Boiler reheater wall temperature prediction method, overtemperature warning method, system and medium

Abstract

This invention discloses a boiler reheater wall temperature prediction method, overtemperature warning method, system, and medium, belonging to the technical field of safe operation of boiler heating surfaces. The method includes: collecting historical operating data and preprocessing it; using the kernel principal component analysis (KPCA) algorithm to filter and reduce the dimensionality of the original variables, selecting the first m principal component variables as model inputs; using a closed continuous time constant (CFC) neural network to capture short-term data characteristics, combined with a Kolmogorov-Arnold representation theorem-based neural network (KAN) to capture long-term dependencies, to construct a boiler reheater wall temperature prediction model based on the CFC-KAN; training and optimizing the model, and then predicting the boiler reheater wall temperature. The invention can accurately predict the boiler reheater wall temperature, providing a decision-making basis for overtemperature warning.

Description

Technical Field

[0001] The present invention relates to the technical field of safe operation of boiler heating surfaces, and in particular to a boiler reheater wall temperature prediction method, an overtemperature early warning method, a system and a medium. Background Art

[0002] The wall temperature of the boiler's heating surface is a critical parameter in unit operation, as tube bursts caused by global or local overheating are common. Monitoring and controlling the metal (wall) temperature of the boiler's heating surface is crucial for ensuring safe operation and extending its service life. With the increasing installed capacity of renewable energy, the peak-shaving role of thermal power units has become increasingly prominent. Frequent load fluctuations complicate the operating conditions of the boiler's heating surface, increasing the risk of overheating. Traditional control methods, such as water spray cooling, suffer from lags and are unable to respond promptly to rapidly changing operating conditions. Therefore, accurate and rapid prediction of the boiler's heating surface wall temperature and its changing trends is crucial for proactively implementing control measures and ensuring safe generator operation. Current methods for acquiring wall temperature data based on actual measurement techniques and temperature measurement points are susceptible to sensor damage or data loss in the context of deep peak-shaving.

[0003] At present, the research on the prediction of boiler heating surface wall temperature mainly focuses on numerical simulation and machine learning. Existing technologies include: "Study on the Overtemperature Characteristics of Boiler Steam Tube Based on Combustion and Hydrodynamic Coupling Model" (Yu Cong, Si Fengqi, Li Min, etc., "Thermal Energy and Power Engineering", 2021, 36(8): 92-98, August 19, 2021) proposed to use the furnace combustion-hydrodynamic coupling model to perform wall temperature numerical simulation; "Coupling Calculation Method of Coal-fired Boiler Platen Superheater Wall Temperature" (Jin Donghao, Liu Xin, Zhang Xiaoyuan, etc., "Proceedings of the Chinese Society of Electrical Engineering", 2022, 42(24): 8951-8960, February 28, 2022) proposed a coupled numerical calculation method for the wall temperature of coal-fired boiler platen superheater; "Data-driven Real-time Prediction Model of Boiler Water-cooled Wall Temperature Distribution" (Yan Jingwen, Liu Xin, Wang Guangli, etc., "Journal of China Coal Society", 2024, 49(10): 4117-4126, September 18, 2024) A coupled heat transfer model was constructed by combining a three-dimensional CFD model on the boiler flue gas side with a one-dimensional hydrodynamic model on the working fluid side.

[0004] The accuracy and robustness of numerical simulation methods are easily limited by model complexity and computing resources. Machine learning methods, especially deep learning, have shown potential in this field due to their powerful data processing and feature learning capabilities. "Boiler wall temperature prediction model based on NARX neural network" (Lu Bin, Liu Xi, Gao Lin, etc., "Thermal Power Generation", 2019, 48(3): 35-40, February 28, 2019) proposed the application of NARX neural network to the prediction of screen superheater wall temperature; "Supercritical boiler heating surface wall temperature prediction method based on data drive" (Wei Xiaobing, Cui Zhipeng, Xu Jing, etc., "Thermal Power Generation", 2023, 52(7): 106-112, May 19, 2023) proposed a wall temperature prediction model based on long short-term memory (LSTM) neural network; "Water-cooled wall temperature prediction based on improved gray wolf algorithm to optimize bidirectional long short-term memory neural network" (Zhan Yi, Feng Leihua, Yang Feng, etc., "Thermal Power Generation", 2024, 53(1):188-196, January 15, 2024) proposed a water-cooled wall temperature prediction model based on improved gray wolf algorithm to optimize bidirectional long short-term memory neural network; "Boiler heating surface wall temperature prediction based on TCN-Attention-BiGRU" (Cao Yiyun, Mao Dajun, Chen Siqin, "Computer Simulation", On April 3, 2024, a wall temperature prediction method based on TCN-Attention-BiGRU was proposed; "Data-driven modeling for overtemperature prediction of screen superheater of coal-fired boiler" (Fan Yuchen, Zhou Yongqing, Wei Chang, etc., "Proceedings of the Chinese Society of Electrical Engineering", September 26, 2024) proposed a deep neural network (GA-DNN) model based on genetic algorithm to optimize hyperparameters.

