A method and related equipment for predicting and classifying the overheating trend of boiler water-cooled walls based on BO-GRU

By using a BO-GRU-based approach that integrates multi-source operating parameters and Bayesian optimization algorithms, accurate prediction and graded early warning of water-cooled wall overheating trends were achieved. This solved the problems of false alarms and missed alarms in existing technologies, and improved the safety and stability of boiler operation.

CN123069313APending Publication Date: 2026-11-10HUANENG SHANGHAI SHIDONGKOU SECOND POWER PLANT +2
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202611212122.6
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-08-11
Publication Date
2026-11-10

AI Technical Summary

Technical Problem

Existing boiler water-cooled wall overheat monitoring systems are unable to accurately reflect the degree of temperature deviation under different loads and combustion conditions. They lack the comprehensive utilization of multi-source information, leading to false alarms or missed alarms. Furthermore, they lack the ability to predict future temperature change trends, leaving operators with limited adjustment time.

Method used

A BO-GRU-based approach is adopted, which integrates multi-source operating parameters. Through data preprocessing, dynamic benchmark calculation, overtemperature feature extraction, and Bayesian optimization of the GRU model, the overtemperature trend of water-cooled walls is predicted and graded early warning is provided. This includes the analysis of temperature difference offset, temperature rise rate, regional temperature gradient, and temperature fluctuation intensity. The hyperparameters of the model are optimized by combining the Bayesian optimization algorithm.

Benefits of technology

It enables accurate identification and timely early warning of water-cooled wall overheating risks under complex operating conditions, reduces false alarms and missed alarms, provides ample adjustment time, reduces the risk of thermal fatigue and tube rupture accidents, and improves the safety and stability of boiler operation.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123069313A_ABST
    Figure CN123069313A_ABST
Patent Text Reader

Abstract

This invention discloses a method and related equipment for predicting and classifying early warning of boiler water-cooled wall overheating trends based on BO-GRU. The method involves acquiring multi-source operating parameters of the boiler and preprocessing the data to obtain preprocessed operating data; calculating the normal wall temperature reference value for water-cooled wall measuring points based on the preprocessed operating data; comparing the real-time wall temperature with the normal wall temperature reference value to obtain the temperature difference offset; extracting overheating feature quantities based on historical temperature sequences; calculating the overheating risk index based on the overheating feature quantities; constructing a time series sample set; inputting the time series sample set into the BO-GRU trend prediction model to output a predicted temperature curve; calculating the predicted overheating probability and the expected time to reach the early warning threshold based on the predicted temperature curve; comparing the overheating risk index, the predicted overheating probability, and the predicted temperature peak of the predicted temperature curve with a preset threshold; and outputting classified early warning information and coordinated response suggestions based on the comparison results.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention belongs to the field of boiler monitoring and relates to a method and related equipment for predicting and classifying the overheating trend of boiler water-cooled walls based on BO-GRU. Background Technology

[0002] As thermal power generating units develop towards higher parameters, larger capacity, and more flexible peak-shaving capabilities, boiler operating conditions are becoming increasingly complex. The safety requirements for water-cooled wall heating surfaces are also constantly increasing. In recent years, water-cooled walls have been subjected to high-temperature flue gas radiation, combustion disturbances, and a complex heat transfer environment. Once combustion deviation, flame center shift, slagging and ash accumulation, coal quality fluctuations, load changes, or uneven airflow distribution occur, localized areas are prone to abnormal temperature increases, a point that deserves serious attention. If not detected and addressed promptly, tube wall strength will decrease, thermal fatigue will intensify, and in severe cases, tube rupture accidents may occur, ultimately affecting the safe and stable operation of the entire unit. Therefore, accurate prediction and early warning of water-cooled wall overheating trends are crucial.

[0003] Looking at existing technical solutions, boiler water-cooled wall overheat monitoring typically works like this: real-time temperature is collected from wall temperature measuring points and compared with a fixed alarm threshold; an alarm is triggered if the set value is exceeded. While this method achieves basic monitoring, the problem lies in the fact that the normal temperature of the water-cooled wall varies under different loads, combustion methods, air volume ratios, feedwater flow rates, and furnace heat loads. Fixed thresholds are difficult to accurately reflect the degree of temperature deviation under current operating conditions, leading to frequent false alarms or missed alarms. Moreover, existing methods mostly rely solely on wall temperature as an indicator, severely lacking the comprehensive utilization of multi-source information such as furnace flue gas temperature, boiler load, main steam parameters, feedwater flow rate, air volume, furnace oxygen content, soot blowing status, flame images, and slagging and ash accumulation status. This results in an incomplete risk assessment.

[0004] Another issue that needs attention is that existing alarm systems mainly focus on whether the temperature is currently exceeding the limit, lacking the ability to predict future temperature trends. By the time an alarm actually occurs, a localized area of ​​the water-cooled wall is often already close to a dangerous state, leaving operators with very little time to adjust the air distribution, control the load increase rate, initiate soot blowing, or adjust the feedwater flow rate. Summary of the Invention

[0005] The purpose of this invention is to overcome the shortcomings of the prior art and provide a method and related equipment for predicting and classifying the overheating trend of boiler water-cooled walls based on BO-GRU. This method can integrate multi-source operating parameters, adapt to different operating conditions, predict the overheating trend of water-cooled walls, and provide classified early warnings.

[0006] To achieve the above objectives, the present invention employs the following technical solution: A method for predicting and classifying the overheating trend of boiler water-cooled walls based on BO-GRU includes the following process: The boiler's multi-source operating parameters are acquired and preprocessed to obtain preprocessed operating data. Calculate the normal wall temperature reference value of the water-cooled wall measuring point based on the pre-processed operating data; The temperature difference offset is obtained by comparing the real-time wall temperature with the normal wall temperature reference value. The over-temperature characteristics are extracted by combining the historical temperature sequence. The over-temperature characteristics include temperature difference offset, temperature rise rate, regional temperature gradient, duration of high temperature, and intensity of temperature fluctuation. The over-temperature risk index is calculated based on the over-temperature characteristics. A time series sample set was constructed by combining preprocessed operational data, historical wall temperature sequences, and over-temperature risk index. Input the time series sample set into the BO-GRU trend prediction model and output the predicted temperature curve; Calculate the predicted over-temperature probability and the estimated time to reach the warning threshold based on the predicted temperature curve; The system compares the over-temperature risk index, the predicted over-temperature probability, and the predicted temperature peak of the predicted temperature curve with the preset threshold, and outputs graded early warning information and joint response suggestions based on the comparison results.

