A boiler superheater wall temperature prediction method, system, device and medium

Abstract

The application discloses a boiler superheater wall temperature prediction method, system, device and medium, and belongs to the technical field of safe operation of a boiler heating surface, and comprises the following steps: collecting historical operation data and preprocessing; screening key variables with a cumulative contribution rate of more than or equal to 95% through a random forest; constructing a BCNN-BAN fusion model, wherein a RoSwish activation function is used in the BCNN to extract multi-scale features, and a double linear attention is used in the BAN to focus on key features; using an improved aurora optimization algorithm to optimize hyperparameters; and inputting real-time data to output a prediction value. The application solves the problems of weak nonlinear fitting capability and poor generalization of a traditional model, can effectively realize over-temperature early warning, provides technical support for safe operation of a thermal power unit under flexible peak shaving, and has remarkable engineering application value.

Description

TECHNICAL FIELD

[0001] The present application belongs to the technical field of safe operation of boiler heating surface, and particularly relates to a boiler superheater wall temperature prediction method, system, device and medium. BACKGROUND

[0002] At present, flexible peak regulation of thermal power units has become an important support for power system operation. Frequent low-load operation and start-stop operation make high-temperature components such as superheaters long-term work in severe fluctuation conditions, which leads to a significant increase in local over-temperature and pipe wall damage risk. Superheater wall temperature anomaly is one of the main causes of boiler tube burst accidents, and its monitoring and prediction are directly related to the safe and stable operation of the unit. In particular, in ultra-supercritical boilers, the continuous rise of superheater wall temperature not only causes metal material creep and oxide scale peeling, but also more likely causes systematic safety hazards such as SCR catalyst poisoning. Therefore, it is of great significance to build an efficient, accurate and deployable superheater wall temperature prediction model for realizing over-temperature early warning, optimizing operation control and prolonging equipment life.

[0003] At present, the research on superheater wall temperature prediction mainly focuses on two directions: one is the numerical simulation method based on physical mechanism, which usually adopts computational fluid dynamics (CFD) model to jointly model the heat transfer process of the flue gas side and the steam side inside the boiler to realize high-precision spatial prediction of wall temperature distribution. The second is the soft measurement method based on machine learning, which establishes a nonlinear mapping relationship between input variables and wall temperature through historical data to realize rapid prediction and dynamic monitoring of wall temperature. The first method has good physical interpretability and working condition adaptability, and can reveal the coupling relationship between local heat transfer process and structural response, but usually relies on a large number of boundary conditions and structural parameter settings, which is complex in modeling and consumes a lot of computing resources, and is difficult to meet the demand for rapid response and real-time analysis in actual operation.

[0004] In recent years, with the rapid development of artificial intelligence technology, data-driven modeling methods have gradually become a research hotspot in the field of superheater wall temperature prediction. Such methods have shown great advantages in improving prediction efficiency and reducing computational cost. Existing research has introduced various neural network structures, including convolutional neural networks (CNN), long short-term memory networks (LSTM), gated recurrent units (GRU), and nonlinear autoregressive models (NARX), to enhance the modeling ability of temperature time series features and improve prediction accuracy and model generalization ability. Among them, CNN and temporal convolutional networks (TCN) have been widely used in temperature sequence modeling tasks due to their outstanding performance in extracting local patterns and short-term dependency features. For example, there is a multi-step wall temperature prediction model based on the encoder-decoder architecture in the prior art, which significantly improves the model's trend capturing ability and prediction accuracy; there is a multi-scale time series modeling framework that integrates TCN, CNN, and Transformer in the prior art, which exhibits excellent stability and accuracy under complex fluctuating conditions. To further improve the model's feature selection ability and generalization performance, the gray correlation analysis method is introduced in the prior art to screen input variables, and a wall temperature time series prediction model is constructed combined with LSTM. There is a genetic algorithm optimized deep neural network (GA-DNN) model in the prior art that can identify and warn the outlet temperature of the screen-type superheater and the over-temperature region, and can accurately locate the over-temperature risk tube position 5 minutes in advance.

[0005] Although the current convolutional neural network has shown significant advantages in modeling ability and nonlinear feature learning, the layer-by-layer structure design adopted by mainstream methods often leads to limited effective receptive field and weak cross-level information fusion mechanism. In particular, in the high-temperature superheater wall temperature modeling scenario with strong nonlinear coupling and dynamic time delay characteristics, the original method still has obvious deficiencies in feature space redundancy suppression, robustness of multi-scale feature expression, and generalization performance. SUMMARY

[0006] To solve the above technical problems, the present application provides a boiler superheater wall temperature prediction method, system, device and medium, by constructing a fusion model based on Broad Convolutional Neural Networks (BCNN) and Bilinear Attention Network (BAN), combining RoSwish activation function, and improving PLO hyperparameter optimization, a multidimensional technical breakthrough is achieved in the boiler superheater wall temperature prediction scene, realizing accurate prediction of the boiler superheater wall temperature, and providing technical support for the safe operation of thermal power generating units. Through experimental verification, the present application performs excellently in prediction accuracy, generalization and engineering adaptability, providing an innovative reference for boiler heating surface wall temperature prediction research and strong support for the safe operation of thermal power generating units.

