CNN-ELM-based boiler efficiency soft measurement method and system, and medium

Abstract

The invention discloses a CNN-ELM-based boiler efficiency soft measurement method and system, and a medium. The method comprises the following steps: step 1, collecting historical operation data; step 2, preprocessing the data; step 3, a CNN-ELM model is constructed; 4, training the CNN-ELM model by using the training set, and determining an optimal network structure and parameter setting; step 5, inputting test set data into the optimized CNN-ELM model, obtaining a prediction result, performing reverse normalization processing, and then analyzing and evaluating the prediction result by using an evaluation function; and step 6, inputting boiler operation data collected in real time into the trained CNN-ELM model, and outputting a dynamic thermal efficiency prediction value. The method can effectively overcome the defects of a single model, achieves the high-precision dynamic measurement of the thermal efficiency of the boiler, and remarkably improves the precision and real-time performance of the measurement of the efficiency of the boiler.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of boiler thermal efficiency measurement, and particularly relates to a boiler efficiency soft measurement method and system based on CNN-ELM and a medium. BACKGROUND

[0002] Coal-fired units are used as stable power sources for power systems at the beginning of their design. In the context of deep peak regulation, units are forced to run at low loads for a long time, deviating from the design operating condition, which seriously damages the thermal efficiency of the boiler. The premise for improving the thermal efficiency of boilers under complex conditions such as deep regulation and transient state is to establish an accurate thermal efficiency prediction model. Therefore, this paper conducts research on dynamic modeling of thermal efficiency.

[0003] The research on modeling of the boiler combustion process includes two branches: mechanism method and data-driven method. The mechanism method is based on the dynamics model, energy conservation law, etc. involved in the combustion and related processes, and then a mathematical model reflecting the process mechanism is established. Previous researchers have established a mathematical model describing the thermal efficiency and NOx emission of the boiler based on the distribution of coal in the furnace and the combustion mechanism of over-fire air (OFA). By simplifying the combustion system, a discrete-time linear state-space model is established. It is also a common research method to establish a mechanism model of the boiler combustion process by computational fluid dynamics (CFD), such as establishing a CFD model of the coupling of the boiler air and flue gas side and the steam-water side based on the mechanism method, using a Dell 3600 workstation with a 3GHz main frequency for 4-thread parallel operation. However, it takes about a week to solve a single operating condition, which is unacceptable for coal-fired power plants. Taking the prediction of the thermal efficiency of the boiler as an example, the coking, wear and tear, and replacement of the heating surface will all affect the characteristics of the thermal process. Therefore, as the running time increases, the mechanism model will gradually mismatch the actual operating condition. Moreover, the coal-fired boiler combustion process is complex, and the coal combustion mechanism in the furnace involves air dynamics, two-phase flow, heat and mass transfer, etc. The online prediction accuracy and real-time performance of the thermal efficiency calculated by the mechanism model are difficult to guarantee. In summary, the applicability of mechanism modeling is limited.

[0004] Compared with the mechanism method, the data-driven method directly excavates the mapping relationship between system parameters from data with the help of machine learning algorithms, and has fast modeling speed and high accuracy. The machine learning algorithm based on neural network is more applicable to industrial processes whose mechanism is not clear, and the data-driven modeling method will be more applicable.

[0005] The previous modeling method does not consider the time scale in the input dimension, belongs to the category of static model, and is suitable for the stable state of the boiler. Specifically, when the boiler starts from the cold state or switches to different loads, the historical working conditions will affect the current thermal efficiency. This is because when the boiler load changes, the heat storage working medium will be released or rebalanced. When the boiler is in a stable running state, the heat storage effect can usually improve the combustion efficiency and thermal efficiency. Therefore, it is necessary to establish a dynamic model for the time sequence characteristics of the boiler thermal efficiency, that is, to include the historical working conditions in the model input.

[0006] Convolutional Neural Network (CNN) has achieved remarkable results in image processing and pattern recognition due to its powerful feature extraction capability; and Extreme Learning Machine (ELM) as a kind of high-efficiency single-hidden-layer feedforward neural network, has the advantages of fast training speed and good generalization performance. However, in the prior art, the method of using CNN or ELM alone still has certain limitations in boiler thermal efficiency measurement. For example, CNN may face overfitting problem when processing high-dimensional data, and ELM is relatively weak in feature extraction capability, making it difficult to fully mine deep information in boiler operation data. SUMMARY

[0007] To solve the above problems, the purpose of the present application is to provide a boiler efficiency soft measurement method, system and medium based on CNN-ELM, which combines the feature extraction capability of CNN and the efficient learning capability of ELM to construct a hybrid model, which can effectively overcome the shortcomings of single model and realize high-precision dynamic measurement of boiler thermal efficiency.