[0007] Optional, multi-source operating parameters include the metal wall temperature at each measuring point of the water-cooled wall, furnace temperature or furnace outlet flue gas temperature, boiler load, main steam pressure, main steam temperature, feedwater flow rate, medium flow rate, primary air volume, secondary air volume, furnace oxygen content, fuel quantity, soot blower operation status, flame image information, and slagging and ash accumulation status information. Data preprocessing for multi-source operating parameters includes: resampling using a unified timestamp alignment method to synchronize the time of multi-source operating parameters; compensation for missing values ​​using forward padding, linear interpolation, or adjacent measurement point correlation compensation methods; removal of outliers using threshold removal, median filtering, or sliding window statistical methods; and obtaining normalized operating parameters by dividing the difference between the original operating parameters and the minimum value of the corresponding parameter by the difference between the maximum and minimum values ​​of the corresponding parameter.

[0008] Optionally, the normal wall temperature reference values ​​for water-cooled wall measuring points can be calculated based on the pre-processed operating data, including: Extract boiler load, feedwater flow rate, air volume parameters, furnace oxygen content, main steam pressure, main steam temperature, medium flow rate, and furnace temperature or furnace outlet flue gas temperature from the pre-processed operating data. The extracted parameters are input into the dynamic benchmark calculation model for processing, and the normal wall temperature reference value of the water-cooled wall measuring point under the current operating conditions is output.

[0009] Optionally, extracting overtemperature features by combining historical temperature sequences includes: The temperature rise rate is obtained by calculating the ratio of the temperature change of the water-cooled wall within the preset sliding time window to the time interval. The temperature gradient of the region is obtained by calculating the ratio of the real-time wall temperature difference between two adjacent water-cooled wall measuring points to the distance between the two measuring points. The average of the sum of squares of the differences between the water-cooled wall temperatures at multiple sampling points within a preset time window and the average water-cooled wall temperatures within the preset time window is used to obtain the intensity of temperature fluctuation. The duration of high temperature is obtained by counting the number of times the water-cooled wall temperature continuously exceeds the preset high temperature threshold.

[0010] Optionally, the over-temperature risk index can be calculated based on over-temperature characteristic quantities, including: The temperature difference offset, temperature rise rate, regional temperature gradient, high temperature duration and temperature fluctuation intensity are normalized to obtain the normalized feature quantities corresponding to each overtemperature feature quantity. The overheating risk index of the water-cooled wall measuring point is obtained by multiplying each normalized feature quantity by its corresponding weight coefficient and then summing the results.

[0011] Optionally, the steps of inputting the time series sample set into the BO-GRU trend prediction model and outputting the predicted temperature curve include: The preprocessed running data, historical wall temperature sequence, and over-temperature risk index are spliced ​​together to form a comprehensive feature vector; The input time series samples are constructed based on the preset sliding window length and the comprehensive feature vector, and the real values ​​of the water-cooled wall temperature at a preset number of future moments are combined as the output to form a time series sample set. The time series sample set is divided into a training set and a validation set; The prediction error of the validation set is used as the objective function of the Bayesian optimization algorithm. The hyperparameters of the GRU trend prediction model are optimized using the Bayesian optimization algorithm. The root mean square error of the objective function is obtained by calculating the square root of the average of the sum of squares of the differences between the predicted wall temperature and the actual wall temperature. The optimal combination of hyperparameters is configured into the BO-GRU trend prediction model to obtain the trained trend prediction model. The current time series sample is input into the trained trend prediction model, which outputs the predicted temperature curve.

[0012] Optionally, the tiered early warning system includes Level 1 alert, Level 2 anomaly alert, and Level 3 danger alert. The specific logic includes: When the over-temperature risk index is greater than or equal to the first risk index threshold and less than the second risk index threshold, or the predicted over-temperature probability is greater than or equal to the first predicted over-temperature probability threshold and less than the second predicted over-temperature probability threshold, or the predicted temperature peak is greater than or equal to the first predicted temperature threshold and less than the second predicted temperature threshold, a first-level warning will be output. When the over-temperature risk index is greater than or equal to the second risk index threshold and less than the third risk index threshold, or the predicted over-temperature probability is greater than or equal to the second predicted over-temperature probability threshold and less than the third predicted over-temperature probability threshold, or the predicted temperature peak is greater than or equal to the second predicted temperature threshold and less than the third predicted temperature threshold, a level 2 abnormal warning will be output. When the over-temperature risk index is greater than or equal to the third risk index threshold, or the predicted over-temperature probability is greater than or equal to the third predicted over-temperature probability threshold, or the predicted temperature peak is greater than or equal to the third predicted temperature threshold, a level three danger warning is output. The first risk index threshold is less than the second risk index threshold and the second risk index threshold is less than the third risk index threshold; the first predicted over-temperature probability threshold is less than the second predicted over-temperature probability threshold and the second predicted over-temperature probability threshold is less than the third predicted over-temperature probability threshold; and the first predicted temperature threshold is less than the second predicted temperature threshold and the second predicted temperature threshold is less than the third predicted temperature threshold.

[0013] A boiler water-cooled wall overheating trend prediction and hierarchical early warning system based on BO-GRU includes: The preprocessing module is used to acquire multi-source operating parameters of the boiler and perform data preprocessing to obtain preprocessed operating data. The dynamic benchmark calculation module is used to calculate the normal wall temperature reference value of the water-cooled wall measuring point based on the pre-processed running data. The feature extraction module is used to compare the real-time wall temperature with the normal wall temperature reference value to obtain the temperature difference offset, and extract the over-temperature feature quantity by combining it with the historical temperature sequence. The over-temperature feature quantity includes temperature difference offset, temperature rise rate, regional temperature gradient, high temperature duration and temperature fluctuation intensity; and calculate the over-temperature risk index based on the over-temperature feature quantity. The sample set construction module is used to combine preprocessed running data, historical wall temperature sequences, and over-temperature risk index to construct a time series sample set; The trend prediction module is used to input the time series sample set into a gated recurrent unit network model that combines a Bayesian optimization algorithm and output the predicted temperature curve. The probability and time calculation module is used to calculate the predicted over-temperature probability and the estimated time to reach the warning threshold based on the predicted temperature curve. The graded early warning module is used to compare the over-temperature risk index, the predicted over-temperature probability, and the predicted temperature peak of the predicted temperature curve with the preset threshold. Based on the comparison results, it outputs graded early warning information and linkage response suggestions.