[0005] However, due to the high degree of coupling between various systems during unit operation and the complex factors affecting wall temperature, the predicted data often suffers from high redundancy and noise. Furthermore, while traditional machine learning methods and some artificial intelligence approaches have advantages in addressing nonlinear problems, further research is needed on deep learning for wall temperature prediction. Summary of the Invention

[0006] To address the above technical issues, this paper proposes a boiler reheater wall temperature prediction method, overtemperature warning method, system, and medium. A fusion model based on a closed-form continuous-time (CFC) neural network and a Kolmogorov-Arnold (KAN) neural network (i.e., a CFC-KAN-based boiler reheater wall temperature prediction model) is constructed. Experimental validation demonstrates the effectiveness of the proposed model, providing a reference for boiler heating surface wall temperature prediction research and supporting the safe operation of thermal power units.

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

[0008] The boiler reheater wall temperature prediction method includes the following steps:

[0009] Step 1: Collect historical operating data of the boiler from the unit's distributed control system;

[0010] Step 2: Preprocess the historical operation data and construct a standardized data set;

[0011] Step 3: Use the kernel principal component analysis algorithm to reduce the dimensionality of the original variables affecting the boiler reheater wall temperature, select the first m principal component variables as the model input, and divide the standardized data set into a training set and a test set;

[0012] Step 4: Construct a boiler reheater wall temperature prediction model based on CFC-KAN;

[0013] Step 5: Use the data from the training set to train the boiler reheater wall temperature prediction model based on CFC-KAN, and optimize the model hyperparameters using the grid search technique to obtain the optimized boiler reheater wall temperature prediction model based on CFC-KAN; the model hyperparameters include the number of neurons and the learning rate;

[0014] Step 6: Input the test set data into the optimized CFC-KAN-based boiler reheater wall temperature prediction model, obtain the prediction results, perform denormalization, and then use the evaluation function to analyze and evaluate the model's prediction performance to verify its prediction performance under different operating conditions.

[0015] Step 7: Input the real-time collected operating parameters of the coal-fired power generation unit into the optimized CFC-KAN-based boiler reheater wall temperature prediction model, and output the predicted value of the boiler reheater wall temperature within the future time T1.

[0016] Furthermore, in step 1, the historical operating data includes boiler reheater wall temperature and auxiliary variables.

[0017] Furthermore, in step 2, preprocessing the historical operation data includes: filling missing values ​​and abnormal values ​​using linear interpolation according to time labels, and then performing normalization.

[0018] Furthermore, in step 3, principal component variables with cumulative contribution rates greater than or equal to 95% are selected as model inputs.

[0019] Furthermore, in step 4, the boiler reheater wall temperature prediction model based on CFC-KAN constructed includes an input layer, a CFC layer, a KAN layer and an output layer, and the activation function is a rectified linear unit ReLU.

[0020] Furthermore, in step 5, the selection range of the number of neurons is [32, 64, 128, 256], and in step 7, T1=5 minutes.

[0021] The present invention also provides a boiler reheater wall temperature over-temperature early warning method, comprising the following steps:

[0022] Step A: obtaining a predicted value of the boiler reheater wall temperature using the boiler reheater wall temperature prediction method according to claim 1;

[0023] Step B: Calculate the average wall temperature rise rate within 1 minute before the current moment, predict the temperature rise δt within the future time T2 based on the average wall temperature rise rate, and superimpose it with the predicted value of the boiler reheater wall temperature to obtain the over-temperature warning temperature;

[0024] Step C: Compare the over-temperature warning temperature with the maximum wall temperature threshold allowed by the reheater material. When the over-temperature warning temperature is greater than or equal to the maximum wall temperature threshold, an over-temperature warning signal is triggered.