[0007] To achieve the above object, the present application adopts the following technical solutions:

[0008] A boiler superheater wall temperature prediction method, comprising the following steps:

[0009] Step 1, collecting the historical operation data of the boiler from the unit decentralized control system, and preprocessing the historical operation data to construct a standardized data set;

[0010] Step 2, using a random forest algorithm to reduce the dimension of the original variables affecting the boiler superheater wall temperature, selecting the first m variables as the model input, and dividing the standardized data set into a training set and a test set;

[0011] Step 3, constructing a boiler superheater wall temperature prediction model based on BCNN-BAN, which is composed of a Broad Convolutional Neural Network in series with a Bilinear Attention Network, and the activation function of BCNN uses a rotating Swish function;

[0012] Step 4, training the boiler superheater wall temperature prediction model based on BCNN-BAN using the data of the training set, and optimizing the model hyperparameters using an improved Aurora optimization algorithm to obtain an optimized boiler superheater wall temperature prediction model based on BCNN-BAN;

[0013] Step 5, inputting the data of the test set into the optimized boiler superheater wall temperature prediction model based on BCNN-BAN, obtaining the prediction results and performing reverse normalization processing, then using an evaluation function to analyze and evaluate the model prediction performance, and verifying the prediction performance under different operating conditions;

[0014] Step 6, inputting the real-time collected coal-fired power generating unit operating parameters into the optimized boiler superheater wall temperature prediction model based on BCNN-BAN, outputting the boiler superheater wall temperature prediction value, and realizing early warning of the boiler superheater over-temperature risk.

[0015] Further, in step 1, the collected historical operation data includes the boiler superheater wall temperature and auxiliary variables; the pre-processing of the historical operation data includes: filling in the missing values and abnormal values according to the time label using linear interpolation, and then performing normalization processing.

[0016] Further, in step 2, the first m variables with a cumulative contribution rate greater than or equal to 95% are selected as the model input, and m is the number of characteristic variables corresponding to the cumulative contribution rate greater than or equal to 95%, which is determined according to the actual operation data characteristics of the unit.

[0017] Further, in step 3, the boiler superheater wall temperature prediction model based on BCNN-BAN includes an input layer, a BCNN layer, a BAN layer and an output layer.

[0018] Further, in step 3, the input layer is used to receive the m-dimensional model input after dimensionality reduction by random forest; the BCNN layer adopts a multi-branch parallel convolution structure, and local feature extraction is performed on the input data through K parallel convolution branches; the activation function adopted by the BCNN layer adopts a rotating Swish function, which is used to dynamically adjust the rotation angle and offset of the function to adapt to the nonlinear feature distribution of the boiler superheater operation data; the BCNN layer splices the feature maps of the K branches in the channel dimension to obtain the final features of the BCNN layer.

[0019] Further, in step 3, the BAN layer receives the final features of the BCNN layer, obtains two groups of features through two linear transformations, and then performs bilinear pooling fusion; the attention mechanism is used to assign weights to each dimension of the fused features; the output layer maps the attention features to the predicted value of the boiler superheater wall temperature through a fully connected layer.

[0020] Further, in step 4, the model hyperparameters include the number of neurons and the learning rate, the optimization range of the number of neurons is [32, 64, 128, 256], and the optimization range of the learning rate is [0.001, 0.0001, 0.00001, 0.000001].

[0021] The application also provides a boiler superheater wall temperature prediction system for implementing the above method, comprising:

[0022] A data acquisition module is configured to acquire historical operation data of a unit distributed control system.

[0023] A pre-processing module is configured to perform missing data filling, abnormal value processing, data dimensionality reduction and normalization processing on the historical operation data.

[0024] A model operation module is configured to deploy the boiler superheater wall temperature prediction model based on the BCNN-BAN and perform prediction calculation to obtain a boiler superheater wall temperature prediction result.

[0025] A visual terminal is configured to display the boiler superheater wall temperature prediction result online.

[0026] The application further provides an electronic device including a processor and a memory storing a computer program, wherein the processor implements the steps of the boiler superheater wall temperature prediction method when executing the program.

[0027] The application further provides a computer readable storage medium storing a computer program, wherein the computer program implements the steps of the above boiler superheater wall temperature prediction method when executed by a processor.

[0028] Advantages:

[0029] (1) The application adopts the BCNN-BAN fusion architecture, breaks through the limitations of the traditional CNN receptive field, single attention mechanism, and insufficient feature extraction breadth, and the BCNN realizes the breadth extraction of global multi-scale features through multi-branch parallel convolution, the BAN excavates the high-order correlation of features through bilinear pooling and combines the attention mechanism to accurately focus on key features, and the two work together to realize the feature extraction effect of breadth coverage and accurate screening. Using only BCNN will cause core influencing factors to be submerged due to feature redundancy; using only BAN is limited by the receptive field and is difficult to cover global features. When the two work together, the BCNN is responsible for widely extracting full-dimensional working condition features, and the BAN accurately focuses on the key features related to wall temperature through bilinear attention, which not only preserves the integrity of global features, but also strengthens the weight of core features, solving the problem of incomplete or redundant feature extraction of traditional models under complex working conditions.