[0014] A computer device includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the BO-GRU-based method for predicting and classifying the overheating trend of boiler water-cooled walls.

[0015] A computer-readable storage medium storing a computer program, which, when executed by a processor, implements the steps of the BO-GRU-based method for predicting and classifying the overheating trend of boiler water-cooled walls.

[0016] Compared with the prior art, the present invention has the following beneficial effects: This invention, by preprocessing multi-source operating data and establishing a dynamic benchmark, restores the true deviation of wall temperature under different combustion states and load conditions. Combining multi-dimensional overtemperature characteristics and historical sequences, this method not only provides a transient assessment of current overtemperature risks but also mines the nonlinear dynamic evolution patterns in the temperature time series using a BO-GRU model, enabling it to predict future temperature change trends in advance. This provides ample adjustment delay for the operating system, overcomes the problem of alarm lag in existing technologies, and helps reduce the risk of accelerated localized pipe wall thermal fatigue and pipe rupture accidents.

[0017] Furthermore, acquiring multi-dimensional operating parameters covering heat load, medium flow rate, and combustion state ensures the integrity of thermodynamic and aerodynamic feature inputs. In the data preprocessing stage, a time synchronization mechanism eliminates the barriers of sampling periods from multi-source heterogeneous sensors, ensuring strict alignment of time-series signals on a unified time axis; missing and outlier removal mechanisms repair distorted data waveforms caused by sensor failures or communication interference; and normalization eliminates dimensional differences between multi-source parameters, mapping input data to the same feature space scale. These preprocessing mechanisms reduce noise and anomalous disturbances in the source data, significantly improving the convergence speed of subsequent model training and the effectiveness of feature extraction.

[0018] Furthermore, the normal operating temperature of the water-cooled wall is not a constant, but highly dependent on the current boiler load, air volume ratio, feedwater flow rate, and the heat load distribution within the furnace. This solution inputs real-time operating parameters into a dynamic benchmark calculation model, mapping in real-time the normal wall temperature reference value that should be present under the current specific thermodynamic boundary conditions. This adaptive dynamic benchmark can isolate the legitimate temperature fluctuation range caused by normal switching of operating conditions, thereby identifying the true temperature difference offset and effectively eliminating the false alarms and missed alarms that are prone to occur under complex variable load conditions.

[0019] Furthermore, local overheating is not only manifested in the absolute value of the current temperature exceeding the limit, but its development rate in the time domain and its diffusion distribution in the spatial domain also determine the degree of damage to the material strength. The temperature rise rate parameter reflects the swiftness of temperature deterioration caused by heat accumulation; the regional temperature gradient parameter is used to characterize the spatial temperature difference between adjacent measuring points, reflecting the degree of uneven heating in the local water-cooled wall and the potential thermal stress risk; the temperature fluctuation intensity parameter quantifies the heat transfer instability caused by local airflow disturbance; the high temperature duration parameter is used to characterize the cumulative time that the water-cooled wall is in an abnormally high temperature state, providing a basis for assessing the material's thermal fatigue and high-temperature damage risk. The extraction of multi-dimensional parameters comprehensively depicts the evolutionary nature of the overheating phenomenon.

[0020] Furthermore, since different overheating characteristic parameters have varying destructive mechanisms and contribution rates to pipe wall strength, directly superimposing a single feature cannot objectively measure the comprehensive hazard equivalent. This scheme normalizes and unifies the scale of multidimensional overheating characteristics, then uses weighted coefficients for linear weighted fusion calculation. This dimensionality reduction mechanism can reasonably map different parameters according to their relative hazard proportions in the overheating disaster evolution chain, converging complex, discrete spatiotemporal multidimensional parameters into a quantitative overheating risk index scalar. This achieves a scientific calibration of the pipe wall health status in local areas, providing a hard quantitative indicator for alarm decision-making and control logic.

[0021] Furthermore, the gating mechanism built into the GRU (Gated Recurrent Unit) network possesses excellent long-short-term dependency capture capabilities, highly fitting the dynamic evolution trajectory of water-cooled wall temperature under complex and variable operating conditions, characterized by large delays and strong coupling in nonlinear time series. Since the fitting performance of deep learning is greatly limited by the hyperparameter structure, traditional manual hyperparameter tuning suffers from strong blindness and a tendency to get trapped in local optima. By introducing the Bayesian optimization (BO) algorithm, and constructing a Gaussian process surrogate model and acquisition function, driven by the root mean square error of the validation set prediction, it can efficiently and adaptively converge to the optimal hyperparameter structure of the GRU model within the global parameter space. The combination of these two approaches endows the model with extremely high regression accuracy and robust generalization ability.

[0022] Furthermore, by setting progressively higher first, second, and third multi-dimensional threshold boundaries, the originally ambiguous over-temperature state is precisely discretized into three clear control levels: alert, abnormal, and dangerous. This completes the mode reversal of status monitoring from delayed alarms to proactive, tiered prevention. Based on the corresponding linkage response strategies triggered by different risk levels, the control system ensures seamless integration with cooling intervention operations such as load reduction, combustion angle adjustment, or soot blowing under different emergency conditions, forming a closed loop for risk suppression. Attached Figure Description

[0023] Figure 1The overall flowchart of the boiler water-cooled wall overheating trend prediction and graded early warning method based on BO-GRU provided by the present invention; Figure 2 This is a schematic diagram of the dynamic benchmark calculation module structure provided by the present invention; Figure 3 The flowchart for overheat feature extraction and risk assessment provided by this invention; Figure 4 The BO-GRU trend prediction model and hierarchical early warning flowchart provided by this invention; Figure 5 The overall structural block diagram of the boiler water-cooled wall overheating trend prediction and graded early warning device based on BO-GRU provided by the present invention.

[0024] Figure 6 This is a schematic diagram of the linkage processing and display alarm module provided by the present invention. Detailed Implementation

[0025] Embodiments of the present invention are described in detail below, examples of which are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention, and should not be construed as limiting the present invention.

[0026] The following disclosure provides many different embodiments or examples for implementing various structures of the invention. To simplify the disclosure, specific examples of components and arrangements are described below. These are merely examples and are not intended to limit the invention. Furthermore, reference numerals and / or letters may be repeated in different examples; such repetition is for simplification and clarity and does not in itself indicate a relationship between the various embodiments and / or arrangements discussed. In addition, examples of various specific processes and materials are provided in this invention, but those skilled in the art will recognize the application of other processes and / or the use of other materials.

