission from thermal power combustion furnace.

[0097] Preferably, the prior guidance loss function and the overall loss function are used as the objective function; the parameter space of the backbone prediction model and the MLP model is determined and the parameters are initialized; the objective function is modeled using the Bayesian algorithm, a probability model is established, and the parameter combination is evaluated using the probability model to obtain the optimal parameter combination;

[0098] It should be noted that by introducing the prior guidance loss function and the overall loss function as the objective function for optimizing the backbone prediction model and the MLP model, this method significantly improves the accuracy and reliability of model prediction; this strategy of joint optimization of multiple loss functions not only considers the model's direct fitting ability to actual data, but also incorporates prior knowledge or domain-specific information, effectively guiding the model to avoid overfitting during training, while better capturing the inherent laws and potential features in the data. In addition, the Bayesian algorithm is used to model the objective function, and a probabilistic model is constructed to evaluate the performance of different parameter combinations, and finally the optimal parameter combination is obtained. This process further enhances the generalization ability and stability of the model.

[0099] S106: combining the first prediction value and the second prediction value to obtain a prediction value of tail gas emissions from a thermal power plant combustion furnace;

[0100] In the embodiment of the present application, the first prediction value and the second prediction value are combined and expressed as follows:

[0101]

[0102] in, is the first predicted value, is the second predicted value, is the predicted value of tail gas emission from thermal power combustion furnace;

[0103] In another possible embodiment, the prediction value of the exhaust gas emissions from the thermal power combustion furnace can be obtained by considering the different weights of the two and performing a weighted summation. This method allows the contribution of the stable component and the dominant component to be dynamically adjusted according to their relative importance in the overall exhaust gas emissions; specifically, a weight coefficient α (0≤α≤1) can be set, where α represents the weight of the stable component prediction value in the final prediction value, and (1-α) represents the weight of the dominant component prediction value. By adjusting the value of α, the influence of the two prediction values ​​on the final prediction result can be flexibly controlled; when α is close to 1, it means that the stable component has a greater influence on the prediction result; and when α is close to 0, it indicates that the effect of the dominant component is more significant.

[0104] It should be noted that the present invention adopts a dual-model parallel prediction strategy, combines the prediction backbone model and the MLP model, and accurately predicts the stationary component and the dominant component of the exhaust gas data of thermal power combustion furnaces, respectively, which significantly improves the accuracy and reliability of the prediction results; the prediction backbone model effectively captures and predicts the future trend of the stationary component through multi-level feature extraction and transformation, while the MLP model directly predicts the future value of the dominant component, avoiding the excessive reliance of traditional methods on statistical data and flexibly responding to non-stationary changes between input and output; by introducing the prior guidance loss function and the overall loss function to jointly optimize the model, and using the Bayesian algorithm for parameter tuning, the generalization ability and stability of the model are further enhanced, and the accurate prediction of exhaust gas emissions from thermal power combustion furnaces is achieved.

[0105] The above is a schematic scheme of a method for predicting the exhaust gas emissions of a thermal power plant combustion furnace in this embodiment. It should be noted that the technical scheme of the system for predicting the exhaust gas emissions of a thermal power plant combustion furnace and the technical scheme of the method for predicting the exhaust gas emissions of a thermal power plant combustion furnace described above belong to the same concept. For details not described in detail in the technical scheme of the system for predicting the exhaust gas emissions of a thermal power plant combustion furnace in this embodiment, please refer to the description of the technical scheme of the method for predicting the exhaust gas emissions of a thermal power plant combustion furnace described above.

[0106] The thermal power plant combustion furnace tail gas emission prediction system in this embodiment includes:

[0107] A data acquisition module, used to obtain tail gas data of thermal power combustion furnaces;

[0108] The classification module is used to perform discrete Fourier transform on the tail gas data of thermal power combustion furnaces to obtain the dominant component data and the stable component data;

[0109] The classification prediction module is used to input the stationary component data into the prediction backbone model for prediction to obtain a first prediction value, and input the dominant component data into the MLP model for prediction to obtain a second prediction value;

[0110] The result prediction module is used to combine the first prediction value and the second prediction value to obtain the prediction value of the tail gas emission of the thermal power combustion furnace.

[0111] This embodiment also provides a computing device, which is applicable to the case of predicting the exhaust gas emission of a thermal power plant combustion furnace, and includes:

[0112] Memory and processor; the memory is used to store computer executable instructions, and the processor is used to execute computer executable instructions to implement the method for predicting tail gas emissions from thermal power combustion furnaces as proposed in the above embodiment.

[0113] This embodiment also provides a storage medium on which a computer program is stored. When the program is executed by a processor, the method for predicting the exhaust gas emission of a thermal power plant combustion furnace as proposed in the above embodiment is implemented.

[0114] The storage medium proposed in this embodiment and the method for predicting exhaust gas emissions from thermal power plant combustion furnaces proposed in the above embodiment belong to the same inventive concept. Technical details not fully described in this embodiment can be found in the above embodiment, and this embodiment has the same beneficial effects as the above embodiment.

