Boiler air pre-heater temperature uniformity regulation and control system based on AI

Abstract

The invention relates to the technical field of thermal equipment operation state dynamic test and compensation control, in particular to an AI-based boiler air pre-heater temperature uniformity regulation and control system, which comprises a collaborative closed-loop control network of a data acquisition module, an AI prediction module, a heat storage system control module, a temperature compensation control module and a thermal stress analysis module. The data acquisition module acquires flue gas temperature, coal quality parameters and load instructions in real time, the flue gas temperature, the coal quality parameters and the load instructions are transmitted to the AI prediction module through an industrial bus, XGBoost, LightGBM and a deep neural network model are used for carrying out spatial-temporal feature extraction on the evolution trend of a temperature field, and the heat storage system control module dynamically switches heat storage and release modes according to the load state. The heat distribution proportion among the multiple units is adjusted by combining the three-dimensional stress field distribution data of the thermal stress analysis module; the bottleneck of lack of thermal inertia regulation of an existing system is broken through, dynamic balance of a temperature field under the condition that thermal stress of a heat transfer element is controllable is achieved, and the heat exchange efficiency and operation stability of the air pre-heater are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of dynamic testing and compensation control of the operating status of thermal equipment, and in particular to an AI-based boiler air preheater temperature uniformity control system. Background Art

[0002] The temperature uniformity test of the boiler air preheater belongs to the technical field of heat transfer equipment performance evaluation. Its core is to verify the matching of the heat transfer element and the flow field by detecting the air flow temperature distribution characteristics in different areas inside the air preheater, and then analyze the heat exchange efficiency and potential operating risks. During the test, a multi-point temperature sensor array is required to be arranged at the inlet and outlet sections of the air preheater to synchronously collect dynamic temperature data, and to construct a temperature gradient distribution model based on the countercurrent heat exchange characteristics of the flue gas and air. If the temperature deviation in a local area exceeds the design threshold, it indicates that there may be problems such as uneven air flow distribution or scaling of the heated surface. Such anomalies will lead to thermal stress concentration, accelerate fatigue damage of metal components and reduce heat recovery efficiency. The test results can provide a quantitative basis for optimizing the layout of the guide device or adjusting the soot blowing strategy.

[0003] The existing air preheater lacks a thermal inertia control mechanism and cannot achieve real-time compensation of heat flux density through dynamic flow distribution. Specifically, when the load of the existing boiler air preheater system suddenly changes, due to the insufficient cross-cycle storage capacity of the heat storage medium, the heat exchange rate between the flue gas side and the air side cannot match the transient heat load changes caused by the load fluctuation, resulting in an imbalance of heat flux density in the local area. For example, when the unit load drops suddenly, the flue gas waste heat drops sharply, but the air intake volume on the air side is not adjusted synchronously. At this time, the heat release function of the heat storage system is not activated in time, causing the front heating surface of the air side to produce a reverse temperature gradient due to the sudden drop in heat input, exacerbating the thermal stress concentration on the surface of the heat transfer element. Summary of the Invention

[0004] In response to the shortcomings of the existing technology, the present invention provides an AI-based boiler air preheater temperature uniformity control system. The present invention solves the problems of temperature field distribution instability and excessive thermal stress of heat transfer elements caused by the lack of thermal inertia control mechanism under load fluctuation conditions in existing air preheaters.