[0008] To achieve the above purpose, the present application adopts the following technical solutions:

[0009] A boiler efficiency soft measurement method based on CNN-ELM, comprising the following steps:

[0010] Step 1, collecting historical operation data;

[0011] Step 2, pre-processing the data obtained in step 1, including missing data processing, data dimension reduction, data reconstruction, and then dividing the data set into training set and test set;

[0012] Step 3, constructing a CNN-ELM model, the CNN-ELM model is composed of an input layer, a CNN network, an ELM and an output layer connected in turn, wherein: the input layer receives data, the CNN network extracts features, the ELM receives the output of the CNN network and predicts the boiler thermal efficiency, and the output is output by the output layer;

[0013] Step 4, training the CNN-ELM model using the training set, optimizing the parameters of the CNN-ELM model through optimization algorithm to determine the best network structure and parameter setting;

[0014] Step 5, input the test set data into the optimized CNN-ELM model, obtain the prediction results and perform inverse normalization processing, then use the evaluation function to analyze and evaluate the prediction results, and verify the prediction performance of the model under different working conditions;

[0015] Step 6, input the real-time collected boiler operation data into the trained CNN-ELM model, and output the dynamic thermal efficiency prediction value.

[0016] Further, in step 1, the collected historical operation data includes thermal efficiency and auxiliary variables related thereto.

[0017] Further, in step 2, the step of missing data processing is: adopting 3σ-rule to detect abnormal data, and replacing the abnormal data based on the following formula:

[0018] Δx i (t) = x i (t) - x i (t-1)

[0019]

[0020] Where, Δx i (t) represents the auxiliary variable x i The change rate of the adjacent two time points, Δx i (t) represents the average change rate of Δx i (t) before time t, σ (t) represents the standard deviation, t = 2, …, n is the sampling time.

[0021] Further, in step 2, the step of data dimension reduction is: using the random forest algorithm to sort the importance of the auxiliary variables, and screening the top 10 important variables as input variables.

[0022] Further, in step 2, the step of data reconstruction is: using the sliding window method to convert the time series data into a two-dimensional tensor, wherein the time dimension includes 20 consecutive sampling time points, and the feature dimension is the screened 10 variables.

[0023] Further, in step 3, the CNN network includes at least two convolutional layers, and a channel equalization block (CE-Block) is connected after each convolutional layer, which is used to suppress the distribution difference of feature values between channels.

[0024] Further, in step 4, the hyperparameters of the CNN-ELM model are set as: batch size 32, optimizer Adam, and activation function ReLU.

[0025] A system for the method, comprising:

[0026] a data acquisition module, which acquires real-time boiler DCS system operation data;

[0027] a preprocessing module, which performs missing data processing, data dimension reduction and data reconstruction;

[0028] a model operation module, which deploys a CNN-ELM model and performs prediction calculation;

[0029] a visual terminal, which displays real-time boiler thermal efficiency prediction results.

[0030] A storage medium, which stores a computer program, the computer program being executed by a processor to implement the CNN-ELM-based boiler efficiency soft measurement method.

[0031] Further, the storage medium is integrated into a coal-fired power plant DCS system or an edge computing terminal, and real-time boiler operation data are received and thermal efficiency prediction results are output.

[0032] Beneficial effects: The application combines the feature extraction capability of CNN and the efficient learning capability of ELM to construct a hybrid model, which can effectively overcome the shortcomings of a single model, realize high-precision dynamic measurement of boiler thermal efficiency, and significantly improve the precision and real-time performance of boiler thermal efficiency measurement. The application designs a dynamic soft measurement method suitable for boiler operation data, which can adapt to multivariate and nonlinear characteristics under complex working conditions, and has strong robustness and generalization capability. Specifically, the application uses CNN to perform multi-level feature extraction on boiler operation data to capture nonlinear relationships in the data; then, the extracted features are input into ELM for rapid modeling and prediction, thereby realizing real-time and dynamic measurement of boiler thermal efficiency; finally, the model is evaluated using real data under variable load conditions of a 600MW coal-fired boiler. In addition, the application also provides a storage medium for realizing automatic operation of the above method, further improving the measurement efficiency and practicality. BRIEF DESCRIPTION OF DRAWINGS