[0115] Through the above description of the implementation methods, the technicians in the relevant field can clearly understand that the present invention can be implemented by means of software and necessary general hardware, and of course can also be implemented by hardware, but in many cases the former is a better implementation method. Based on such an understanding, the technical solution of the present invention is essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as a computer floppy disk, read-only memory (ReadOnly, Memory, ROM), random access memory (RandomAccess Memory, RAM), flash memory (FLASH), hard disk or optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, server, or network device, etc.) to perform the methods of various embodiments of the present invention.

[0116] Example 2

[0117] Referring to Table 1, an embodiment of the present invention provides a method for predicting tail gas emissions from a thermal power plant combustion furnace. In order to verify its beneficial effects, comparison results of several schemes are provided.

[0118] In order to verify the effectiveness of the proposed method for predicting tail gas emissions from thermal power plant combustion furnaces, a series of experiments were designed and implemented;

[0119] Detailed exhaust gas data are collected from the monitoring system of thermal power furnaces, including but not limited to raw flue gas SO 2 Concentration, net flue gas SO 2 Key indicators such as concentration, net flue gas dust concentration, CO concentration, and CO content in flue gas on sides A and B. These data cover different operating conditions and time periods, ensuring the diversity and representativeness of the samples.

[0120] Next, the collected exhaust data were preprocessed, including data cleaning and normalization, to eliminate the influence of outliers and dimensional differences on the prediction results. Discrete Fourier Transform (DFT) technology was applied to process these data. By converting the exhaust data into complex Fourier components, the frequency component set with the largest amplitude was successfully identified and selected. These components represent the dominant components in the exhaust data.

[0121] These frequency component sets are restored to the time domain to obtain the dominant component data; at the same time, the dominant component data are subtracted from the original Fourier component to obtain the stationary component data.

[0122] In the prediction stage, the stationary component data is input into the pre-trained prediction backbone model, which is built on existing frameworks such as DLinear, FEDformer or Informer, and extracts stationary features through multiple fully connected layers, and finally outputs the first prediction value of the stationary component through a linear activation function. At the same time, the dominant component data is input into the MLP (multi-layer perceptron) model, and the nonlinear processing capability of the MLP is used to predict the future value of the dominant component and obtain the second prediction value.

[0123] Finally, the first prediction value and the second prediction value are fused to obtain the final prediction value of the tail gas emissions of the thermal power plant combustion furnace. In order to evaluate the prediction effect, multiple performance indicators are calculated, including mean square error (MSE) and mean absolute error (MAE), and these methods are compared with other baseline methods (such as direct prediction using DLinear, FEDformer, and Informer).

[0124] The experimental results show that the method for predicting the exhaust gas emission of thermal power generation furnaces of the present invention has achieved significant improvements in various indicators. 2 Concentration, net flue gas SO2 When the concentration, net flue gas dust concentration, CO concentration, and the CO content of flue gas on the A side and the B side are measured, the MSE and MAE values ​​of this method are lower than those of other baseline methods, proving the superiority of this method in prediction accuracy. This finding not only verifies the effectiveness of the method of the present invention, but also provides new ideas and technical means for the accurate prediction of tail gas emissions from thermal power combustion furnaces.

[0125] In the parameter search stage of the prediction backbone model and the simple MLP model, the Bayesian optimization algorithm is used to initialize the parameter space, set the reasonable value range of multiple parameters including the learning rate, the number of hidden layer nodes and the activation function type, and use the Gaussian process to build a probability model to capture the complex relationship between parameters and predict the performance of different parameter combinations;

[0126] During the iterative search process, the Bayesian optimization algorithm intelligently selects the parameter combination that is most likely to improve the model performance for evaluation based on the prediction results of the probability model. After each evaluation, the new parameter combination and its performance results are added to the training set to update the probability model. When the improvement in model performance in multiple consecutive iterations is lower than the preset threshold or reaches the maximum number of iterations, the search process has converged and the optimal parameter combination is obtained.

[0127] In order to verify the effectiveness of the proposed method in predicting the exhaust gas emissions of thermal power plant combustion furnaces, we selected three existing prediction model frameworks, DLinear, FEDformer and Informer, as benchmarks. In the prediction process, the predicted values ​​of the two component data were obtained respectively, and they were combined to obtain the final predicted value of the exhaust gas emissions of thermal power plant combustion furnaces;

[0128] The comparison is carried out through indicators such as mean square error (MSE) and mean absolute error (MAE), and the results are shown in Table 1;

[0129] Table 1 Prediction results of tail gas emissions from thermal power plant combustion furnaces using different methods

[0130]

[0131]

[0132] As shown in Table 1, in the prediction task of the exhaust gas emissions from thermal power plant combustion furnaces, no matter which framework is used, after applying the method of the present invention, its prediction performance has been significantly improved, which is specifically manifested in that the values ​​of various prediction indicators, MSE and MAE, are at a relatively low level.