[0005] In order to solve the above technical problems, the specific technical solutions of the present invention are as follows: The AI-based boiler air preheater temperature uniformity control system provided by the present invention includes a heat storage system end, a self-circulation system end, a heat collection system end, an ash collection system, an ash removal system, an ammonium bisulfate capture system end, a warm air heating system end, and an optimized boiler rapid start-stop system end. The heat storage system end establishes a communication connection with the self-circulation system end, the heat collection system end, the ammonium bisulfate capture system end, the warm air heating system end, and the optimized boiler rapid start-stop system end. The heat storage system end includes; Data acquisition module, used to collect real-time flue gas temperature, coal quality information and dispatching load instructions from the flue gas side temperature measurement point, boiler coal quality monitoring equipment and dispatching load instruction receiving end; An AI prediction module is in communication with the data acquisition module, receives the real-time flue gas temperature, coal quality information, and dispatch load instructions, predicts the flue gas temperature and load using a preset machine learning model, generates prediction results, and outputs them to the heat storage system control module and the temperature compensation control module; The thermal stress analysis module communicates with the AI prediction module, receives the prediction results and the actual operating parameters of the heat transfer element, and calculates the thermal stress distribution through the finite element thermal-mechanical coupling model; A heat storage system control module is in communication with the AI prediction module and the thermal stress analysis module. It controls the heat input and output of the heat storage system based on the prediction results and the load status of the unit. It stores flue gas heat and residual steam heat under high load conditions, and outputs heat to improve boiler efficiency or assist in unit startup under low load conditions. It also fine-tunes heat distribution based on feedback from the thermal stress analysis module. A temperature compensation control module is in communication with the AI prediction module, receives the prediction results and the temperature monitoring data from the circulation system end, and controls the temperature uniformity of the flue gas side and the exhaust temperature by adjusting the forced circulation speed of the medium; The thermal stress analysis module is in communication with the AI prediction module, the heat storage system control module, and the temperature compensation control module. It analyzes the thermal stress distribution based on the prediction results and the actual operating parameters of the heat transfer element, and feeds the analysis results back to the heat storage system control module and the temperature compensation control module to adjust the heat output and the medium circulation speed.

[0006] Furthermore, in the AI-based boiler air preheater temperature uniformity control system of the present invention, the data acquisition module is configured as follows: Temperature data is collected from the temperature measuring points on the flue gas side, the coal element content and combustion characteristic parameters are obtained through coal quality monitoring equipment, the load adjustment instructions are obtained in real time from the dispatch instruction receiving end, and the data are integrated and transmitted to the AI prediction module.

[0007] Furthermore, in the AI-based boiler air preheater temperature uniformity control system of the present invention, the preset machine learning models in the AI prediction module include XGBoost, LightGBM and deep neural network; The flue gas temperature change trend is predicted based on coal quality information and load instructions, and high-resolution load prediction results are generated. The prediction results are transmitted to the heat storage system control module for heat storage strategy adjustment and to the temperature compensation control module for medium circulation speed optimization.

[0008] Furthermore, in the AI-based boiler air preheater temperature uniformity control system of the present invention, the heat storage system control module is configured as follows: Under high load conditions, the flue gas heat and residual steam heat are input into the heat storage system for storage. Under low load conditions, the heat is output to the warm air heating system to increase the inlet air temperature, or to the optimized boiler rapid start-stop system to accelerate the start-up of the unit. Based on the stress distribution data of heat transfer elements fed back by the thermal stress analysis module, the heat distribution ratio among multiple units is dynamically adjusted.

[0009] Furthermore, in the AI-based boiler air preheater temperature uniformity control system of the present invention, the temperature compensation control module is configured as follows: Based on the flue gas temperature distribution data fed back from the circulation system, the medium circulation speed in the high-temperature zone is controlled to accelerate heat absorption. The stress threshold data of the thermal stress analysis module is simultaneously received to limit the adjustment range of the medium circulation speed to avoid excessive thermal stress.

[0010] Furthermore, in the AI-based boiler air preheater temperature uniformity control system of the present invention, the thermal stress analysis module is configured as follows: Based on the difference between the predicted temperature and the actual operating temperature by the AI prediction module, the instantaneous thermal stress distribution of the heat transfer element is calculated, and a stress adjustment instruction is generated and transmitted to the heat storage system control module to reduce heat output, or transmitted to the temperature compensation control module to reduce the medium circulation speed.

[0011] Furthermore, in the AI-based boiler air preheater temperature uniformity control system of the present invention, the ammonium bisulfate capture system includes: The corrosion-resistant independent circulation pipeline maintains the temperature in the low-temperature section of the air preheater within the ammonium bisulfate condensation range. The heat is input through the heat storage system to gasify the condensed ammonium bisulfate. The soot blowing device is used to remove the lumps on the pipe wall, and the gaseous ammonium bisulfate is transported to the ash removal system through the ash collection system.