[0033] Figure 1 Flowchart of the CNN-ELM-based boiler efficiency soft measurement method in the embodiment;

[0034] Figure 2 Schematic diagram of data reconstruction in the embodiment, wherein (a) is a time series variable, and (b) is a tensor graph obtained by a sliding window;

[0035] Figure 3 Schematic diagram of data set segmentation in the embodiment;

[0036] Figure 4 Structural schematic diagram of a convolutional neural network (CNN);

[0037] Figure 5 is a schematic diagram of an ELM structure;

[0038] Figure 6 is a prediction result graph of a CNN-ELM in the embodiment on a test set;

[0039] Figure 7 is a prediction result graph of a CNN in the embodiment on a test set;

[0040] Figure 8 is a prediction result graph of an ELM in the embodiment on a test set. DETAILED DESCRIPTION

[0041] The application will be further explained in connection with the accompanying drawings and embodiments.

[0042] Embodiment

[0043] As shown in the figure, the boiler efficiency soft measurement method based on a CNN-ELM in the embodiment comprises the following steps: Figure 1 (1) Collect historical operation data from a boiler DCS system.

[0044] The research object of the embodiment is a 600 MW supercritical thermal power unit W flame boiler. The boiler is a supercritical parameter, W flame combustion, variable pressure operation direct current boiler. The main design parameters of the boiler are shown in Table 1. The power plant is located in the hinterland of southwest China, the climate is humid, and the low-quality anthracite with a volatile content of less than 12% is used, which leads to a large amount of coking in the furnace. The coking phenomenon seriously affects the combustion atmosphere and heat transfer in the furnace, making it more difficult to predict the boiler thermal efficiency through a mechanism model, so a data-driven modeling method is more suitable for such industrial scenarios.

[0045] Table 1 Main design parameters of the boiler

[0046]

[0047] According to the combustion mechanism and the suggestions of the operation engineer, the application collects 53 related auxiliary variables such as wind, coal and water, and the prediction target is the boiler thermal efficiency, as shown in Table 2. The selection of auxiliary variables comprehensively considers the influence factors such as the furnace aerodynamic field, the wind-coal two-phase flow field and the working medium heat absorption. Among them, the secondary air controls the main air required for combustion in the furnace, the mill inlet capacity wind flow controls the distribution of the wind-coal two-phase flow field in the furnace, and the main steam pressure, main steam temperature and other variables reflect the working medium heat absorption level in the furnace.

[0048] (2) Preprocess the obtained data.

[0049]

[0050] ​The historical data of 5019 samples with a sampling period of 2 minutes were collected from the target boiler DCS for about 7 days of continuous operation. The original data were preprocessed to meet the model input requirements, and the data preprocessing process included missing data processing, data dimensionality reduction, data reconstruction, and data set division.

[0051] (2.1) Missing data processing

[0052] Due to the complex production site environment of the boiler, the signal is contaminated with noise, resulting in different degrees of abnormal values in the original data stored in the DCS. In this embodiment, 3σ-rule is used to detect abnormal data, and abnormal data is replaced based on the following formula:

[0053] Table 2 Auxiliary variable table

[0054]

[0055]

[0056] △x i (t)=x i (t)-x i (t-1)

[0057]

[0058]

[0059] wherein, △x i (t) represents the auxiliary variable x i The change rate of the adjacent two time points, represents the average change rate of △x i (t) before time t, σ i (t) represents the standard deviation, t = 2, …, n is the sampling time.

[0060] (2.2) Data dimensionality reduction

[0061] This embodiment proposes to use feature engineering to process the auxiliary variables based on the high-dimensional nonlinearity and large inertia of the boiler combustion system, and analyzes the rationality of supervised dimensionality reduction in the combustion process based on real historical data.

[0062] Random Forest (RF) is an extended variant of Bagging, which is particularly suitable for assessing the importance of high-dimensional nonlinear variables. Because RF uses several independent decision trees for parallel prediction, random and with replacement sampling makes out-of-bag data naturally capable of evaluating the deterioration of model prediction error, which indirectly indicates the importance of the variable. For each decision-making decision tree, the in-bag data is used to train the tree, and the out-of-bag data (OOB) is used to evaluate the tree. The deterioration of prediction accuracy will be achieved by adding noise interference to important variables. The importance of input variables before and after interference is measured by the following formula:

[0063]

[0064] Wherein, The prediction error of the out-of-bag data on the decision tree b before and after interference, respectively, B is the number of decision trees. The sum of the importance of all input variables is 1, that is, i=1,2,…,m represents the number of auxiliary variables.