[0030] (2) The application proposes a RoSwish activation function with learnable parameters , which can adaptively adapt to the strong nonlinearity of boiler operation data compared with ReLU and traditional Swish, and improve the fitting ability of the model to complex working condition data. The dynamic width of the BCNN (number of network branches / number of channels) can adaptively adjust the breadth of feature extraction according to the change of the boiler working condition; and the learnable parameters , of the RoSwish activation function can dynamically adjust the amplitude of feature response. The two form a two-dimensional amplitude-channel adaptive mechanism, so that the model can always accurately fit the change of the wall temperature under dynamic working conditions such as unit peak shaving and fuel switching, and the nonlinearity expression ability is improved by more than 20% compared with ReLU and traditional Swish activation functions.

[0031] (3) The application solves the problems of early maturity convergence of the traditional genetic algorithm and low high-dimensional optimization precision of the Bayesian optimization through the aurora ellipse wandering value updating, particle collision mechanism, adaptive weight and other improvement strategies, and in the scene of collaborative optimization of multiple super parameters (neuron number, learning rate, convolution kernel size, branch number, etc.) of the BCNN-BAN fusion model, the convergence speed is improved by more than 30%, the global optimality of the model is significantly improved, the global convergence ability is improved, and the convergence efficiency is ensured. Finally, the optimal super parameter combination selected for the BCNN-BAN model can stably control the prediction error (MAE) to be minimum, which is much better than the effect of the traditional optimization algorithm.

[0032] (4) The application integrates the integrated process of random forest feature dimension reduction and BCNN-BAN fusion model precise prediction, first selects the key variables with a cumulative contribution rate of more than 95% through the random forest, and then inputs the BCNN-BAN fusion model to realize wall temperature prediction, which guarantees the effectiveness of the features and improves the engineering practicability.

[0033] In summary, the application realizes breakthroughs in prediction accuracy, working condition adaptability and engineering practicability through the synergistic innovation of multiple technologies, provides a new technical scheme for boiler heating surface wall temperature prediction, and provides strong support for the safe and efficient operation of the thermal power unit in the flexible peak regulation scene. BRIEF DESCRIPTION OF DRAWINGS

[0034] Figure 1 It is a flowchart of the boiler superheater wall temperature prediction method of the application;

[0035] Figure 2 It is the change of the boiler load and the superheater wall temperature in the selected operation data in the embodiment;

[0036] Figure 3 It is the random forest dimension reduction feature contribution degree in the embodiment;

[0037] Figure 4 It is a super parameter optimization flowchart of the improved aurora optimization algorithm used in the embodiment;

[0038] Figure 5 It is a prediction result graph of the verification set before the super parameter optimization in the embodiment;

[0039] Figure 6 It is a prediction result graph of the verification set after the super parameter optimization in the embodiment;

[0040] Figure 7 It is the BCNN-BAN prediction result of the super parameter optimization in the embodiment;

[0041] Figure 8The prediction results of the boiler superheater wall temperature prediction model based on BCNN-BAN in the embodiment on the validation set;

[0042] Figure 9 The prediction results of the boiler superheater wall temperature prediction model based on BCNN in the embodiment on the validation set;

[0043] Figure 10 The prediction results of the boiler superheater wall temperature prediction model based on BAN in the embodiment on the validation set;

[0044] Figure 11 The figure of the change of the model training loss function with the number of iterations in the embodiment. DETAILED DESCRIPTION

[0045] In order to make the objects, technical solutions and advantages of the present application clearer, the present application will be further described in detail below with reference to the drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and do not limit the present application. In addition, the technical features involved in each embodiment of the present application described below can be combined with each other as long as they do not conflict with each other.

[0046] As shown in Figure 1 The present application provides a boiler superheater wall temperature prediction method, comprising the following steps:

[0047] Step 1: Collecting the historical operation data of the boiler from the unit distributed control system, pre-processing the historical operation data, and constructing a standardized data set;

[0048] Step 2: Using the random forest algorithm to reduce the dimension of the original variables affecting the boiler superheater wall temperature, determining the characteristic value, selecting the first m variables as the model input, and finally dividing the standardized data set into a training set and a test set; Step 3: Constructing a boiler superheater wall temperature prediction model based on BCNN-BAN (four-layer network: input layer, BCNN layer, BAN layer, output layer connected in order); which is composed of a wide convolutional neural network connected in series with a bilinear attention network, wherein the activation function of BCNN adopts a rotating Swish function; Step 4: Training the boiler superheater wall temperature prediction model based on BCNN-BAN using the data of the training set, and optimizing the model hyperparameters through the improved aurora optimization algorithm to obtain the optimized boiler superheater wall temperature prediction model based on BCNN-BAN;

[0049] Step 5, input the data of the test set into the optimized BCNN-BAN-based boiler superheater wall temperature prediction model, obtain the prediction results and perform reverse normalization processing, then analyze and evaluate the prediction performance of the model using the evaluation function, and verify the prediction performance of the model under different working conditions;

[0050] Step 6, input the real-time collected operation parameters of the coal-fired power generating unit into the optimized BCNN-BAN-based boiler superheater wall temperature prediction model, output the predicted value of the boiler superheater wall temperature, and realize the early warning of the over-temperature risk of the boiler superheater.