[0025] Furthermore, in step C, after the over-temperature warning signal is triggered, at least one of the following adjustment measures is performed:

[0026] Reduce fuel input and increase oxygen setpoint to optimize air-to-coal ratio;

[0027] Adjust the burner swing angle to reduce the flame height in the furnace;

[0028] Adjust the reheater damper opening to reduce the steam temperature.

[0029] The present invention also provides a boiler reheater wall temperature prediction and overtemperature warning system, including the following modules:

[0030] Data acquisition module, used to collect historical operation data of the boiler from the unit's distributed control system;

[0031] The preprocessing module is used to fill missing values, process outliers, and normalize the collected historical operating data. The model building module is used to build and train the boiler reheater wall temperature prediction model based on CFC-KAN.

[0032] The prediction module outputs the predicted value of the boiler reheater wall temperature and calculates the temperature rise rate based on the boiler wall temperature data 1 minute before the current moment;

[0033] The over-temperature warning module triggers an over-temperature warning signal and generates an adjustment instruction based on the comparison result of the predicted value of the boiler reheater wall temperature, the temperature rise rate and the maximum wall temperature threshold allowed by the reheater material.

[0034] The present invention also provides a computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, the above-mentioned boiler reheater wall temperature overtemperature early warning method is implemented.

[0035] Beneficial effects:

[0036] 1. The boiler reheater wall temperature prediction model based on CFC-KAN proposed in this paper combines the advantages of closed continuous time constant (CFC) neural network and Kolmogorov-Arnold (KAN) neural network, and is suitable for complex time series modeling tasks. The model improves the network's modeling ability for dynamic systems by introducing continuous time representation and learnable kernel functions. Among them, the CFC structure uses differential equations to represent the system dynamics and can handle inputs of any time interval; the time dependency is modeled by parameterized kernel functions. The KAN architecture is based on the Kolmogorov-Arnold theorem and disperses nonlinear transformations into weight functions at the edge of the network, with high model expression ability and low computational complexity. The boiler reheater wall temperature prediction model based on CFC-KAN has the following advantages: (1) It is suitable for processing long time series data and has good long-term prediction effect; (2) It has high computational efficiency; (3) It has strong generalization ability and good adaptability to irregular sampling data; (4) It has a theoretical basis.

[0037] 2. The early warning method of the present invention takes into account the time delay in the impact of adjustments to the furnace combustion conditions on the wall temperature. By calculating the average temperature rise rate (°C / min) over a period of time prior to the current moment (e.g., within the previous minute), it calculates the future temperature rise δt. This is then added to the predicted wall temperature to obtain the overtemperature warning temperature, which is then compared with the maximum allowable wall temperature threshold. When this temperature exceeds or equals the wall temperature threshold, an overtemperature warning signal is triggered, prompting operators to take prompt adjustment measures (such as reducing fuel input, increasing the oxygen setpoint to optimize the air-to-coal ratio, adjusting the burner swing angle, appropriately lowering the furnace flame height, or adjusting the reheater damper opening). This method helps reduce the risk of overheating in the boiler reheater tube wall, promotes safe operation of power plant boilers, and extends the service life of equipment. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] Figure 1 Flowchart of the boiler reheater wall temperature prediction method of the present invention;

[0039] Figure 2 This is the wall temperature response curve of measuring point 7 of the high-temperature reheater;

[0040] Figure 3 This is the result graph of kernel principal component analysis (KPCA);

[0041] Figure 4 This is a schematic diagram of the closed continuous time constant (CFC) neural network structure;

[0042] Figure 5 Schematic diagram of KAN network structure;

[0043] Figure 6 Comparison of reheater wall temperature prediction curves based on CFC-KAN;

[0044] Figure 7 Comparison of reheater wall temperature prediction curves based on CNN;

[0045] Figure 8 This is a comparison chart of the reheater wall temperature prediction curve based on LSTM;

[0046] Figure 9 This is a comparison chart of the reheater wall temperature prediction curve based on GRU;

[0047] Figure 10 The comparison chart of the reheater wall temperature prediction curve based on CFC;

[0048] Figure 11 The comparison chart of reheater wall temperature prediction curve based on KAN;

[0049] Figure 12 This is a graph showing the change in the training loss function of the boiler reheater wall temperature prediction model based on CFC-KAN as the number of iterations increases. DETAILED DESCRIPTION

[0050] In order to make the objectives, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely for the purpose of explaining the present invention and are not intended to limit the present invention. In addition, the technical features involved in the various embodiments of the present invention described below may be combined with each other as long as they do not conflict with each other.