[0027] Example 1 This embodiment provides a method for predicting and classifying early warning of boiler water-cooled wall overheating trends based on BO-GRU. This method is specifically designed to predict the temperature change trend of boiler water-cooled walls and output classified early warning information based on the prediction results and risk assessment results.

[0028] like Figure 1 As shown, the method mainly includes the following steps: S1: Collect multi-source operating parameters required for predicting the overheating trend of boiler water-cooled walls.

[0029] Specifically, the multi-source operating parameters include the metal wall temperature at each measuring point of the water-cooled wall, and include at least one of the following: furnace temperature or furnace outlet flue gas temperature, boiler load, main steam pressure, main steam temperature, feedwater flow rate, medium flow rate, primary air volume, secondary air volume, furnace oxygen content, fuel quantity, soot blower operation status, flame image information, and slagging and ash accumulation status information.

[0030] In this embodiment, the metal wall temperature at each measuring point of the water-cooled wall is denoted as Tw_i(t), where i represents the measuring point number of the water-cooled wall and t represents the sampling time; the furnace temperature or the flue gas temperature at the furnace outlet is denoted as Tg(t); the boiler load is denoted as L(t); the feedwater flow rate is denoted as Q(t); and the furnace oxygen content is denoted as O(t).

[0031] S2: Perform data preprocessing on the collected multi-source operating parameters to obtain preprocessed boiler operating data.

[0032] The data preprocessing includes time synchronization, missing value compensation, outlier removal, measurement point reliability assessment, and normalization. In some implementations, it also includes noise reduction filtering and measurement point reliability assessment. The purpose of these processes is to improve data quality.

[0033] For data from different sampling periods, resampling is performed using a unified timestamp alignment method, which allows data from different sources to form a sample sequence at the same time scale.

[0034] For missing data, forward imputation, linear interpolation, or compensation by association with adjacent measurement points can be used to fill in the missing data. For outliers that are significantly outside the reasonable range, threshold removal, median filtering, or sliding window statistical methods can be used to identify and correct them.

[0035] In a preferred embodiment, the operating parameters are normalized using the following normalization formula:

[0036] in, Indicates the original operating parameters. This represents the minimum value of the corresponding parameter. This indicates the maximum value of the corresponding parameter. This represents the normalized operating parameters. This facilitates subsequent modeling and calculations.

[0037] S3: Calculate the normal wall temperature reference value of the water-cooled wall measuring point under the current operating conditions based on the pre-processed boiler operating data, and establish a dynamic benchmark.

[0038] Specifically, such as Figure 2As shown, based on parameters such as current boiler load, feedwater flow rate, air volume, furnace oxygen content, main steam pressure, main steam temperature, medium flow rate, and furnace temperature or flue gas temperature, the system outputs the normal wall temperature reference value for the corresponding water-cooled wall measuring point under the current operating conditions. .

[0039] The calculation model for the normal wall temperature reference value Tref_i(t) can be expressed as:

[0040] in, This represents the reference value of the normal wall temperature at time t under the current operating conditions corresponding to the i-th water-cooled wall measuring point; Represents the dynamic benchmark calculation model; Indicates air volume parameter; Indicates the main steam pressure; Indicates the main steam temperature; Indicates the flow rate of the medium.

[0041] Furthermore, the real-time wall temperature Tw_i(t) is compared with the normal wall temperature reference value Tref_i(t) to obtain the temperature difference offset ΔT_i(t), which is calculated using the following formula:

[0042] in, This represents the temperature difference offset at the i-th water-cooled wall measuring point at time t. The larger this offset, the closer the risk of overheating.

[0043] S4: Extract overheating features based on real-time wall temperature, normal wall temperature reference values, and historical temperature sequences.

[0044] Specifically, the overheating characteristics include multiple aspects such as temperature difference offset, temperature rise rate, regional temperature gradient, duration of high temperature, and intensity of temperature fluctuation. A comprehensive consideration of these characteristics can more fully reflect the state and trend of overheating.

[0045] Among them, the rate of temperature rise The calculation formula is as follows: It is determined based on the temperature change of the water-cooled wall within the preset sliding time window and the time interval.

[0046] in, This indicates the length of the sliding time window. This metric reflects the rate of temperature increase very well.

[0047] Regional temperature gradient The temperature difference between adjacent measuring points and the distance between measuring points are determined, and the calculation formula is as follows:

[0048] in, These represent the real-time wall temperatures at two adjacent water-cooled wall measuring points. This indicates the distance between two measuring points.

[0049] Temperature fluctuation intensity The duration of high temperature is determined based on the standard deviation of the water-cooled wall temperature within a preset time window. The value is determined based on the duration for which the water-cooled wall temperature continuously exceeds a preset high-temperature threshold. This indicator effectively reflects the intensity of temperature fluctuations. The calculation formula is as follows:

[0050] in, For the intensity of temperature fluctuation, The number of sampling points within a preset time window. For the first Water-cooled wall temperature at each sampling point This represents the average water-cooled wall temperature within that time window. This metric is used to assess the severity of temperature fluctuations.

[0051] S5: Calculate the water-cooled wall overheat risk index based on the overheat characteristic quantity, and conduct a risk level assessment.

[0052] like Figure 3 As shown, in this embodiment, the temperature difference offset, temperature rise rate, regional temperature gradient, duration of high temperature, and intensity of temperature fluctuation are normalized to obtain normalized feature quantities. .

[0053] The overheating risk index The calculation formula is as follows:

[0054] in, This represents the overheating risk index of the i-th water-cooled wall measuring point at time t. These represent the weight coefficients of the corresponding feature quantities. These are the normalized temperature difference offset, normalized temperature rise rate, normalized regional temperature gradient, normalized high temperature duration, and normalized temperature fluctuation intensity, respectively. All weighting coefficients are non-negative, and the sum of the weighting coefficients is 1.

[0055] Furthermore, the weighting coefficients satisfy the following constraints:

[0056] S6: Construct a BO-GRU trend prediction model to predict the trend of water-cooled wall temperature change within a preset time period.

[0057] Specifically, such as Figure 4 As shown, this is based on multi-source historical operating data, historical wall temperature sequences, and over-temperature risk indices. The time series sample set is constructed by using sliding time window samples as model input.