[0051] Preferably, in step 1, the pre-processing of the historical operation data of the unit distributed control system includes: firstly, filling in the missing values and abnormal values in the historical operation data using linear interpolation according to the time label, and then performing normalization processing.

[0052] Preferably, in step 2, the first m variables with a cumulative contribution rate greater than or equal to 95% are selected as the model input (m is the number of characteristic variables corresponding to the cumulative contribution rate greater than or equal to 95%, which is determined according to the characteristics of the actual operation data of the unit).

[0053] Preferably, in step 3, the BCNN-BAN-based boiler superheater wall temperature prediction model constructed includes an input layer, a BCNN layer, a BAN layer and an output layer, and the activation function adopts a rotating Swish function (RoSwish).

[0054] Preferably, in step 4, the model hyperparameters include the number of neurons and the learning rate, and the hyperparameter optimization of the neural network uses an intelligent optimization algorithm (improved aurora optimization algorithm PLO).

[0055] Preferably, in step 4, the optimization range of the number of neurons is [32, 64, 128, 256], the optimization range of the learning rate is [0.001, 0.0001, 0.00001, 0.000001], and finally the number of neurons is determined as 128 and the learning rate is determined as 0.0001.

[0056] Embodiment

[0057] The embodiment of the application realizes a BCNN-BAN-based boiler superheater wall temperature prediction method, which includes the following steps:

[0058] Step 1: collect historical operation data from a unit distributed control system (DCS) and perform pre-processing:

[0059] (1) Collecting data. Taking the high-temperature superheater of a 1000 MW ultra-supercritical once-reheat generator unit as the research object, the historical data of the unit from 21:17:00 on May 22, 2025 to 19:57:00 on May 29 were extracted from the unit distributed control system (DCS), a total of 10,000 groups of samples, with a sampling time interval of 60 seconds. Through in-depth analysis of the historical operation data, it was found that at the wall temperature No. 2 measuring point position of the high-temperature superheater, the wall temperature showed a high phenomenon regardless of which load stage it was in. Therefore, the wall temperature at this measuring point position was selected as the object for analysis.

[0060] The load and superheater wall temperature change in the selected data are shown in Figure 2 , the load change range is [295 MW, 1002 MW], and the wall temperature change range is [581℃, 628℃], so it can be seen that the selected data covers various complex working conditions, and the superheater wall temperature is over-temperature.

[0061] (2) Data preprocessing. Due to the complex production site environment of thermal power generator units and the influence of factors such as noise interference on the collected signals, there are inevitably missing values and outliers in a large amount of raw data obtained. In this embodiment, 3σ-rule is used to detect abnormal data of the boiler operation data, and the abnormal data is replaced based on the following formula:

[0062] ;

[0063] ;

[0064] ;

[0065] ;

[0066] wherein, represents an auxiliary variable the change rate at adjacent time t and time t-1, represents the average change rate before time t, represents the standard deviation at time t, t = 2, …, n is the sampling time, and n represents the total sampling time. represents the time index value.

[0067] The data with replaced abnormal data is linearly normalized by the following formula to standardize the data size:

[0068] ;

[0069] wherein, , respectively represent the minimum value and the maximum value of the i-th variable, The raw data is represented, The normalized results are shown in Table 1.

[0070] Step 2: Data dimensionality reduction based on random forest and division of data set:

[0071] In combination with historical data, 106 main parameters such as unit load, total fuel quantity, total air quantity, feedwater flow, secondary air flow, main steam pressure, main steam flow, and main steam temperature, which affect the superheater wall temperature, were analyzed. The random forest (RF) algorithm was used to calculate the score of feature importance, and the key feature variables were selected according to the score, thereby reducing the input dimension. The random forest analysis results of the candidate variables are shown in FIG. 1. Figure 3 Figure 3 The middle ordinate shows the importance (or contribution rate) of each feature, and the curve on the bottom plane shows the cumulative importance of different feature quantities. According to the contribution rate curve, the cumulative importance of the first 67 variables is 95%, which basically covers the main information of the original data. According to the mechanism analysis, five input variables, “Reheat steam temperature B side temperature ℃”, “Reheater outlet temperature ℃”, “Reheater pressure MPa”, “Reheat steam temperature A side temperature ℃”, and “Coal feeder E (t / h)”, were removed. Finally, the first m = 62 variables were selected as the input variables of the model, as shown in Table 1.