[0051] like Figure 1 As shown, the present invention provides a boiler reheater wall temperature prediction method, comprising the following steps: Step 1: Collect historical operating data of the unit distributed control system (DCS) (i.e. Figure 1 Step 1: Collect historical operation data);

[0052] Step 2: Pre-process the historical operating data of the distributed control system (DCS) of the unit, including: missing data processing, data normalization (i.e. Figure 1 Step 2: Preprocess historical operation data); Step 3: Use the Kernel Principal Component Analysis (KPCA) algorithm to reduce the dimension of the original candidate variables affecting the boiler reheater wall temperature, select the first m principal component variables as the input features of the model, and divide the data set into a training set and a test set; (i.e. Figure 1Step 3 in

[15] : Data dimensionality reduction, divided into training set and test set); Step 4: Construct a boiler reheater wall temperature prediction model based on CFC-KAN;

[0053] Step 5: Use the training data to train the boiler reheater wall temperature prediction model based on CFC-KAN, and optimize the model hyperparameters using grid search technology, including the number of neurons and learning rate. Step 6: Input the test data into the boiler reheater wall temperature prediction model based on CFC-KAN, obtain the prediction results, denormalize them, and evaluate and analyze them.

[0054] Step 7: Input the real-time collected generator set operating parameters into the optimized CFC-KAN-based boiler reheater wall temperature prediction model, and output the predicted value of the boiler reheater wall temperature in the future time T1 (i.e. Figure 1 Step 7 in the process: real-time data collection is input into the boiler reheater wall temperature prediction model based on CFC-KAN, and the predicted value of the reheater wall temperature is output).

[0055] Preferably, pre-processing the historical operation data of the distributed control system of the unit includes: firstly filling the missing values ​​and abnormal values ​​in the historical operation data using linear interpolation according to the time tags, and then performing normalization processing.

[0056] Preferably, in step 3, the KPCA algorithm is used to reduce the dimension and select the first 26 principal component variables, with a cumulative contribution rate greater than or equal to 95% (ie, m=26).

[0057] Preferably, in step 4, the boiler reheater wall temperature prediction model based on CFC-KAN includes an input layer, a CFC layer, a KAN layer and an output layer, adopts a forward propagation method, and the activation function is a rectified linear unit (ReLU).

[0058] Preferably, in step 5, 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 the final optimized learning rate is 0.0001.

[0059] Preferably, in step 7, the future time T1 is 5 minutes.

[0060] The present invention also provides a boiler reheater wall temperature over-temperature early warning method, comprising the following steps: Step A: using the above-mentioned boiler reheater wall temperature prediction method to predict the boiler reheater wall temperature, to obtain a boiler reheater wall temperature prediction value;

[0061] Step B: Calculate the average wall temperature rise rate within 1 minute before the current moment, predict the temperature rise δt within the future set time T2 based on the temperature rise rate, and superimpose it with the predicted value of the boiler reheater wall temperature to obtain the over-temperature warning temperature;

[0062] Step C: comparing the over-temperature warning temperature with the maximum wall temperature threshold allowed by the reheater material, and triggering an over-temperature warning signal when the over-temperature warning temperature is greater than or equal to the maximum wall temperature threshold.

[0063] Preferably, T2=5 minutes.

[0064] Preferably, in step C, after the over-temperature warning signal is triggered, at least one of the following adjustment measures is performed:

[0065] Reduce fuel input and increase oxygen setpoint to optimize air-to-coal ratio;

[0066] Adjust the burner swing angle to reduce the flame height in the furnace;

[0067] Adjust the reheater damper opening to reduce the steam temperature.

[0068] Example 1

[0069] The embodiment of the present invention implements a boiler reheater wall temperature prediction method based on CFC-KAN, comprising the following steps:

[0070] Step 1: Collect historical operating data from the unit's distributed control system (DCS).