[0012] Beneficial effects of the present invention: The present invention realizes dynamic balanced control of the air preheater temperature field by constructing a closed-loop control network consisting of a data acquisition module, an AI prediction module, a heat storage system control module, a temperature compensation control module, and a thermal stress analysis module. The AI prediction module uses XGBoost, LightGBM, and a deep neural network model to extract spatiotemporal features of flue gas temperature and load fluctuations, generating high-resolution prediction results to drive the heat storage and release strategy of the heat storage system, and compensating for the control lag caused by thermal inertia through a feedforward mechanism. The thermal stress analysis module converts the difference between the predicted temperature and the measured temperature into a three-dimensional stress field distribution based on a finite element thermomechanical coupling model, dynamically adjusts the heat distribution ratio and medium circulation speed of the heat storage system, and realizes temperature gradient optimization under the stress threshold constraint of the heat transfer element. The ammonium bisulfate capture system controls the removal of scale deposits in the low-temperature section through a corrosion-resistant pipeline and heat linkage with the heat storage system. The multiple modules work together to break through the bottleneck of heat flux density imbalance in traditional systems when the load suddenly changes, thereby improving the heat exchange efficiency of the air preheater and extending the service life of the heat transfer element. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] In order to more clearly illustrate the technical solution of the present invention, the following is a brief introduction to the drawings required for use in the embodiments. Obviously, for ordinary technicians in this field, other drawings can be obtained based on the drawings without paying any creative labor.

[0014] Figure 1 An AI-based boiler air preheater temperature uniformity control system is provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0015] In order to make the purpose, technical solutions and advantages of the present invention clearer, the technical solutions of the present invention will be clearly and completely described below in conjunction with the specific embodiments of the present invention and the corresponding drawings. Obviously, the embodiments described are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention. The technical solutions provided by each embodiment of the present invention are described in detail below in conjunction with the drawings. In order to better understand the purpose of the present invention, the present invention is further described in detail below.

[0016] See also Figure 1 The AI-based boiler air preheater temperature uniformity control system provided by the present invention includes a heat storage system end, a self-circulation system end, a heat collection system end, an ash collection system, an ash removal system, an ammonium bisulfate capture system end, a warm air heating system end, and an optimized boiler rapid start-stop system end. The heat storage system end establishes a communication connection with the self-circulation system end, the heat collection system end, the ammonium bisulfate capture system end, the warm air heating system end, and the optimized boiler rapid start-stop system end. The heat storage system end includes; Data acquisition module, used to collect real-time flue gas temperature, coal quality information and dispatching load instructions from the flue gas side temperature measurement point, boiler coal quality monitoring equipment and dispatching load instruction receiving end; An AI prediction module is in communication with the data acquisition module, receives the real-time flue gas temperature, coal quality information, and dispatch load instructions, predicts the flue gas temperature and load using a preset machine learning model, generates prediction results, and outputs them to the heat storage system control module and the temperature compensation control module; The thermal stress analysis module communicates with the AI prediction module, receives the prediction results and the actual operating parameters of the heat transfer element, and calculates the thermal stress distribution through the finite element thermal-mechanical coupling model; A heat storage system control module is in communication with the AI prediction module and the thermal stress analysis module. It controls the heat input and output of the heat storage system based on the prediction results and the load status of the unit. It stores flue gas heat and residual steam heat under high load conditions, and outputs heat to improve boiler efficiency or assist in unit startup under low load conditions. It also fine-tunes heat distribution based on feedback from the thermal stress analysis module. A temperature compensation control module is in communication with the AI prediction module, receives the prediction results and the temperature monitoring data from the circulation system end, and controls the temperature uniformity of the flue gas side and the exhaust temperature by adjusting the forced circulation speed of the medium; The thermal stress analysis module is in communication with the AI prediction module, the heat storage system control module, and the temperature compensation control module. It analyzes the thermal stress distribution based on the prediction results and the actual operating parameters of the heat transfer element, and feeds the analysis results back to the heat storage system control module and the temperature compensation control module to adjust the heat output and the medium circulation speed.