[0072] Table 1: Selected 62 variables and their contribution

[0073]

[0074] The preprocessed sub-sequence data set was divided into a training set of 7976 groups and a test set of 1994 groups according to a ratio of 8:2.

[0075] Step 3: Construction of a boiler superheater wall temperature prediction model based on BCNN-BAN.

[0076] The boiler superheater wall temperature prediction model based on BCNN-BAN includes an input layer, a BCNN layer, a BAN layer, and an output layer, and the activation function uses a rotated Swish function (RoSwish).

[0077] (1) Input layer: receiving 62-dimensional input feature values after dimensionality reduction in step 2, the data is represented in matrix form as:

[0078] ;

[0079] In the formula, represents the input original monitoring data matrix; H is the height of the input data, corresponding to the time sequence length or spatial row number of the monitoring data; W is the width of the input data, corresponding to the feature dimension or spatial column number of the monitoring data; and C is the channel number of the input data, corresponding to different types of monitoring parameters. ​

[0080] (2) BCNN layer:

[0081] In view of the deficiencies of the width learning system (BLS) and the one-dimensional convolution (1D-CNN), the present application proposes a broad convolutional neural network (BCNN). Unlike the traditional layer-by-layer convolution structure, the algorithm improves the network structure by drawing on the "width" idea in the width learning system (BLS), adopts parallel extraction of feature nodes and enhanced nodes, and then adopts a fusion mechanism to improve the feature learning ability of the model. Among them, the feature extraction adopts one-dimensional convolution operation with adjustable convolution kernel size, and this flexible feature extraction mode enhances the adaptability of the model to multi-scale features.

[0082] (2.1) Multi-branch convolution layer. The BCNN adopts a multi-branch parallel convolution structure, and local feature extraction is performed on the input data through K parallel convolution branches. The convolution operation of each branch is as follows:

[0083] ;

[0084] In the formula, represents the output feature map of the kth convolution branch, , K represents the total number of convolution branches is a convolution operation, including a local window sliding and a weighted summation process; is the convolution kernel parameter of the kth branch, represents a real set, and are the height and width of the kth convolution kernel, respectively, and C is the input channel number, is the output channel number of the kth branch; is the bias term of the kth branch; is a residual connection, which is used to alleviate the gradient vanishing problem of a deep network and enhance feature transmission.

[0085] (2.2) In the multi-branch convolution layer of the BCNN, the activation function adopts a rotating Swish function (RoSwish), and its expression is as follows:

[0086] ;

[0087] In the formula, is the feature value after convolution operation; is a Sigmoid function, and are learnable parameters, which are used to dynamically adjust the rotation angle and offset of the function to adapt to the nonlinear feature distribution of the boiler superheater operation data. The activation function introduces learnable parameters and Compared with the traditional Swish function, the proposed function has stronger adaptability and can better capture the nonlinear variation of the wall temperature of the superheater, thus improving the feature extraction ability of the BCNN for monitoring data under complex working conditions.

[0088] (2.3) Feature fusion and output. The feature maps of the K branches are spliced in the channel dimension to obtain the final feature output of the BCNN:

[0089] ;

[0090] wherein, is the fusion feature output by the BCNN; is the channel dimension splicing operation; and are the height and width of the feature map after fusion, respectively, which are determined by the convolution step, padding, and other parameters.

[0091] (3) BAN layer:

[0092] BAN is a deep learning architecture that fuses feature interaction and focuses on key influencing factors through bilinear pooling and attention mechanism.

[0093] (3.1) Bilinear feature fusion. BAN receives the feature output by the BCNN, obtains two groups of features through two linear transformations, and then performs bilinear pooling fusion:

[0094] ;

[0095] wherein, denotes bilinear fusion, is the first group of linear transformation features; is the second group of linear transformation features; and are the weight matrices of linear transformation; and are the bias terms of linear transformation; is the bilinear pooling operation, which can use element multiplication or outer product, etc., to enhance the interaction between features; is the feature vector after bilinear fusion.

[0096] (3.2) Attention weight calculation

[0097] The attention mechanism assigns weights to each dimension of the feature vector after bilinear fusion, highlighting the features that play a key role in predicting the wall temperature of the boiler superheater:

[0098] ;

[0099] ;

[0100] where, is the i-th dimension of the bilinear fused feature vector ; , D is the dimension of the bilinear fused feature vector ; is the j-th dimension of the bilinear fused feature vector , j ; is the scoring function, which measures the importance of the feature through linear transformation, where is the weight vector of the scoring function, is the bias term of the scoring function, is the scoring function; is the attention weight of the i-th feature dimension, which is normalized by the softmax function to ensure that the sum of the weights is 1; is the weighted attention feature, which focuses on key features, and exp(·) represents the exponential function.

[0101] (4) Output layer:

[0102] The attention feature is mapped to the predicted value of the boiler superheater wall temperature through a fully connected layer:

[0103] ;

[0104] where, is the predicted value of the boiler superheater wall temperature; is the weight matrix of the output layer; is the bias term of the output layer.