[0058] Let the sliding window length be m and the prediction step size be h, then the input sequence of the t-th sample can be represented as:

[0059] Its corresponding output can be represented as:

[0060] in, This represents the input time series sample. Let represent the true water-cooled wall temperature over the next h time points, and let x(t) represent the comprehensive feature vector at time t, composed of operating parameters, wall temperature parameters, and risk characteristics. Based on this, an input feature sequence for predicting the water-cooled wall temperature trend is constructed.

[0061] This embodiment uses a GRU neural network to predict the time series of water-cooled wall temperatures, and simultaneously employs the Bayesian optimization algorithm to optimize the hyperparameters of the GRU model. This method leverages the powerful nonlinear fitting capabilities of neural networks while using Bayesian optimization to find the optimal model parameters.

[0062] Furthermore, the hyperparameters of the GRU model include at least one of the following: learning rate, number of hidden layer nodes, number of network layers, batch size, sliding window length, Dropout ratio, and number of training epochs; the number of training epochs can be determined based on the validation set error or early stopping condition.

[0063] The BO algorithm uses the prediction error on the validation set as the objective function, which can be represented by the root mean square error (RMSE), and its calculation formula is as follows:

[0064] Where N represents the number of samples in the validation set. This indicates that the model predicts the wall temperature. This indicates the corresponding actual wall temperature.

[0065] The Boolean algorithm constructs a surrogate model based on existing hyperparameter combinations and their corresponding objective function values. Then, it selects the next set of hyperparameter combinations to be evaluated using a sampling function, continuing until the maximum number of iterations is reached or the error convergence condition is met. Finally, it outputs the optimal hyperparameter combination. This iterative optimization approach can efficiently find good parameter configurations.

[0066] The optimal hyperparameter combination is input into the GRU model to obtain the optimized BO-GRU trend prediction model. This model is then used to output the predicted temperature curves for a predetermined future time period, providing a basis for subsequent risk assessment and early warning.

[0067] S7: Calculate the predicted over-temperature probability and the estimated time to reach the warning level based on the predicted temperature curve.

[0068] In a preferred embodiment, the predicted over-temperature probability is calculated based on the relationship between the predicted peak temperature within a future preset time window and the safe temperature threshold of the water-cooled wall. Its expression is as follows:

[0069] in, This represents the predicted over-temperature probability at the i-th water-cooled wall measuring point; This represents the predicted peak temperature at the i-th measurement point within a future preset time window. represents the safe temperature threshold of the water-cooled wall; k represents the probability adjustment coefficient.

[0070] The expected time to reach the warning level is determined based on the moment when the predicted temperature curve first reaches the corresponding warning threshold.

[0071] S8: Classified early warning based on over-temperature risk index, predicted over-temperature probability, and predicted temperature peak.

[0072] In this embodiment, the warning levels include Level 1 alert, Level 2 anomaly warning, and Level 3 danger warning. Specifically, Level 1 alert indicates a localized temperature rise trend in the water-cooled wall; Level 2 anomaly warning indicates a significant risk of overheating in a localized area of ​​the water-cooled wall; and Level 3 danger warning indicates a high level of danger posed by the water-cooled wall.

[0073] Let the risk index thresholds be R1, R2, and R3, where R1 < R2 < R3; the predicted over-temperature probability thresholds be P1, P2, and P3, where P1 < P2 < P3; and the predicted temperature thresholds be T1, T2, and T3, where T1 < T2 < T3.

[0074] When satisfied ,or ,or When this happens, a level one warning message will be output.

[0075] When satisfied ,or ,or At that time, a level-two abnormality warning will be issued.

[0076] When satisfied ,or ,or At that time, a level three danger warning will be issued.

[0077] S9: Output joint response suggestions and display alarm information according to the warning level.

[0078] Specifically, the coordinated response recommendations include adjusting the ratio of primary and secondary air, adjusting the burner tilt angle or operating mode, controlling the boiler load increase rate, starting the soot blower in the corresponding area, adjusting the feedwater flow rate, or adjusting the operating parameters of the circulation loop.

[0079] The alarm information displayed includes the location of high-risk areas, real-time wall temperature, reference value of normal wall temperature, temperature difference deviation, over-temperature risk index, predicted temperature change trend, predicted over-temperature probability, warning level, and corresponding handling suggestions.

[0080] If the over-temperature risk index, predicted over-temperature probability, and predicted peak temperature all decrease to within a safe range after the coordinated response, the warning will be lifted and continuous monitoring will begin; if the risk still has not decreased to a safe range, the series of processes including monitoring, assessment, prediction, warning, and response will continue.

[0081] This embodiment can obtain a reference value for the normal wall temperature of the water-cooled wall under the current operating conditions based on the boiler's current load, combustion status, air volume ratio, medium flow rate, and furnace heat load. Then, by combining multiple characteristics such as real-time wall temperature deviation, temperature rise rate, regional temperature gradient, and duration of high temperature, a comprehensive assessment of the risk of water-cooled wall overheating is performed, thereby improving the accuracy of overheating risk identification.

[0082] This embodiment can predict the temperature change trend of the water-cooled wall within a preset time period. This allows for early warning before the water-cooled wall reaches a dangerous temperature. Based on the level of risk, different levels of alarm information and corresponding handling suggestions are output, providing operators with strong decision-making support for operations such as adjusting combustion, controlling load, initiating soot blowing, and adjusting feedwater flow. This significantly improves the safety, stability, and intelligence level of boiler operation.

[0083] In summary, this embodiment, by integrating multi-source operating parameters, dynamic benchmark calculation, multi-dimensional over-temperature feature extraction, BO-GRU trend prediction, and a tiered early warning strategy, can predict the future temperature change trend of the water-cooled wall before it reaches a dangerous temperature. Based on the level of risk, it outputs tiered early warnings and corresponding handling suggestions, which greatly improves the accuracy of boiler water-cooled wall over-temperature risk identification and the timeliness of early warning.

[0084] Example 2 This embodiment also provides a dynamic early warning device for boiler water-cooled wall overheating, such as... Figure 5As shown, the device includes: a data acquisition module, a data preprocessing module, a dynamic benchmark calculation module, an over-temperature feature extraction module, a risk assessment module, a trend prediction module, a graded early warning module, a linkage response module, and a display and alarm module.

[0085] The data acquisition module is responsible for collecting multi-source operating parameters during boiler operation. These parameters include: metal wall temperature at various measuring points on the water-cooled wall, furnace temperature or flue gas temperature at the furnace outlet, boiler load, main steam temperature, main steam pressure, feedwater flow rate, primary air volume, secondary air volume, furnace oxygen content, fuel quantity, sootblower operation status, flame image information, and slagging and ash accumulation status information, etc.