[0017] The data acquisition module acquires real-time temperature distribution data through a distributed array of temperature sensors placed on the flue gas side of the air preheater. The temperature sensor array consists of thermocouples and infrared temperature measuring devices to synchronously collect dynamic temperature values at different measuring points. The coal quality monitoring equipment uses an online spectrometer to continuously detect the elemental content and calorific value of the coal, generating coal combustion characteristic parameters. The dispatch load instruction receiver communicates with the power plant's DCS system via the OPC protocol to obtain real-time unit load adjustment instructions. The data acquisition module aligns the timestamps of the above temperature data, coal quality parameters, and load instructions, and then transmits them to the AI prediction module via the industrial bus.

[0018] The machine learning model built into the AI prediction module was trained using historical operating data. The training data includes flue gas temperature sequences under different coal quality conditions, load instruction change curves, and corresponding actual temperature response data. The model extracts features from the input real-time data and uses the XGBoost algorithm to predict flue gas temperature change trends within the future time window. It also uses a deep neural network to analyze the correlation between load instructions and temperature fluctuations to generate load forecast results with high time resolution. The forecast results are divided into short-term forecast data and medium-term forecast data, which are transmitted to the thermal storage system control module for thermal storage medium scheduling planning and to the temperature compensation control module for pre-adjustment of medium circulation parameters.

[0019] Based on the predicted high-load periods, the thermal storage system control module activates the phase-change thermal storage unit in the thermal storage system to absorb excess heat from the flue gas and the main steam pipeline. The thermal storage medium is a molten salt and ceramic composite thermal storage material. During low-load periods, the control module activates the thermal storage medium's heat release circuit, transferring the stored heat via a heat exchanger to the air preheating line in the warm air heating system. Simultaneously, based on the startup requirements of the optimized boiler rapid start-stop system, some heat is allocated to the air duct in front of the burner to accelerate fuel ignition. The thermal storage system control module receives the stress distribution map of the heat transfer element from the thermal stress analysis module. If the stress value in a local area approaches the material's yield limit, the module automatically reduces the heat output rate in that area.

[0020] The temperature compensation control module controls the circulation flow rate of the medium, which is high-temperature-resistant silicone oil, by adjusting the speed of the variable-frequency centrifugal pump on the self-circulating system side. This module receives data from a fiber-optic temperature sensor located in the high-temperature zone of the air preheater on the self-circulating system side. If the standard deviation of the temperature distribution exceeds a set threshold, it increases the pumping frequency in the corresponding pipeline in the high-temperature zone to increase the medium circulation speed, accelerate heat absorption, and reduce temperature gradients. Simultaneously, the module obtains real-time information on the allowable stress fluctuation range calculated by the thermal stress analysis module. If the stress change rate exceeds a safety threshold due to adjustments to the medium circulation speed, a speed limiting mechanism is triggered to maintain the current flow rate.

[0021] Based on a finite element thermomechanical coupling model, the thermal stress analysis module spatially maps the temperature prediction data output by the AI prediction module with the measured temperature data at the self-circulating system end to calculate the transient temperature field distribution on the heat transfer element surface. By solving the thermoelastic equations, an analysis report is generated that includes the location of stress concentration areas and the rate of stress change. This module transmits a stress overlimit warning signal to the thermal storage system control module to trigger heat distribution optimization. It also sends a medium circulation speed correction factor to the temperature compensation control module, forming a closed-loop control system.

[0022] The ammonium bisulfate capture system features a stainless steel serpentine cooling pipeline in the low-temperature section of the air preheater. By adjusting the cooling medium flow rate to maintain pipe wall temperature, the ammonium bisulfate vapor in the flue gas condenses on the outer walls of the pipeline. When the ash accumulation reaches a detection threshold, the thermal storage system switches to delivering high-temperature medium to the serpentine pipeline, causing the condensate to vaporize and separate from the pipe walls. The vaporized ammonium bisulfate vapor is then transported through a negative pressure ash collection pipeline to an electrostatic precipitator for targeted capture and removal.

[0023] Through real-time data exchange and command collaboration, each module builds a dynamic balance control network for the air preheater temperature field. The data acquisition module provides basic monitoring data, the AI prediction module generates feedforward control parameters, the heat storage system and temperature compensation module optimizes heat storage and distribution, the thermal stress analysis module implements safety boundary constraints, and the ammonium bisulfate capture module addresses low-temperature corrosion. Ultimately, this achieves active temperature uniformity control while controlling the thermal stress of the heat transfer element.