[0071] The study focused on the high-temperature reheater of a 1000MW ultra-supercritical single-reheat power generation unit. Historical data from 00:00:00 on March 6, 2025, to 22:40:00 on March 12, 2025, was extracted from the unit's distributed control system (DCS). A total of 10,000 samples were collected, with a sampling interval of 60 seconds. To identify locations with significant heat load deviations and a high risk of overheating, it was necessary to determine the water spray cooling response time corresponding to different heating surface wall temperatures. Analysis of historical operating data revealed that the wall temperature at measuring point 7 on the high-temperature reheater was generally elevated. Therefore, this study selected measuring point 7 on the high-temperature reheater as the research target.

[0072] Analyze the response characteristics of the wall temperature at the No. 7 measuring point of the reheater to the reheating desuperheating water spray rate, such as Figure 2 As shown in the image, the reheater spray desuperheating water flow rate increased at 01:18:00. At 01:23:00, the wall temperature at measuring point 7 began to decline, indicating a wall temperature response time of approximately 300 seconds. Therefore, to enable proactive action, the wall temperature prediction window was set to 5 minutes.

[0073] Step 2: Preprocess the historical operation data and build a standardized data set.

[0074] For the original historical operation data collected, linear interpolation is used to fill missing values ​​and outliers. Then, the cleaned input data is normalized using the following formula:

[0075] ;

[0076] Where, 、 Represent the minimum and maximum values ​​of the i-th variable, respectively. Represents the original data, is the normalized result.

[0077] Step 3: Based on the kernel principal component analysis algorithm, the original candidate variables affecting the reheater wall temperature are reduced in dimension and the principal component variables are selected as the model input.

[0078] Combined with historical data, the kernel principal component analysis algorithm was used to reduce the dimension of 36 parameters, including unit load, total fuel quantity, total air volume, primary air volume, secondary air volume percentage, burner swing angle command, Close-Coupled Over-Fire Air (CCOFA) damper opening, Separated Over-Fire Air (SOFA) swing angle, main steam pressure, main steam temperature, reheat steam temperature, first reheater left attemperator water volume, second reheater left attemperator water volume, first reheater right attemperator water volume, and second reheater right attemperator water volume. The analysis results are as follows: Figure 3 shown.

[0079] Figure 3 The contribution rate of the first 28 principal component variables is shown. The cumulative contribution rate of the first 26 principal component variables is 95%, so the first 26 principal component variables are selected as the input features of the model.

[0080] Step 4: Construct a boiler reheater wall temperature prediction model based on CFC-KAN.

[0081] The CFC-KAN-based boiler reheater wall temperature prediction model includes an input layer, a CFC layer, a KAN layer and an output layer, and the activation function is ReLU.

[0082] (1) Input layer: Receives the 26-dimensional input features after dimensionality reduction in step 3.

[0083] (2) CFC layer: It is built based on the closed-form continuous time constant (CFC) principle. CFC is a closed-form approximate solution of the liquid time constant (LTC) neural network, such as Figure 4 shown. Figure 4The backbone is the primary processing module for input data entering the LTC neurons. It performs preliminary operations such as feature extraction and transformation on the input data, providing the foundational data for subsequent processing by different branches. The backbone distributes the processed data to three branches, labeled g, f, and h. These branches perform different computations on the backbone output data. The g branch performs specific calculations on the input data, and its output participates in subsequent operations. The f branch performs calculations on the input data, negates the result, and then multiplies it with other calculation results (a cross indicates multiplication). The h branch also performs calculations on the input data, and its output also participates in subsequent operations. Figure 4 middle It is an activation function, common ones include Sigmoid function. and The relevant calculation results are processed separately and then fused with the results of other branches through the Hadamard product (the dots in the circle represent the multiplication of the corresponding elements). After the calculations and activation function processing of each branch, the results of different paths are combined through addition (indicated by the plus sign) to obtain the final output.

[0084] Compared to the LTC model that requires a numerical solver, CFC significantly improves training and inference efficiency. This layer efficiently extracts time-related dynamic features, and its core operation can be expressed as:

[0085] ;

[0086] Where f, g and h are trainable neural layers (parameters are θ f ,θ g and θ h ), represents input, σ represents sigmoid function, ⊙ represents Hadamard product (element-wise product), is the hidden state, and t represents the time.