[0086] The data preprocessing module is connected to the data acquisition module. Its responsibility is to perform time synchronization, noise reduction filtering, missing value compensation, outlier removal, and measurement point reliability assessment on the acquired multi-source operating parameters, and finally obtain the preprocessed boiler operating data for subsequent use.

[0087] As a preferred approach, the data preprocessing module normalizes the collected data.

[0088] The dynamic baseline calculation module is connected to the data preprocessing module. It is used to calculate the normal wall temperature reference value of the water-cooled wall measuring point under the current operating conditions based on the preprocessed boiler operating data, combined with the current boiler load, combustion conditions, medium flow rate, furnace heat load, and air volume ratio. This is a key step in establishing a dynamic baseline.

[0089] As a preferred embodiment, the dynamic baseline calculation module includes a baseline model unit and a temperature difference calculation unit. The baseline model unit is used to output a reference value for the normal wall temperature of the water-cooled wall based on the current operating parameters.

[0090] The temperature difference calculation unit is used to calculate the temperature difference offset between the real-time wall temperature of the water-cooled wall and the normal wall temperature reference value.

[0091] The over-temperature feature extraction module is connected to the dynamic benchmark calculation module. It extracts over-temperature features based on the real-time wall temperature of the water-cooled wall, the reference value of the normal wall temperature, and the historical temperature sequence. These features mainly include: temperature difference offset, temperature rise rate, regional temperature gradient, duration of high temperature, and intensity of temperature fluctuation, among other aspects.

[0092] The risk assessment module is connected to the over-temperature feature extraction module. It is used to calculate the over-temperature risk index of the water-cooled wall based on the temperature difference offset, temperature rise rate, regional temperature gradient, duration of high temperature, and intensity of temperature fluctuation, and then to comprehensively assess the over-temperature risk of different measuring points or different regions.

[0093] The trend prediction module is connected to the data preprocessing module and the risk assessment module. Its function is to predict the trend of water-cooled wall temperature changes within a preset time period based on real-time operating data and historical temperature sequences, and finally obtain important information such as predicted temperature, predicted overheating probability, and the expected time to reach the warning level.

[0094] like Figure 6 As shown, the graded early warning module is connected to the risk assessment module and the trend prediction module, and is used to predict the over-temperature risk index. Predicting the probability of overheating and predicted peak temperature Determine the warning level; when or When, output a level 1 warning; when When, output a level 2 anomaly warning; when or At that time, a level three danger warning will be issued, among which, The risk thresholds are progressively larger. These are progressively increasing probability thresholds. This tiered design allows operators to take appropriate action based on different warning levels.

[0095] The coordinated response module is connected to the tiered early warning module. When the early warning conditions are met, it will output response suggestions or control instructions. These suggestions or instructions include the following: adjusting the ratio of primary and secondary air, adjusting the burner's tilt angle or operating mode, controlling the rate of boiler load increase, activating the soot blowers in the corresponding area, adjusting the feedwater flow rate, or adjusting the operating parameters of the circulation loop, etc.

[0096] The alarm display module is connected to the graded early warning module. Its responsibilities include displaying information such as the location of high-risk areas in the water-cooled wall, real-time wall temperature, normal wall temperature reference value, temperature difference deviation, over-temperature risk index, predicted temperature change trend, early warning level, and corresponding handling suggestions. Based on different early warning levels, it outputs corresponding pop-up prompts, audible and visual alarms, or danger alarm signals.

[0097] The data acquisition module includes various hardware devices such as wall temperature sensors, furnace temperature sensors, pressure sensors, flow sensors, oxygen detection devices, airflow detection devices, image acquisition devices, and soot blowing status acquisition interfaces. Among these, the wall temperature sensors are arranged along different heights and areas of the boiler water-cooled wall to acquire the metal wall temperature at different locations. This arrangement ensures the comprehensiveness and accuracy of temperature measurements.

[0098] The image acquisition system includes an infrared thermal imaging device and a flame image acquisition device. The infrared thermal imaging device is used to acquire heat distribution information in the water-cooled wall area, while the flame image acquisition device is used to acquire information about the offset of the flame center inside the furnace. These two devices provide important image information.

[0099] The linkage response module is connected via communication with the boiler combustion control system, soot blowing control system, feedwater control system, or load control system. Based on the results of tiered early warnings, the system can output linkage response suggestions; and, under conditions that meet control authority, equipment interlocking, and unit safety constraints, it can also output control commands to the corresponding control systems.

[0100] This embodiment, by incorporating multiple functional modules including a data acquisition module, a data preprocessing module, a dynamic benchmark calculation module, an over-temperature feature extraction module, a risk assessment module, a trend prediction module, a tiered early warning module, a coordinated response module, and a display and alarm module, transforms the traditional single-source wall temperature alarm into a multi-source parameter fusion judgment. The advantage of this approach is that it reduces the possibility of false alarms or missed alarms arising from fixed threshold alarms under different operating conditions.

[0101] Example 3: This embodiment provides a BO-GRU-based system for predicting and classifying overheating trends in boiler water-cooled walls. This BO-GRU-based system can be used to implement the aforementioned BO-GRU-based method for predicting and classifying overheating trends in boiler water-cooled walls. Specifically, the BO-GRU-based system includes a preprocessing module, a dynamic benchmark calculation module, a feature extraction module, a sample set construction module, a trend prediction module, a probability and time calculation module, and a classifying early warning module.

[0102] The preprocessing module is used to acquire multi-source operating parameters of the boiler and perform data preprocessing to obtain preprocessed operating data.

[0103] The dynamic benchmark calculation module is used to calculate the normal wall temperature reference value of the water-cooled wall measuring point based on the preprocessed running data.

[0104] The feature extraction module is used to compare the real-time wall temperature with the normal wall temperature reference value to obtain the temperature difference offset, and to extract the over-temperature feature quantity by combining the historical temperature sequence. The over-temperature feature quantity includes temperature difference offset, temperature rise rate, regional temperature gradient, duration of high temperature, and intensity of temperature fluctuation; and calculates the over-temperature risk index based on the over-temperature feature quantity.

[0105] The sample set construction module is used to combine preprocessed running data, historical wall temperature sequences, and over-temperature risk indices to construct a time series sample set.

[0106] The trend prediction module is used to input the time series sample set into a gated recurrent unit network model that combines a Bayesian optimization algorithm and output the predicted temperature curve.