[0024] Specifically, in the AI-based boiler air preheater temperature uniformity control system of the present invention, the data acquisition module is configured as follows: Temperature data is collected from the temperature measuring points on the flue gas side, the coal element content and combustion characteristic parameters are obtained through coal quality monitoring equipment, the load adjustment instructions are obtained in real time from the dispatch instruction receiving end, and the data are integrated and transmitted to the AI prediction module.

[0025] The data acquisition module deploys a distributed temperature sensor array on the flue gas side of the air preheater. The sensor array consists of K-type thermocouples and infrared temperature measuring devices. The thermocouples are installed at upstream, midstream, and downstream measuring points in the flue gas flow section, and the infrared temperature measuring devices are located in observation windows on the outer wall of the tubular air preheater for non-contact temperature monitoring. The coal quality monitoring equipment integrates an online laser-induced breakdown spectroscopy analyzer to perform real-time elemental content testing on the coal conveyor belt. A thermogravimetric analysis unit is also used to simultaneously obtain coal combustion characteristic parameters, including volatile matter release rate and fixed carbon reaction activity. The dispatch instruction receiver communicates with the power plant's distributed control system via the OPC-UA protocol, analyzing the target load value and adjustment rate parameters in the unit load adjustment instructions in real time.

[0026] The data integration process uses an Industrial Internet of Things (IIoT) time-series database to synchronize temperature data, coal quality parameters, and load instructions. Temperature data is stored at a sampling rate of one second, coal quality parameters are cached at a minute-by-minute update rate, and load instructions are updated instantly based on DCS system signals. The integrated dataset is encapsulated via the Modbus-TCP protocol and transmitted to the AI prediction module's data preprocessing unit. This unit performs linear interpolation to fill in missing data and applies a sliding window mean filter to outliers, generating a standardized input data stream.

[0027] The standardized data stream is fed into the feature extraction layer of the AI prediction module. This layer decomposes the high-frequency fluctuation and low-frequency trend components of the temperature series through wavelet transform. The sulfur and ash content data in the coal quality parameters are normalized and concatenated with the load command's adjustment gradient parameters to form a multidimensional feature vector. This feature vector is then passed through a temporal convolutional network to capture local features. The resulting output is high-level representation data that includes the spatiotemporal distribution of flue gas temperature and the correlation between load and coal quality, providing the input foundation for subsequent prediction models.

[0028] Specifically, in the AI-based boiler air preheater temperature uniformity control system of the present invention, the preset machine learning models in the AI prediction module include XGBoost, LightGBM and deep neural network; The flue gas temperature change trend is predicted based on coal quality information and load instructions, and high-resolution load prediction results are generated. The prediction results are transmitted to the heat storage system control module for heat storage strategy adjustment and to the temperature compensation control module for medium circulation speed optimization.

[0029] The AI prediction module's machine learning model is trained using a historical boiler operation dataset. The training data includes flue gas temperature time series data under different coal quality conditions, load instruction change curves, and corresponding heat transfer element temperature response records. The XGBoost model selects features based on sulfur, ash, and volatile matter content in coal quality parameters to generate a weight matrix that correlates coal combustion characteristics with temperature fluctuations. The LightGBM model processes the gradient change data of load instructions and extracts thermal inertia characteristics under load mutation scenarios using a time series decomposition algorithm. The deep neural network uses an encoder-decoder architecture, with the encoder performing convolution processing on the spatial distribution data of flue gas temperature, and the decoder outputting a predicted temperature field map within a future time window.

[0030] During the prediction process, the output feature vectors of the XGBoost and LightGBM models are fused through a feature concatenation layer and fed into a deep neural network's temporal convolution model for spatiotemporal feature extraction, generating a high-resolution load forecast. This forecast is divided into short-term and medium-term prediction data. The short-term prediction data includes the flue gas temperature gradient trend over the next five minutes and is transmitted to the thermal storage system control module via the OPC-UA protocol to trigger the dynamic charging and discharging strategy of the thermal storage medium. The medium-term prediction data covers the load fluctuation pattern over the next 30 minutes and is transmitted to the temperature compensation control module via the Modbus-TCP protocol to drive the pre-adjustment mechanism of the medium circulation speed.