[0065] The RF algorithm is used to evaluate the feature importance of 53 comprehensive auxiliary variables affecting the boiler thermal efficiency, and the decision tree is set to 500. The sum of the importance of the top 5 variables is more than 0.9. The author combines the prior knowledge of the boiler, and the F layer secondary air door which mainly affects the combustion in the furnace is also included in the input variable. The load has a greater impact on the overall unit, so the input variable is determined to be 10. The total weight of the top 10 variables is 0.933, and the reordering is shown in Table 3.

[0066] The variable importance calculated by RF follows the change rule of boiler thermal efficiency, and the complexity and universality of the model are balanced. In this embodiment, the top 10 variables are selected as the input variables of the thermal efficiency model, and the flue gas temperature has the highest importance.

[0067] Table 3 Importance of variables related to boiler thermal efficiency (mean ± standard value)

[0068]

[0069] (2.3) Data reconstruction

[0070] The input format of the two-dimensional CNN is a continuous two-dimensional tensor. In order to meet the input requirements, the input variables need to be reconstructed. As time series data, the sliding window method is used to divide the data according to the feature dimension and the time dimension. The time dimension contains 20 sampling time points, i.e. H=20; the feature dimension is 10 variable types, i.e. W=10; and the window sliding step is 1. The label corresponding to each two-dimensional tensor is the thermal efficiency of the last time dimension of the tensor, Figure 2 The data reconstruction schematic diagram is shown in the figure.

[0071] (2.4) Dataset Splitting

[0072] The original 5019 samples are split into 5000 standard input matrices, each corresponding to the original sample sequence number of the label (thermal efficiency) 20-5019, which is reordered to 1-5000 to make it consistent with the matrix sequence number, i.e. the input matrix and the label sequence number are counted from 1, as shown in Figure 3 The data set splitting diagram is shown. The split data is divided into training set, validation set and test set according to the ratio of 6:2:2, and the data set maintains the original sampling order.

[0073] (3) Constructing a CNN-ELM model composed of an input layer, a CNN network, an ELM and an output layer connected in turn, wherein: the input layer receives data, the CNN network extracts features, the ELM receives the output of the CNN network and predicts the boiler thermal efficiency, and the output is output by the output layer.

[0074] Convolutional Neural Network (CNN) is a neural network specially designed for processing data with grid-like structure, which is composed of input layer, convolutional layer, activation layer, pooling layer and fully connected layer, as shown in Figure 4 .

[0075] The convolutional layer is the core part of CNN, each convolutional layer is composed of multiple convolutional kernels, each convolutional kernel is like a filter, the main function is to extract features from input data. With the help of convolution operation, the convolutional layer can effectively capture important features in the data and enhance the key features in the original signal, while effectively reducing the dimension of the features. The pooling layer is generally connected after the convolutional layer, and the main function is to extract data again, reduce the data required for prediction, and thus reduce the complexity of the network. Pooling operations mainly include average pooling and maximum pooling. Average pooling is to take the average value in a region, and maximum pooling is to take the maximum value in a region.

[0076] The activation layer enhances the non-linear ability of the model by applying an activation function to the output of the fully connected layer or the output of the convolutional layer. Commonly used activation functions include sigmoid function, ReLU function and tanh function, whose formulas are as follows:

[0077]

[0078] ReLU(x) = max(0, x)

[0079]

[0080] Compared with the other two, ReLU function shows stronger robustness, with the advantages of fast speed, good sparsity, and good derivation characteristics, so most CNNs use ReLU function as the activation function.

[0081] The fully connected layer is located at the end of the CNN, connected with the neurons of the upper structure, realizing the data dimension transformation while preserving the useful information of the data. Its main function is to learn the output features of the previous layers, integrate globally, and map to the label space of the sample.

[0082] ELM is a single hidden layer feedforward neural network algorithm. Before the algorithm is executed, the network randomly generates input weight and hidden layer node bias. The output weight matrix can be calculated to obtain the unique optimal solution under the corresponding structure, so ELM has the advantages of fast network training speed, good generalization, and strong robustness. The network structure of ELM is shown in Figure 5 .

[0083] Assuming that the input layer has n neurons, the hidden layer has l neurons, and the output layer has m neurons, the randomly generated bias is B, and the connection weight matrix between the input layer and the hidden layer is W:

[0084]

[0085] The connection weight matrix between the hidden layer and the output layer is β:

[0086]

[0087] At this time, ELM with k nodes can be described as

[0088]

[0089] In the formula, g(x) is the activation function.