[0087] (3) KAN layer: constructed based on the Kolmogorov-Arnold representation theorem. Figure 5 This is a schematic diagram of the KAN neural network structure. Figure 5 As can be seen, the input layer receives the original data, and the internal function on the connection line between it and the hidden layer (Small rectangle with curved pattern) is used as a learnable activation function, which is adaptively adjusted during training and transforms the input data and then sums it. The operation is passed to the hidden layer; the hidden layer further processes the data to extract features, and the external function on the connection line between it and the output layer (Small rectangle with curved pattern) After transforming the data again, it is passed to the output layer, which outputs the final result. Unlike traditional MLP (Multi-layer Perceptron) that places activation functions at the nodes, KAN places learnable activation functions at the edges of the network (weight positions), achieving high expressiveness and parameter efficiency through spline function fitting. This layer is used to deeply explore complex nonlinear relationships in input features. KAN model It consists of external functions and internal functions, expressed as:

[0088] ;

[0089] Among them, x is the n-dimensional input vector, representing the input to The data contains n components x1, x2, ... x n Where x p Represents the pth component in the input vector x, where p ranges from 1 to n and is used to index each element of the input vector; It is an internal function (learnable) with two subscripts q and p. p is used to index the components of the input vector, and q is used to number different internal functions. Function The domain is [0, 1], and the range is the real number set R; is an external function (learnable), whose domain and range are both real number sets R; Indicates x1, x2, ... x n is a function of a variable.

[0090] (4) Output layer: Receives the features processed by the KAN layer and outputs the predicted value of the boiler reheater wall temperature.

[0091] Step 5: Use the training set data to train the boiler reheater wall temperature prediction model based on CFC-KAN, and optimize the model hyperparameters through grid search technology.

[0092] The preprocessed standardized data set is divided into 8000 groups of data as training set and 2000 groups of data as test set according to the ratio of 8:2. The mean absolute percentage error (MAPE), root mean square error (RMSE), mean absolute error (MAE) and coefficient of determination (R) are used. 2 Evaluate model performance:

[0093] ;

[0094] Where, They represent the actual measured value, predicted value, and mean of the actual measured value, respectively; n represents the total number of samples; and i represents the index value.

[0095] Grid search technology was used to optimize its hyperparameters: Neurons Number optimization: the value range is [32, 64, 128, 256]. The results of the comparative model evaluation indicators are shown in Table 1. When the number of neurons is 128, the prediction effect is the best.

[0096] Table 1 Comparison of neuron numbers

[0097]

[0098] Learning rate optimization: The Adam optimization algorithm is used, with a value range of [0.001, 0.0001, 0.00001, 0.000001]. The results of the model evaluation indicators are shown in Table 2. When the learning rate is 0.0001, the prediction effect is the best.

[0099] Table 2 Comparison of learning rates

[0100]

[0101] Step 6: Input the test set data into the optimized CFC-KAN-based boiler reheater wall temperature prediction model, obtain the prediction results and perform denormalization processing, and use the evaluation function to analyze and evaluate the model prediction performance.

[0102] Step 6.1 Verify the wall temperature prediction: Use the test set data to make predictions, and compare the predicted values ​​with the actual measured values. Figure 6 As shown in the figure (blue: measured values, red line with stars: predicted values). The results show that the boiler reheater wall temperature prediction model based on CFC-KAN can accurately predict the wall temperature fluctuation trend and turning point, and the predicted values ​​are in good agreement with the measured values ​​during the stable period.

[0103] Step 6.2: Comparative analysis of models: The boiler reheater wall temperature prediction model based on CFC-KAN is compared with the CNN, LSTM, GRU, CFC, and KAN models. In order to reduce the impact of random factors on the prediction results of the above models, each neural network model uses the same historical data set for multiple repeated experiments. Figure 7-11 As shown, Figure 7 Comparison of reheater wall temperature prediction curves based on CNN; Figure 8 This is a comparison chart of the reheater wall temperature prediction curve based on LSTM; Figure 9 This is a comparison chart of the reheater wall temperature prediction curve based on GRU; Figure 10 The comparison chart of the reheater wall temperature prediction curve based on CFC; Figure 11The comparison of the reheater wall temperature prediction curve based on KAN is shown in Table 3. The comparison of quantitative evaluation indicators is shown in Table 3. As shown in Table 3, the error indicators of the boiler reheater wall temperature prediction model based on CFC-KAN (RMSE=1.05℃, MAE=0.92℃, MAPE=0.5015%) are the smallest, and the determination coefficient R 2 The value closest to 1 indicates that its prediction performance is the best.