[0031] The temporal convolution model uses a dilated causal convolution structure to process time series data. Multi-layer atrous convolutions expand the receptive field to capture long-term dependencies, while an attention mechanism is introduced to weight load instruction mutation points. The decoder outputs a temperature field prediction map, which is then compared with real-time monitoring data, for error correction via residual connections. The residual connection layer uses gated recurrent units to dynamically compensate for prediction deviations, improving the model's prediction accuracy in scenarios with rapidly changing loads.

[0032] The prediction results undergo structured processing before transmission. Short-term prediction data is encapsulated as a set of control instructions containing parameters such as heat storage rate and medium flow demand. The thermal storage system control module interprets the instructions and activates the heat input valve opening adjustment of the phase change thermal storage unit. Medium-term prediction data is converted into a matrix of medium circulation speed adjustment coefficients. The temperature compensation control module generates the speed control signal for the variable frequency centrifugal pump based on the regional weight parameters in the matrix. The transmission process uses the Industrial Ethernet communication protocol. The data packets are accompanied by timestamps and data checksums, which are synchronized with the clock signals of the thermal storage system control module and the temperature compensation control module, forming a feedforward and feedback composite control link.

[0033] Specifically, in the AI-based boiler air preheater temperature uniformity control system of the present invention, the heat storage system control module is configured as follows: Under high load conditions, the flue gas heat and residual steam heat are input into the heat storage system for storage. Under low load conditions, the heat is output to the warm air heating system to increase the inlet air temperature, or to the optimized boiler rapid start-stop system to accelerate the start-up of the unit. Based on the stress distribution data of heat transfer elements fed back by the thermal stress analysis module, the heat distribution ratio among multiple units is dynamically adjusted.

[0034] The heat storage system control module activates the phase-change heat storage unit under high-load conditions. This unit utilizes a honeycomb-shaped heat storage body made of molten salt and ceramic-based composite thermal storage materials. Waste heat from the flue gas is transferred to the heat storage body via a radiation heat exchanger for storage. Waste heat from the main steam line is indirectly exchanged with the heat storage medium via a fin-and-tube heat exchanger. The temperature gradient of the heat storage medium is monitored in real time by a distributed fiber-optic temperature measurement system. When the critical phase change temperature of the heat storage unit is detected, the control module closes the flue gas inlet valve and activates the aerogel insulation device in the thermal insulation layer, completing closed-loop control of the heat storage phase.

[0035] Under low-load conditions, the control module switches to heat output mode. The heat stored in the thermal storage medium is exchanged with the air preheating pipeline on the warm air heating system side through a plate heat exchanger. The outlet air temperature of the heat exchanger is precisely controlled by a PID controller that adjusts the cold air mixing ratio. A multi-channel diversion design is used to optimize heat delivery to the boiler's rapid start-stop system. The heat distribution valve group dynamically adjusts the heat delivery ratio to the burner's front air duct and the tail flue reheater based on the air temperature demand curve during the unit's startup phase, shortening the unit startup time and reducing unburned carbon generation.

[0036] The stress distribution data fed back by the thermal stress analysis module is transmitted to the heat storage system control module via industrial Ethernet. The dynamic allocation algorithm built into the control module analyzes the coordinates of high-risk areas in the stress distribution map and generates heat output limit instructions for the corresponding unit's heat storage unit. The heat distribution ratio between multiple units is achieved through cross-system collaboration through the OPC-UA protocol. When the stress value of the heat transfer element of a unit approaches the safety threshold, the control module automatically increases the output flow of the heat storage medium of other units to maintain the balance of the total heat supply while reducing the risk of local thermal stress concentration. The heat distribution instruction and the feedback data from the heat storage medium flow meter form a closed-loop verification mechanism. When the flow meter detects that the actual output value deviates from the instruction by more than 5%, it triggers an alarm and starts the backup pump group to compensate for the flow.