[0105] Step 4, train the boiler superheater wall temperature prediction model based on BCNN-BAN using the data of the training set, and optimize the model hyperparameters through the improved aurora optimization algorithm.

[0106] All experiments are carried out under the PyTorch deep learning framework, with an AMD Ryzen98945HS CPU processor, 32GB of memory, 8GB of GPU memory, and a Windows 11 operating system. The 7976 group data has been selected as the training set in step 2 above. In order to better evaluate the performance of the model, the present application uses the mean absolute percentage error MAPE, the root mean square error RMSE, the mean absolute error MAE and the determination coefficient R 2 Evaluate the performance of the model:

[0107]

[0108] where, , 、 respectively represent the actual measured value, the predicted value, the mean value of the actual measured value, n represents the total number of samples, and t represents the index value.

[0109] To further improve the prediction accuracy and fitting effect of the network model, the hyperparameters of the model are optimized. The improved polar light optimization algorithm is used to optimize the hyperparameters of the model. The polar light optimization algorithm (Polar Lights Optimization, PLO) is a meta-heuristic optimization algorithm inspired by the natural polar light phenomenon, which realizes global optimization by simulating the motion characteristics, energy conversion and spatial distribution law of polar light particles. Figure 4 The hyperparameter optimization flowchart of the improved polar light optimization algorithm used in the embodiment is shown in Table 1, and the specific steps are as follows: ① Initialize the time, the maximum number of iterations (set to 200), the particle swarm (particles correspond to hyperparameter combinations, and the number of particles is set to 30), calculate the fitness of each particle (based on the weighted value of MAE+MAPE of the model) and sort; ② Update the gyration motion speed, polar light ellipse wandering value and adaptive weight (the optimal weight range is 0.4~0.9) of each particle; ③Judge whether the particle collision mechanism is triggered (triggered when the particle fitness is continuously improved for 5 generations), if yes, then regenerate part of the particles to jump out of the local optimum; ④ Use the greedy selection mechanism to iterate the particles, and retain 50% of the particles with the best fitness; ⑤ Calculate the fitness of the new particle swarm, and update the global optimal particle; ⑥Judge whether the end condition is met (the maximum number of iterations is reached or the optimal fitness is ≤1.5℃), if yes, then output the optimal particle (corresponding to the optimal hyperparameter), otherwise return to ② for continuous iteration.

[0110] The comparison of model prediction results based on improved polar light optimization algorithm parameter optimization is shown in FIG. 1 (using the data of the test set). Before the hyperparameter optimization ( Figures 5-6 ), the model shows obvious deviation in the sample interval corresponding to the two subgraphs, and the tracking delay of the dynamic trend of the wall temperature is about 2~3 minutes; the peak capture ability of the local wall temperature change is poor, and the difference between the predicted value and the measured value is 1%~2%. After the hyperparameter optimization by the improved polar light optimization algorithm ( Figure 5 ), the prediction performance of the model on the validation set is significantly improved, and the tracking delay of the dynamic trend is less than 1 minute; the peak capture ability of the wall temperature change is greatly improved, and the deviation between the predicted value and the measured value is less than 1%. Figure 6

[0111] The iteration process of hyperparameter optimization is shown in Table 2, and one optimal record is selected every five iterations for display, and a total of 20 data are counted. After selecting the optimal hyperparameters, the final prediction effect of the model on the training set and the validation set is shown in FIG. 2. Figure 7 ​​

[0112] From Table 2, after optimization by the improved aurora optimization algorithm, the selection of optimal hyperparameters and its results are as follows: lr=0.0522, N1=189, N2=N 1= 189, Weight_Decay=1e-05, MAE=1.691, MAPE=0.277, RMSE=2.117, R 2 =0.883, and the training time is 25.11 s. Among them, Lr (Learning Rate) is the model learning rate, which controls the parameter update step; Weight_Decay is the weight decay coefficient, which is used to prevent model overfitting; N1 is the number of neurons in the first stage of the model; N2 is the number of neurons in the second stage of the model.

[0113] Table 2 Hyperparameters in the iteration process of the improved aurora optimization algorithm

[0114]

[0115] For other parameters, when the number of neurons is 189, the prediction effect is the best. The learning rate is optimized using the Adam optimization algorithm, and when the learning rate is 0.0001, the prediction effect is the best. Figure 7 When the wall temperature of the superheater changes, the boiler superheater wall temperature prediction model based on BCNN-BAN can accurately predict the change trend. The model prediction result can predict the turning point of the measured value change of the superheater wall temperature, and can remain stable in the period of relatively stable temperature, and is in good agreement with the actual measured value.

[0116] Step 5: input the test set data into the optimized boiler superheater wall temperature prediction model based on BCNN-BAN, obtain the prediction result and perform inverse normalization processing, and analyze and evaluate the prediction performance of the model by using the evaluation function.