[0107] The probability and time calculation module is used to calculate the predicted over-temperature probability and the estimated time to reach the warning threshold based on the predicted temperature curve.

[0108] The graded early warning module is used to compare the over-temperature risk index, the predicted over-temperature probability, and the predicted temperature peak of the predicted temperature curve with the preset threshold. Based on the comparison results, it outputs graded early warning information and linkage response suggestions.

[0109] Example 4 This embodiment provides a computer device including a processor and a memory. The memory stores a computer program, which includes program instructions. The processor executes the program instructions stored in the computer storage medium. The processor may be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. It is the computing and control core of the computer device, suitable for implementing one or more instructions, specifically suitable for loading and executing one or more instructions to achieve a corresponding method flow or function. The processor described in this embodiment can be used in the operation of a BO-GRU-based method for predicting and classifying the overheating trend of boiler water-cooled walls, including: acquiring multi-source operating parameters of the boiler and performing data preprocessing to obtain preprocessed operating data; calculating the normal wall temperature reference value of the water-cooled wall measuring points based on the preprocessed operating data; and comparing the real-time wall temperature with the normal wall temperature reference value. The temperature difference offset is obtained, and over-temperature features are extracted by combining them with historical temperature sequences. These features include temperature difference offset, temperature rise rate, regional temperature gradient, duration of high temperature, and intensity of temperature fluctuation. An over-temperature risk index is calculated based on these features. A time series sample set is constructed by combining the preprocessed operational data, historical wall temperature sequences, and the over-temperature risk index. This time series sample set is then input into the BO-GRU trend prediction model, which outputs a predicted temperature curve. The predicted over-temperature probability and the estimated time to reach the warning threshold are calculated based on the predicted temperature curve. The over-temperature risk index, predicted over-temperature probability, and predicted temperature peak value of the predicted temperature curve are compared with the preset threshold. Based on the comparison results, tiered warning information and coordinated response suggestions are output.

[0110] Example 5 This embodiment provides a computer-readable storage medium (Memory), which is a memory device in a computer device used to store programs and data. It is understood that the computer-readable storage medium here can include both the built-in storage medium in the computer device and extended storage media supported by the computer device. The computer-readable storage medium provides storage space that stores the operating system of the computer device. Furthermore, the storage space also stores one or more instructions suitable for loading and execution by a processor. These instructions can be one or more computer programs (including program code). It should be noted that the computer-readable storage medium here can be high-speed RAM or non-volatile memory, such as at least one disk storage device.

[0111] One or more instructions stored in the computer-readable storage medium can be loaded and executed by the processor to implement the corresponding steps of the boiler water-cooled wall overheating trend prediction and graded early warning method based on BO-GRU in the above embodiments; one or more instructions in the computer-readable storage medium are loaded and executed by the processor to perform the following steps: acquiring multi-source operating parameters of the boiler and performing data preprocessing to obtain preprocessed operating data; calculating the normal wall temperature reference value of the water-cooled wall measuring point based on the preprocessed operating data; comparing the real-time wall temperature with the normal wall temperature reference value to obtain the temperature difference offset, and extracting overheating feature quantities by combining historical temperature sequences. The parameters include temperature difference offset, temperature rise rate, regional temperature gradient, duration of high temperature, and intensity of temperature fluctuation; the over-temperature risk index is calculated based on the over-temperature characteristics; a time series sample set is constructed by combining preprocessed operational data, historical wall temperature sequences, and the over-temperature risk index; the time series sample set is input into the BO-GRU trend prediction model, and the predicted temperature curve is output; the predicted over-temperature probability and the expected time to reach the warning threshold are calculated based on the predicted temperature curve; the over-temperature risk index, the predicted over-temperature probability, and the predicted temperature peak of the predicted temperature curve are compared with the preset threshold, and graded warning information and joint response suggestions are output based on the comparison results.

[0112] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, optical storage, etc.) containing computer-usable program code.

[0113] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations 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, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0114] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0115] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0116] In the above embodiments of this application, the descriptions of each embodiment have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0117] The above description is only a preferred embodiment of this application. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of this application, and these improvements and modifications should also be considered within the scope of protection of this application.

[0118] It should be understood that the above description is for illustrative purposes and not for limitation. Many embodiments and applications beyond the examples provided will become apparent to those skilled in the art upon reading the above description.

Claims

1. A method for predicting and classifying the overheating trend of boiler water-cooled walls based on BO-GRU, characterized in that, The process includes the following: The boiler's multi-source operating parameters are acquired and preprocessed to obtain preprocessed operating data. Calculate the normal wall temperature reference value of the water-cooled wall measuring point based on the pre-processed operating data; The temperature difference offset is obtained by comparing the real-time wall temperature with the normal wall temperature reference value. The over-temperature characteristics are extracted by combining the historical temperature sequence. The over-temperature characteristics include temperature difference offset, temperature rise rate, regional temperature gradient, duration of high temperature, and intensity of temperature fluctuation. The over-temperature risk index is calculated based on the over-temperature characteristics. A time series sample set was constructed by combining preprocessed operational data, historical wall temperature sequences, and over-temperature risk index. Input the time series sample set into the BO-GRU trend prediction model and output the predicted temperature curve; Calculate the predicted over-temperature probability and the estimated time to reach the warning threshold based on the predicted temperature curve; The system compares the over-temperature risk index, the predicted over-temperature probability, and the predicted temperature peak of the predicted temperature curve with the preset threshold, and outputs graded early warning information and joint response suggestions based on the comparison results.

2. The method for predicting and classifying the overheating trend of boiler water-cooled walls based on BO-GRU according to claim 1, characterized in that, The multi-source operating parameters include the metal wall temperature at each measuring point of the water-cooled wall, the furnace temperature or the flue gas temperature at the furnace outlet, the boiler load, the main steam pressure, the main steam temperature, the feedwater flow rate, the medium flow rate, the primary air volume, the secondary air volume, the furnace oxygen content, the fuel quantity, the sootblower operation status, the flame image information, and the slagging and ash accumulation status information. Data preprocessing for multi-source operating parameters includes: resampling using a unified timestamp alignment method to synchronize the time of multi-source operating parameters; compensation for missing values ​​using forward padding, linear interpolation, or adjacent measurement point correlation compensation methods; removal of outliers using threshold removal, median filtering, or sliding window statistical methods; and obtaining normalized operating parameters by dividing the difference between the original operating parameters and the minimum value of the corresponding parameter by the difference between the maximum and minimum values ​​of the corresponding parameter.