[0037] The control module integrates a safety interlock protection function. When the thermal stress analysis module sends an emergency shutdown signal, it immediately cuts off the heat output path of all thermal storage units and simultaneously activates the emergency cooling system to inject low-temperature nitrogen into the thermal storage medium circuit for rapid cooling. The switching between the emergency cooling system and the regular heat output system is achieved through a three-way solenoid valve group. The valve group status is controlled by the control module's FPGA logic controller with millisecond-level response, forming a multi-level safety protection system.

[0038] Specifically, in the AI-based boiler air preheater temperature uniformity control system of the present invention, the temperature compensation control module is configured as follows: Based on the flue gas temperature distribution data fed back from the circulation system, the medium circulation speed in the high-temperature zone is controlled to accelerate heat absorption. The stress threshold data of the thermal stress analysis module is simultaneously received to limit the adjustment range of the medium circulation speed to avoid excessive thermal stress.

[0039] The temperature compensation control module acquires flue gas temperature distribution data from an array of fiber-optic temperature sensors deployed at the end of the circulation system. The sensors are evenly spaced along the axial direction of the air preheater's high-temperature tube bundle, utilizing wavelength division multiplexing technology to enable parallel monitoring of multiple measurement points. Temperature data is transmitted via the RS-485 bus to the control module's data processing unit, which uses a moving average filtering algorithm to eliminate measurement noise and generate the standard deviation of the temperature distribution, which serves as a benchmark parameter for media circulation speed control.

[0040] The medium circulation speed is controlled using a variable frequency drive centrifugal pump unit. The control module's built-in fuzzy PID controller receives the temperature distribution standard deviation parameter and generates a pump speed adjustment command. When the temperature standard deviation in a high-temperature zone exceeds a set threshold, the controller increases the speed of the centrifugal pump in the corresponding zone, increasing the forced circulation flow of the high-temperature silicone oil medium, accelerating heat absorption from the flue gas, and reducing temperature gradients. The speed adjustment command is transmitted to the inverter via the Profibus-DP protocol, which outputs a three-phase PWM modulation signal to drive the centrifugal pump motor.

[0041] Stress threshold data transmitted by the thermal stress analysis module is transmitted via industrial Ethernet to the control module's safety constraint unit, which converts the stress threshold into an upper limit for the medium's circulation speed. The control module's speed control command generation unit compares the current speed setting with the upper limit in real time. If it detects that the setting exceeds the safety threshold, it automatically limits the speed command to within the permitted range. This limit status is fed back to the main control system via a hardwired signal, triggering a warning on the human-machine interface.

[0042] The medium circulation pipeline is equipped with a pressure differential compensation device. When a sudden change in the centrifugal pump speed causes pipeline pressure fluctuations, the pressure differential compensation device uses a bidirectional hydraulic cylinder to adjust the opening of the branch valve to maintain overall system pressure balance. The pressure balance signal is fed back to the control module via the HART protocol, forming a composite control loop with the speed adjustment command to prevent water hammer caused by sudden flow changes. The pipeline also integrates a solenoid control valve group, which dynamically adjusts the medium distribution ratio of each branch pipeline based on temperature distribution data, achieving precise control of the heat absorption rate in each local area.

[0043] The control module establishes a data exchange channel with the thermal storage system control module. When the thermal stress analysis module detects stress concentration in the heat transfer element, it synchronously adjusts the medium circulation speed and the heat output ratio of the thermal storage system. The two modules utilize a timestamp synchronization mechanism, using the IEEE 1588 precision clock protocol to maintain the timing consistency of control commands, forming a cross-system collaborative temperature control network. Historical control command data is stored on an industrial-grade solid-state drive, supporting fault tracing and iterative optimization of control parameters.

[0044] Specifically, in the AI-based boiler air preheater temperature uniformity control system of the present invention, the thermal stress analysis module is configured as follows: Based on the difference between the predicted temperature and the actual operating temperature by the AI prediction module, the instantaneous thermal stress distribution of the heat transfer element is calculated, and a stress adjustment instruction is generated and transmitted to the heat storage system control module to reduce heat output, or transmitted to the temperature compensation control module to reduce the medium circulation speed.