[0117] In order to verify the optimal model proposed in the embodiment, the boiler superheater wall temperature prediction model based on BCNN-BAN is compared with the BCNN and BAN models. In order to reduce the influence of random factors on the prediction results of the above various models, the same historical data set is used for multiple repeated experiments for each neural network model. The prediction results of different models are compared as shown in Figures 8-10 Figure 8 The prediction result of the boiler superheater wall temperature prediction model based on BCNN-BAN on the validation set is shown in the table. Figure 9 The prediction result of the boiler superheater wall temperature prediction model based on BCNN in the embodiment on the validation set is shown in the table. Figure 10 The prediction result of the boiler superheater wall temperature prediction model based on BAN in the embodiment on the validation set is shown in the table. Figures 8-10 ​The middle blue curve represents the measured value of the superheater wall temperature, the red curve represents the predicted value of the superheater wall temperature, and the green columnar mark represents the error corresponding to the measured value and the predicted value at each sample point. The quantitative evaluation index comparison is shown in Table 3. As can be seen from Table 3, the error indicators (RMSE=2.117 ℃, MAE=1.691 ℃, MAPE=0.277 %) of the boiler superheater wall temperature prediction model based on BCNN-BAN are all the smallest, and the determination coefficient (R 2 =0.883) is closest to 1, indicating that the prediction performance is optimal.

[0118] From Figures 8-10 and the comparison results in Table 3, it can be seen that the proposed BCNN-BAN model performs best in terms of prediction accuracy. BCNN can capture multi-dimensional and multi-scale original features in the input data through the width convolution structure, solving the limitations of single network in feature coverage breadth; and the bilinear attention mechanism of BAN can accurately locate the key features that most significantly affect the wall temperature through bilinear pooling and weight allocation of features, and suppress the interference of irrelevant or noise features, making up for the shortcomings of single width convolution in feature importance distinction. The combination of the two is not a simple structure stacking, but through the “wide coverage” of BCNN to provide a rich feature base, and then through the “precise selection” of BAN to strengthen the contribution of key information, so that the model can not only capture potential influencing factors comprehensively, but also focus on core driving variables when dealing with the complex scene of boiler superheater multivariate strong coupling and dynamic change of working conditions, thereby being superior to single BCNN or BAN model in terms of prediction accuracy and adaptability to variable working conditions.

[0119] Table 3 Comparison of performance indicators of BCNN-BAN and benchmark model prediction

[0120]

[0121] In addition, Figure 11 The figure showing the change of the model training loss function with the number of iterations (normalized results) in the embodiment is shown. The boiler superheater wall temperature prediction model based on BCNN-BAN gradually stabilizes and converges after about 20 iterations. Therefore, compared with the baseline prediction model, the boiler superheater wall temperature prediction model based on BCNN-BAN proposed in the present application not only has more accurate prediction effect, but also shows better dynamic performance in actual boiler heating surface wall temperature prediction.

[0122] Step 6, input the real-time collected generator set operation parameters into the trained boiler superheater wall temperature prediction model based on BCNN-BAN, and output the predicted value of the superheater wall temperature in the future time.

[0123] The present application also provides a boiler superheater wall temperature prediction system for realizing the above-mentioned method, comprising the following modules:

[0124] a data collection module configured to acquire historical operation data of the unit decentralized control system;

[0125] a preprocessing module configured to perform missing data filling, outlier processing, data dimension reduction and normalization processing on the historical operation data;

[0126] a model operation module configured to deploy a boiler superheater wall temperature prediction model based on BCNN-BAN, and perform prediction calculation to obtain a boiler superheater wall temperature prediction result;

[0127] a visual terminal configured to display the boiler superheater wall temperature prediction result online.

[0128] The application further provides an electronic device comprising a processor and a memory, wherein the memory stores a computer program, and the processor implements the steps of the boiler superheater wall temperature prediction method when executing the program.

[0129] The application further provides a computer readable storage medium, wherein the storage medium stores a computer program, and the computer program implements the boiler superheater wall temperature prediction method when executed by a processor.

[0130] Those skilled in the art should understand that the embodiments of the application can be provided as a method, a system or a computer program product. Therefore, the application can be in the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Moreover, the application can be in the form of a computer program product implemented on one or more computer usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer usable program code. The solutions in the embodiments of the application can be implemented in various computer languages, such as object-oriented programming language Java and interpreted scripting language JavaScript.

[0131] The application is described with reference to flowcharts and / or block diagrams according to the embodiments of the application. It should be understood that each flow and / or block in the flowcharts and / or block diagrams, and the combination of the flows and / or blocks in the flowcharts and / or block diagrams can be implemented by computer program instructions. These computer program instructions can be provided to a general-purpose computer, a special-purpose computer, an embedded processor or other programmable data processing devices to produce a machine, so that the instructions executed by the computer or other programmable data processing devices produce a device that implements the functions specified in the flowcharts and / or block diagrams. Figure 1 The device that implements the functions specified in one flow or multiple flows and / or blocks Figure 1 The device that implements the functions specified in one flow or multiple flows and / or blocks

[0132] These computer program instructions can also be stored in a computer- readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instructions which implement the Figure 1 function specified in the flow or flows and / or blocks Figure 1 of the block or blocks.