[0104] Table 3 Comparison of wall temperature prediction performance of different models

[0105]

[0106] also, Figure 12 The training loss function of the CFC-KAN-based boiler reheater wall temperature prediction model is shown as a function of the number of iterations (normalized results). The CFC-KAN-based boiler reheater wall temperature prediction model gradually stabilizes and converges after approximately 20 iterations. Therefore, compared to the baseline prediction model, the proposed CFC-KAN-based boiler reheater wall temperature prediction model not only achieves more accurate predictions but also exhibits superior dynamic performance when predicting actual boiler heating surface wall temperatures.

[0107] Step 7: Input the real-time collected generator set operating parameters into the trained CFC-KAN-based boiler reheater wall temperature prediction model to output the predicted value of the reheater wall temperature within a future time (5 minutes in this embodiment).

[0108] Example 2

[0109] Taking a certain in-service transformer-operated once-through boiler in Example 1 as an example, the present invention further provides a boiler reheater wall temperature overtemperature warning method, comprising the following steps: Step A: using the boiler reheater wall temperature prediction method described in Example 1 to predict the boiler reheater wall temperature, to obtain a predicted value of the boiler reheater wall temperature;

[0110] Step B: Calculate the average wall temperature rise rate within 1 minute before the current moment, and predict the temperature rise within the future set time (5 minutes in this embodiment) based on the temperature rise rate And superimposed with the predicted value of boiler reheater wall temperature to obtain the over-temperature warning temperature;

[0111] Step C: Compare the over-temperature warning temperature with the maximum wall temperature threshold (e.g., 641°C) allowed by the reheater material (e.g., T92). When the over-temperature warning temperature is greater than or equal to the maximum wall temperature threshold, trigger an over-temperature warning signal and perform adjustment measures (e.g., adjusting the opening of the reheater damper to reduce the steam temperature).

[0112] Example 3

[0113] The present invention provides a boiler reheater wall temperature prediction and overtemperature warning system, comprising the following modules:

[0114] The data acquisition module is used to collect historical operation data (implementing step 1 of embodiment 1).

[0115] The preprocessing module is used to fill missing values, process outliers and normalize the historical operation data (implementing step 2 of embodiment 1).

[0116] The model building module is used to build and train a boiler reheater wall temperature prediction model based on CFC-KAN (implementing steps 3 to 5 of Example 1).

[0117] The prediction module outputs the future boiler reheater wall temperature prediction value (implementing step 7 of Example 1, where the prediction time is 5 minutes) and the temperature rise rate (calculated based on the wall temperature data 1 minute before the current moment) based on the trained boiler reheater wall temperature prediction model based on CFC-KAN.

[0118] The over-temperature warning module calculates the temperature rise rate based on the wall temperature data one minute before the current moment. Combining the wall temperature prediction value, the temperature rise rate, and the maximum wall temperature threshold allowed by the reheater material, it calculates the temperature rise δt within a set time in the future (5 minutes in this embodiment). This temperature rise δt is then added to the wall temperature prediction value to obtain the over-temperature warning temperature. When the over-temperature warning temperature is greater than or equal to the maximum wall temperature threshold, a warning signal is triggered and an adjustment instruction is generated.

[0119] Example 4

[0120] The present invention also provides a computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, the above-mentioned boiler reheater wall temperature overtemperature early warning method is implemented.

[0121] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the above-mentioned boiler reheater wall temperature overtemperature warning method.

[0122] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Thus, the present invention may take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk drives, CD-ROMs, optical storage devices, etc.) containing computer-usable program code. The solutions in the embodiments of the present invention may be implemented using various computer languages, such as the object-oriented programming language Java and the interpreted scripting language JavaScript.

[0123] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowcharts and / or block diagrams, as well as combinations of processes 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 processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowcharts and / or block diagrams. Figure 1 a process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0124] These computer program instructions may also be stored in a computer readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 a process or multiple processes and / or boxes Figure 1 The function specified in one or more boxes.

[0125] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operational steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing the instructions executed on the computer or other programmable device for implementing the process. Figure 1 a process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0126] Although the preferred embodiments of the present invention have been described, those skilled in the art may make additional changes and modifications to these embodiments once they have learned the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications that fall within the scope of the present invention.

[0127] Obviously, those skilled in the art may make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if such changes and modifications fall within the scope of the claims and their equivalents, the present invention is intended to include such changes and modifications.