[0133] In summary, the model of the present invention has shown obvious advantages in the field of prediction of tail gas emissions from thermal power combustion furnaces, which not only improves the prediction accuracy, but also enhances the stability and generalization ability of the model. This achievement is of great significance to environmental protection supervision, emission control, and energy conservation and emission reduction in the thermal power industry, and is expected to provide strong support for the green development and sustainable development of the industry.

[0134] In summary, this experiment successfully verified the effectiveness and accuracy of the proposed method for predicting tail gas emissions from thermal power plant combustion furnaces. This method innovatively decomposes tail gas data into stable components and dominant components, and uses targeted models for prediction respectively, and then combines the prediction results to achieve accurate prediction of tail gas emissions. This achievement not only provides strong technical support for the management of tail gas emissions from thermal power plant combustion furnaces, but also provides new ideas and methods for the design of prediction models in related fields.

[0135] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A method for predicting tail gas emissions from a thermal power plant combustion furnace, characterized in that: include: Obtain exhaust data from thermal power plant combustion furnaces; Performing discrete Fourier transform on the tail gas data of the thermal power plant combustion furnace to obtain dominant component data and stable component data; Input the stationary component data into the prediction backbone model for prediction to obtain a first prediction value, and input the dominant component data into the MLP model for prediction to obtain a second prediction value; The first predicted value and the second predicted value are combined to obtain a predicted value of tail gas emissions from a thermal power plant combustion furnace.

2. The method for predicting tail gas emissions from a thermal power plant combustion furnace according to claim 1, characterized in that: The exhaust gas data of the thermal power plant combustion furnace is subjected to discrete Fourier transform to obtain the dominant component data, including: Convert the tail gas data of the thermal power plant combustion furnace into Fourier components in complex form, and select a frequency component set with the largest amplitude; The frequency component set is restored to the time domain component to obtain the dominant component data, which is expressed as: in, is the dominant component data, K t is the frequency component set, Z t Convert the exhaust gas data of thermal power plant into Fourier components in complex form.

3. The method for predicting tail gas emissions from a thermal power plant combustion furnace according to claim 2, characterized in that: Obtaining stationary composition data includes, The dominant component data is subtracted from the Fourier component to obtain the stationary component data.

4. The method for predicting tail gas emissions from a thermal power plant combustion furnace according to claim 3, characterized in that: Obtaining the first prediction value includes, The prediction backbone model receives the stationary component data as input and passes it through multiple fully connected layers; In each fully connected layer, the stationary component data is weighted and summed with the neurons in the previous layer using the learned weights, and a bias term is added to extract stationary features; The stationary features are weighted and transformed by a nonlinear activation function to obtain enhanced stationary features; The enhanced stationary features are transmitted to the output layer, and the output layer obtains the first prediction value by using a linear activation function calculation based on the learned weights and biases.

5. The method for predicting tail gas emissions from a thermal power plant combustion furnace according to claim 2, characterized in that: Obtaining the second prediction value includes, The MLP model is used to predict the future value of the dominant component data to obtain a second predicted value, which is expressed as: Among them, W1, W2, and W3 are all learnable parameters, and X non,t is the dominant component data, X t The exhaust gas data of thermal power plant is converted into Fourier components in complex form, ReLU is the activation function, is the second predicted value.

6. The method for predicting tail gas emissions from a thermal power plant combustion furnace according to claim 1, characterized in that: The backbone prediction model and MLP model include, The prior guidance loss function and the overall loss function are introduced to optimize the backbone prediction model and the MLP model, which are expressed as: Among them, L nonstat,φ is the overall loss function, L forecast,θ,φ is the prior guidance loss function, is the second predicted value, Y non,t is the offset prediction value, It is the predicted value of tail gas emission from thermal power combustion furnace.

7. The method for predicting tail gas emissions from a thermal power plant combustion furnace according to claim 6, characterized in that: Also includes, Taking the prior guidance loss function and the overall loss function as objective functions; Determine the parameter space of the backbone prediction model and the MLP model and initialize the parameters; The objective function is modeled using a Bayesian algorithm to establish a probability model, and the parameter combination is evaluated using the probability model to obtain the optimal parameter combination.

8. A system for predicting tail gas emissions from thermal power plant combustion furnaces, characterized in that: include, A data acquisition module, used to obtain tail gas data of thermal power combustion furnaces; A classification module, used for performing discrete Fourier transform on the tail gas data of the thermal power combustion furnace to obtain dominant component data and stable component data; A classification prediction module, used for inputting the stationary component data into a prediction backbone model for prediction to obtain a first prediction value, and inputting the dominant component data into an MLP model for prediction to obtain a second prediction value; The result prediction module is used to combine the first prediction value and the second prediction value to obtain a prediction value of tail gas emissions from a thermal power plant combustion furnace.

9. An electronic device, comprising: Memory and processor; The memory is used to store computer executable instructions, and the processor is used to execute the computer executable instructions. When the computer executable instructions are executed by the processor, the steps of the method for predicting exhaust gas emissions from a thermal power plant combustion furnace as described in any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, implement the steps of the method for predicting exhaust gas emissions from a thermal power plant combustion furnace as claimed in any one of claims 1 to 7.

Images