[0133] The computer program instructions can also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions that are executed on the computer or other programmable apparatus provide steps for implementing the Figure 1 function specified in the flow or flows and / or blocks Figure 1 of the block or blocks.

[0134] While the preferred embodiments of the application have been described, additional variations and modifications can be made to the embodiments by those skilled in the art once they learn of the basic inventive concepts. Therefore, the appended claims are intended to encompass within their scope all such variations and modifications as are included within the scope of the application.

[0135] Obviously, numerous modifications and variations of the present application are possible in light of the above teachings. It is therefore to be understood that within the scope of the appended claims and their equivalents, the application can be practiced otherwise than as specifically described.

Claims

1. A method for predicting the wall temperature of a boiler superheater, characterized by, The method comprises the following steps: Step 1, collecting historical operation data of the boiler from a unit decentralized control system, and preprocessing the historical operation data to construct a standardized data set; Step 2, using a random forest algorithm to reduce the dimension of original variables affecting the wall temperature of the boiler superheater, selecting the first m variables as the model input, and dividing the standardized data set into a training set and a test set; Step 3, constructing a boiler superheater wall temperature prediction model based on BCNN-BAN, which is composed of a wide convolutional neural network in series with a bilinear attention network, and the activation function of the BCNN is a rotating Swish function; Step 4, training the boiler superheater wall temperature prediction model based on BCNN-BAN using the data of the training set, and optimizing the model hyperparameters through an improved aurora optimization algorithm to obtain an optimized boiler superheater wall temperature prediction model based on BCNN-BAN; Step 5, inputting the data of the test set into the optimized boiler superheater wall temperature prediction model based on BCNN-BAN, obtaining the prediction results and performing reverse normalization processing, then analyzing and evaluating the prediction performance of the model using an evaluation function, and verifying the prediction performance of the model under different working conditions; Step 6, inputting the real-time collected operation parameters of the coal-fired power generating unit into the optimized boiler superheater wall temperature prediction model based on BCNN-BAN, outputting the predicted value of the boiler superheater wall temperature, and realizing early warning of the over-temperature risk of the boiler superheater.

2. The boiler superheater wall temperature prediction method of claim 1, wherein, In step 1, the historical operation data includes the wall temperature of the boiler superheater and auxiliary variables; the preprocessing of the historical operation data includes: filling the missing values and abnormal values using linear interpolation according to the time label, and then performing normalization processing.

3. The method of claim 1, wherein, In step 2, the first m variables with a cumulative contribution rate greater than or equal to 95% are selected as the model input.

4. The method of claim 1, wherein, In step 3, the boiler superheater wall temperature prediction model based on BCNN-BAN comprises an input layer, a BCNN layer, a BAN layer and an output layer.

5. The method of claim 4, wherein, The input layer is used to receive the m-dimensional model input after random forest dimension reduction; the BCNN layer adopts a multi-branch parallel convolution structure, and locally extracts features of the input data through K parallel convolution branches; the rotating Swish function adopted by the BCNN layer is used to dynamically adjust the rotation angle and offset of the function to adapt to the nonlinear feature distribution of the boiler superheater operation data; the BCNN layer splices the feature maps of the K branches in the channel dimension to obtain the final features of the BCNN layer.

6. The boiler superheater wall temperature prediction method of claim 5, wherein, The BAN layer receives the final features of the BCNN layer, obtains two groups of features through two linear transformations, and then performs bilinear pooling fusion; the attention mechanism is used to assign weights to each dimension of the fused features; the output layer maps the attention features to the predicted value of the boiler superheater wall temperature through a fully connected layer.

7. The method of claim 1, wherein, In step 4, the model hyperparameters include the number of neurons and the learning rate, the optimization range of the number of neurons is [32, 64, 128, 256], and the optimization range of the learning rate is [0.001, 0.0001, 0.00001, 0.000001].

8. System for implementing the method for predicting the wall temperature of a boiler superheater according to any one of claims 1-7, characterized in that, It comprises: A data collection module is configured to acquire historical operation data of the unit decentralized control system. A preprocessing module is configured to perform missing data filling, outlier processing, data dimension reduction and normalization processing on the historical operation data. A model operation module is configured to deploy a boiler superheater wall temperature prediction model based on BCNN-BAN, and perform prediction calculation to obtain a boiler superheater wall temperature prediction result. A visual terminal is configured to display the boiler superheater wall temperature prediction result online. 9.An electronic device comprising a processor and a memory, the electronic device characterized by: The memory stores a computer program, and the processor executes the program to implement the steps of the boiler superheater wall temperature prediction method according to any one of claims 1-7. 10.A non-transitory computer-readable storage medium having stored thereon a computer program, characterized in that, The computer program is executed by the processor to implement the steps of the boiler superheater wall temperature prediction method according to any one of claims 1-7.

Images