Claims

1. A boiler reheater wall temperature prediction method, characterized in that: The following steps are involved: Step 1: Collect historical operating data of the boiler from the unit's distributed control system; Step 2: Preprocess the historical operation data and construct a standardized data set; Step 3: Use the kernel principal component analysis algorithm to reduce the dimensionality of the original variables affecting the boiler reheater wall temperature, select the first m principal component variables as the model input, and divide the standardized data set into a training set and a test set; Step 4: Construct a boiler reheater wall temperature prediction model based on CFC-KAN; The boiler reheater wall temperature prediction model based on CFC-KAN includes an input layer, a CFC layer, a KAN layer and an output layer, and the activation function is a rectified linear unit ReLU; Step 5: Use the data from the training set to train the boiler reheater wall temperature prediction model based on CFC-KAN, and optimize the model hyperparameters using the grid search technique to obtain the optimized boiler reheater wall temperature prediction model based on CFC-KAN; the model hyperparameters include the number of neurons and the learning rate; Step 6: Input the test set data into the optimized CFC-KAN-based boiler reheater wall temperature prediction model, obtain the prediction results, perform denormalization, and then use the evaluation function to analyze and evaluate the model's prediction performance to verify its prediction performance under different operating conditions. Step 7: Input the real-time collected operating parameters of the coal-fired power generation unit into the optimized CFC-KAN-based boiler reheater wall temperature prediction model, and output the predicted value of the boiler reheater wall temperature within the future time T1.

2. The boiler reheater wall temperature prediction method according to claim 1, characterized in that: In step 1, the historical operating data includes the boiler reheater wall temperature and auxiliary variables.

3. The boiler reheater wall temperature prediction method according to claim 1, characterized in that: In step 2, preprocessing the historical operation data includes: filling missing values ​​and abnormal values ​​using linear interpolation according to time labels, and then performing normalization.

4. The boiler reheater wall temperature prediction method according to claim 1, characterized in that: In step 3, the principal component variables with cumulative contribution rates greater than or equal to 95% are selected as model inputs.

5. The boiler reheater wall temperature prediction method according to claim 1, characterized in that: In step 5, the selection range of the number of neurons is [32, 64, 128, 256], and in step 7, T1=5 minutes.

6. The boiler reheater wall temperature over-temperature early warning method is characterized by: The steps include: Step A: obtaining a predicted value of the boiler reheater wall temperature using the boiler reheater wall temperature prediction method according to claim 1; Step B: Calculate the average wall temperature rise rate within 1 minute before the current moment, predict the temperature rise δt within the future time T2 based on the average wall temperature rise rate, and superimpose it with the predicted value of the boiler reheater wall temperature to obtain the over-temperature warning temperature; Step C: Compare the over-temperature warning temperature with the maximum wall temperature threshold allowed by the reheater material. When the over-temperature warning temperature is greater than or equal to the maximum wall temperature threshold, an over-temperature warning signal is triggered.

7. The boiler reheater wall overtemperature early warning method according to claim 6, characterized in that: In step C, after the over-temperature warning signal is triggered, at least one of the following adjustment measures is performed: Reduce fuel input and increase oxygen setpoint to optimize air-to-coal ratio; Adjust the burner swing angle to reduce the flame height in the furnace; Adjust the reheater damper opening to reduce the steam temperature.

8. Boiler reheater wall temperature prediction and over-temperature warning system, characterized by: The boiler reheater wall temperature prediction and over-temperature warning system is constructed based on the boiler reheater wall temperature prediction method according to claim 1, and the boiler reheater wall temperature prediction and over-temperature warning system includes the following modules: Data acquisition module, used to collect historical operation data of the boiler from the unit's distributed control system; The preprocessing module is used to fill missing values, process outliers and normalize the collected historical operation data; The model building module is used to build and train the boiler reheater wall temperature prediction model based on CFC-KAN. The prediction module outputs the predicted value of the boiler reheater wall temperature and calculates the temperature rise rate based on the boiler wall temperature data one minute before the current moment. The over-temperature warning module triggers an over-temperature warning signal and generates an adjustment instruction based on the comparison result of the predicted value of the boiler reheater wall temperature, the temperature rise rate and the maximum wall temperature threshold allowed by the reheater material.

9. A computer-readable storage medium, characterized in that The storage medium stores a computer program, which, when executed by a processor, implements the steps of the boiler reheater wall temperature overtemperature warning method according to claim 6 or 7.

Images