[0090] If the initial β is determined, the above formula can be optimized as

[0091]

[0092] That is:

[0093] H0β=Y0

[0094] Where H0 is the hidden layer output matrix:

[0095]

[0096] Y0 is the network output matrix:

[0097]

[0098] For easy calculation, the above formula is converted to

[0099]

[0100] In the formula, Moore-Penrose generalized inverse of matrix H0, thus the global optimal output weight matrix can be obtained.

[0101] The activation function commonly used in ELM includes Sigmoid, Sine, Hardlim, etc., and the Sigmoid function is selected in the application.

[0102]

[0103] The CNN network mainly extracts object features through multiple stacked convolutional layers and sampling layers, and thus the CNN network can extract better classification features. However, the CNN network adopts a fully connected BP neural network as a perceiver, and the gradient descent method is used to find the minimum global error of the network during network training, so the training time is long and the network generalization ability is poor, so the perceiver of the CNN is not a very good classifier.

[0104] ELM is a single-hidden-layer feedforward neural network, the hidden layer weights and bias are randomly generated before training, and need not be adjusted during the training process, only the number of hidden layer neurons needs to be set, and a unique optimal solution can be obtained, so ELM has the advantages of fast training speed, good generalization and high classification accuracy. However, since ELM is a shallow neural network, its feature learning ability is limited, and only high accuracy can be obtained on the premise that the training data has good enough features.

[0105] In view of the various excellent performances of the convolutional neural network CNN and the single-hidden-layer feedforward neural network ELM, the CNN and the ELM are combined to construct a CNN-ELM network. The CNN network is composed of two convolutional layers (C1, C2) and two sampling layers (S1, S2), and the ELM is the last output layer. The entire network is composed of two parts, the features are first extracted through the CNN model, and then the ELM model receives the features.

[0106] The process of the CNN-ELM model is as follows: first, the network data is collected through the network data capture module, and the collected historical network data is preprocessed, including feature mapping digitization, normalization and dimension reconstruction, etc., to form a training data set; then, the CNN network is trained to convergence using the training data set, and the ELM network is trained using the highly abstract features extracted by the CNN, and finally the optimal CNN-ELM measurement model can be obtained. During online measurement, real-time network data generated in the first few seconds is collected through the network data capture module, and the real-time network data is preprocessed, and then the preprocessed test data set is input into the CNN-ELM model to complete the measurement classification of the input network connection.

[0107] (4) Train the CNN-ELM model using the training set, optimize the parameters of the CNN-ELM model through the optimization algorithm, and determine the optimal network structure and parameter settings;

[0108] Table 4 Structure of CNN-ELM with 10 variables

[0109]

[0110] To ensure the lightweight of the model structure, the CNN-ELM is only equipped with 2 convolutional layers and 2 sampling layers. As shown in Table 4, the hyperparameters of the CNN-ELM are determined through multiple training and verification. In addition, the batch size is 32, the epoch is 200, the optimizer is Adam, and the learning rate is 0.005.

[0111] (5) Input the test set data into the optimized CNN-ELM model, obtain the prediction results and perform inverse normalization processing, and then analyze and evaluate the prediction results using the evaluation function to verify the prediction performance of the model under different working conditions;

[0112] (6) Input the real-time collected boiler operation data into the trained CNN-ELM model to output the dynamic thermal efficiency prediction value.

[0113] To evaluate the performance of the thermal efficiency dynamic prediction model based on the CNN-ELM model, different modeling methods are compared with the CNN-ELM in this section, and the mean and standard deviation of the model prediction error come from twenty independent experiments.

[0114] The CNN-ELM, baseline CNN, and ELM models are compared on the test set. The CNN models all adopt a single hidden layer structure, and since the input is a 20x10 tensor, the hidden layer nodes contain 20 convolutional neurons respectively. The mean and standard deviation of the prediction error of CNN, ELM, and CNN-ELM are shown in Table 5.

[0115] Table 5 Prediction performance of LSTM, GRU, and CNN-ELM on the test set

[0116]

[0117] The RMSE of CNN-ELM is 0.202±0.006%, the MAE is 0.0895±0.008%, the R 2 2 is 0.966±0.005, and all performance indicators are better than the baseline CNN and ELM models. The prediction curve of a randomly selected experiment from twenty experiments is shown in Figure 6 , and the corresponding boiler load curve is shown in Figure 6 . Figure 7and Figure 8 The prediction results of the CNN, ELM on the test set. Among the three models, the deviation between the predicted value and the true value of CNN-ELM is the smallest, indicating that CNN-ELM can also be competent for various working conditions.