3. The method for predicting and classifying the overheating trend of boiler water-cooled walls based on BO-GRU according to claim 1, characterized in that, The normal wall temperature reference values ​​for water-cooled wall measuring points are calculated based on the pre-processed operating data, including: Extract boiler load, feedwater flow rate, air volume parameters, furnace oxygen content, main steam pressure, main steam temperature, medium flow rate, and furnace temperature or furnace outlet flue gas temperature from the pre-processed operating data. The extracted parameters are input into the dynamic benchmark calculation model for processing, and the normal wall temperature reference value of the water-cooled wall measuring point under the current operating conditions is output.

4. The method for predicting and classifying the overheating trend of boiler water-cooled walls based on BO-GRU according to claim 1, characterized in that, Extracting overheat features from historical temperature sequences includes: The temperature rise rate is obtained by calculating the ratio of the temperature change of the water-cooled wall within the preset sliding time window to the time interval. The temperature gradient of the region is obtained by calculating the ratio of the real-time wall temperature difference between two adjacent water-cooled wall measuring points to the distance between the two measuring points. The average of the sum of squares of the differences between the water-cooled wall temperatures at multiple sampling points within a preset time window and the average water-cooled wall temperatures within the preset time window is used to obtain the intensity of temperature fluctuation. The duration of high temperature is obtained by counting the number of times the water-cooled wall temperature continuously exceeds the preset high temperature threshold.

5. The method for predicting and classifying the overheating trend of boiler water-cooled walls based on BO-GRU according to claim 1, characterized in that, The over-temperature risk index is calculated based on over-temperature characteristic quantities, including: Normalize the temperature difference offset, temperature rise rate, regional temperature gradient, high temperature duration, and temperature fluctuation intensity to obtain the normalized characteristic quantities corresponding to each overtemperature characteristic quantity. The overheating risk index of the water-cooled wall measuring point is obtained by multiplying each normalized feature quantity by its corresponding weight coefficient and then summing the results.

6. The method for predicting and classifying the overheating trend of boiler water-cooled walls based on BO-GRU according to claim 1, characterized in that, The steps for inputting a time series sample set into the BO-GRU trend prediction model and outputting the predicted temperature curve include: The preprocessed running data, historical wall temperature sequence, and over-temperature risk index are spliced ​​together to form a comprehensive feature vector; The input time series samples are constructed based on the preset sliding window length and the comprehensive feature vector, and the real values ​​of the water-cooled wall temperature at a preset number of future moments are combined as the output to form a time series sample set. The time series sample set is divided into a training set and a validation set; The prediction error of the validation set is used as the objective function of the Bayesian optimization algorithm. The hyperparameters of the GRU trend prediction model are optimized using the Bayesian optimization algorithm. The root mean square error of the objective function is obtained by calculating the square root of the average of the sum of squares of the differences between the predicted wall temperature and the actual wall temperature. The optimal combination of hyperparameters is configured into the BO-GRU trend prediction model to obtain the trained trend prediction model. The current time series sample is input into the trained trend prediction model, which outputs the predicted temperature curve.

7. The method for predicting and classifying the overheating trend of boiler water-cooled walls based on BO-GRU according to claim 1, characterized in that, The tiered early warning system includes Level 1 alert, Level 2 anomaly alert, and Level 3 danger alert. The specific logic includes: When the over-temperature risk index is greater than or equal to the first risk index threshold and less than the second risk index threshold, or the predicted over-temperature probability is greater than or equal to the first predicted over-temperature probability threshold and less than the second predicted over-temperature probability threshold, or the predicted temperature peak is greater than or equal to the first predicted temperature threshold and less than the second predicted temperature threshold, a first-level warning will be output. When the over-temperature risk index is greater than or equal to the second risk index threshold and less than the third risk index threshold, or the predicted over-temperature probability is greater than or equal to the second predicted over-temperature probability threshold and less than the third predicted over-temperature probability threshold, or the predicted temperature peak is greater than or equal to the second predicted temperature threshold and less than the third predicted temperature threshold, a level 2 abnormal warning will be output. When the over-temperature risk index is greater than or equal to the third risk index threshold, or the predicted over-temperature probability is greater than or equal to the third predicted over-temperature probability threshold, or the predicted temperature peak is greater than or equal to the third predicted temperature threshold, a level three danger warning is output. The first risk index threshold is less than the second risk index threshold and the second risk index threshold is less than the third risk index threshold; the first predicted over-temperature probability threshold is less than the second predicted over-temperature probability threshold and the second predicted over-temperature probability threshold is less than the third predicted over-temperature probability threshold; and the first predicted temperature threshold is less than the second predicted temperature threshold and the second predicted temperature threshold is less than the third predicted temperature threshold.

8. A boiler water-cooled wall overheating trend prediction and graded early warning system based on BO-GRU, characterized in that, include: The preprocessing module is used to acquire multi-source operating parameters of the boiler and perform data preprocessing to obtain preprocessed operating data. The dynamic benchmark calculation module is used to calculate the normal wall temperature reference value of the water-cooled wall measuring point based on the pre-processed running data; The feature extraction module is used to compare the real-time wall temperature with the normal wall temperature reference value to obtain the temperature difference offset, and extract the over-temperature feature quantity by combining it with the historical temperature sequence. The over-temperature feature quantity includes temperature difference offset, temperature rise rate, regional temperature gradient, high temperature duration and temperature fluctuation intensity; and calculate the over-temperature risk index based on the over-temperature feature quantity. The sample set construction module is used to combine preprocessed running data, historical wall temperature sequences, and over-temperature risk index to construct a time series sample set; The trend prediction module is used to input the time series sample set into a gated recurrent unit network model that combines a Bayesian optimization algorithm and output the predicted temperature curve. The probability and time calculation module is used to calculate the predicted over-temperature probability and the estimated time to reach the warning threshold based on the predicted temperature curve. The graded early warning module is used to compare the over-temperature risk index, the predicted over-temperature probability, and the predicted temperature peak of the predicted temperature curve with the preset threshold. Based on the comparison results, it outputs graded early warning information and linkage response suggestions.

9. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the method for predicting and classifying the overheating trend of boiler water-cooled walls based on BO-GRU as described in any one of claims 1 to 7.

10. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the steps of the method for predicting and classifying the overheating trend of boiler water-cooled walls based on BO-GRU as described in any one of claims 1 to 7.