[0118] The application also provides a system for realizing the CNN-ELM-based boiler efficiency soft measurement method, comprising:

[0119] A data acquisition module acquires the boiler DCS system operation data in real time;

[0120] A preprocessing module performs missing data processing, data dimension reduction and data reconstruction;

[0121] A model operation module deploys the CNN-ELM model and performs prediction calculation;

[0122] A visual terminal displays the real-time boiler thermal efficiency prediction result.

[0123] The application also provides a storage medium storing a computer program, which realizes the steps of the CNN-ELM-based boiler thermal efficiency soft measurement method when executed by a processor. Further, the storage medium is integrated into a coal-fired power plant DCS system or an edge computing terminal, receives the boiler operation data in real time and outputs the boiler thermal efficiency prediction result.

[0124] The above only describes the preferred embodiments of the application, and it should be noted that those skilled in the art can make several improvements and refinements without departing from the principles of the application, and these improvements and refinements should also be considered within the protection scope of the application.

Claims

1. A soft measurement method for boiler efficiency based on CNN-ELM, characterized in that: Includes the following steps: Step 1: Collect historical operational data; Step 2 involves preprocessing the data obtained in Step 1, including: handling missing data, data dimensionality reduction, and data reconstruction. Then, the dataset is divided into training and testing sets. Step 3: Construct a CNN-ELM model. The CNN-ELM model consists of an input layer, a CNN network, an ELM, and an output layer connected in sequence. The input layer receives data, the CNN network extracts features, the ELM receives the output of the CNN network and predicts the boiler thermal efficiency, and the output layer outputs the results. Step 4: Train the CNN-ELM model using the training set, optimize the CNN-ELM model parameters using the optimization algorithm, and determine the optimal network structure and parameter settings; Step 5: Input the test set data into the optimized CNN-ELM model, obtain the prediction results and perform inverse normalization, and then use the evaluation function to analyze and evaluate the prediction results to verify the prediction performance of the model under different working conditions. Step 6: Input the real-time collected boiler operation data into the trained CNN-ELM model and output the dynamic thermal efficiency prediction value.

2. The method according to claim 1, characterized in that: In step 1, the collected historical operating data includes thermal efficiency and related auxiliary variables.

3. The method according to claim 1, characterized in that: In step 2, the missing data processing steps are as follows: 3σ-rule is used to detect outlier data, and outlier data is replaced based on the following formula: △x i (t)=x i (t)-x i (t-1) Where, △x i (t) represents the auxiliary variable x i The rate of change between two adjacent time points Indicate △x i (t) represents the average rate of change before time t, σ i (t) represents the standard deviation, and t = 2, ..., n represents the sampling time.

4. The method according to claim 1, characterized in that: In step 2, the data dimensionality reduction step is as follows: the auxiliary variables are ranked by importance using the random forest algorithm, and the top 10 most important variables are selected as input variables.

5. The method according to claim 1, characterized in that: In step 2, the data reconstruction steps are as follows: the time series data is converted into a two-dimensional tensor using the sliding window method, wherein the time dimension contains 20 consecutive sampling times and the feature dimension consists of 10 filtered variables.

6. The method according to claim 1, characterized in that: In step 3, the CNN network includes at least two convolutional layers, each followed by a channel equalization block (CE-Block) to suppress the distribution differences of feature values ​​between channels.

7. The method according to claim 1, characterized in that: In step 4, the hyperparameters of the CNN-ELM model are set as follows: batch size 32, optimizer Adam, and activation function ReLU.

8. A system for implementing the method of any one of claims 1-7, characterized in that, include: The data acquisition module acquires real-time operating data from the boiler's DCS system. The preprocessing module performs missing data processing, data dimensionality reduction, and data reconstruction. The model computation module deploys the CNN-ELM model and performs prediction calculations. The visualization terminal displays real-time boiler thermal efficiency prediction results.

9. A storage medium, characterized in that: The storage medium stores a computer program, which, when executed by a processor, implements the CNN-ELM-based soft measurement method for boiler efficiency as described in any one of claims 1-7.

10. The storage medium according to claim 9, characterized in that: The storage medium is integrated into the DCS system or edge computing terminal of a coal-fired power plant to receive boiler operation data in real time and output thermal efficiency prediction results.

Images