[0045] The thermal stress analysis module collects actual operating temperature data from the heat transfer element's surface using a distributed array of fiber-optic temperature sensors. The sensor array utilizes wavelength division multiplexing (WDM) technology for simultaneous monitoring of multiple measurement points, with a data acquisition frequency of 10 times per second. The actual temperature data is aligned with the predicted temperature dataset transmitted by the AI prediction module using a timestamp matching module. This matching module uses a sliding window algorithm to eliminate clock drift errors and generate a temperature difference matrix as input for stress calculations.

[0046] The temperature difference matrix is input into a finite element thermomechanical coupling analysis model, which includes a built-in database of heat transfer element material properties, including elastic modulus, thermal expansion coefficient, and Poisson's ratio. The analysis model uses an explicit time integration algorithm to convert temperature gradient data into a three-dimensional transient thermal stress field distribution. The coordinates of stress concentration areas and stress change rates are annotated in a cloud-like format. After dimensionality reduction, the stress field distribution data is converted into a two-dimensional projection map, facilitating rapid interpretation of control commands.

[0047] The stress adjustment command generation unit receives the stress field distribution map and uses a threshold comparison algorithm to identify the coordinates of areas where the material's allowable stress value exceeds the specified value. It then generates hierarchical control commands for these areas: a primary command triggers the thermal storage system control module to reduce the heat output rate of the corresponding unit's thermal storage unit; a secondary command drives the temperature compensation control module to reduce the medium circulation rate in the corresponding pipeline. Command priority is dynamically assigned based on the degree of stress exceeding the specified value, and when the stress change rate exceeds a preset threshold, it automatically escalates to an emergency control command.

[0048] Command transmission utilizes a hard real-time industrial Ethernet protocol, and the transmission protocol stack integrates CRC checksums and retransmission mechanisms to ensure the integrity and timeliness of control signals. After receiving the command, the thermal storage system control module and the temperature compensation control module analyze the region coordinates and adjustment parameters contained in the command using an FPGA logic controller, ensuring actuator response time within 50 milliseconds. Execution status feedback signals are returned to the thermal stress analysis module via a shared memory area, forming a closed-loop control verification mechanism.

[0049] The safety protection unit monitors the stress field analysis process in real time. If the stress value fails to fall back to the safe range after three consecutive executions, it triggers a coordinated multi-module emergency response mechanism. This mechanism simultaneously sends a full heat output cutoff command to the thermal storage system control module and a maximum cycle speed command to the temperature compensation control module. Simultaneously, it activates the independent emergency cooling system for forced air cooling of the heat transfer elements. The emergency response timing of each module is coordinated through a precise clock synchronization protocol to avoid control conflicts.

[0050] Historical stress data is stored in an industrial-grade solid-state storage array using a time-series database format, supporting multi-dimensional retrieval by region coordinates, stress range, and other factors. Complete stress field distribution data is regularly backed up to a remote server using an incremental storage strategy, retaining the full data for the last three months for subsequent model optimization and training.

[0051] Specifically, the AI-based boiler air preheater temperature uniformity control system of the present invention, the ammonium bisulfate capture system end includes: The corrosion-resistant independent circulation pipeline maintains the temperature in the low-temperature section of the air preheater within the ammonium bisulfate condensation range. The heat is input through the heat storage system to gasify the condensed ammonium bisulfate. The soot blowing device is used to remove the lumps on the pipe wall, and the gaseous ammonium bisulfate is transported to the ash removal system through the ash collection system.

[0052] The ammonium bisulfate capture system features a stainless steel serpentine cooling pipe in the low-temperature section of the air preheater. The pipe's exterior is sprayed with a silicon carbide wear-resistant coating for enhanced corrosion resistance. The cooling medium is an ethylene glycol-water solution, which exchanges heat with the heat medium in the thermal storage system via a plate heat exchanger, maintaining the pipe wall temperature within the ammonium bisulfate crystallization temperature range.

[0053] Condensate thickness is detected using a laser ranging sensor array. The sensors are mounted at a 45-degree angle of incidence on the outer observation window of the pipeline and measure the change in the distance between the agglomerate surface and the pipe wall. When the agglomerate thickness reaches a preset threshold, the control module activates the heat transfer program on the thermal storage system side. High-temperature heat medium is introduced into the serpentine pipeline through a switching valve group, raising the pipe wall temperature to above the vaporization point of ammonium bisulfate. The heat transfer rate is regulated by a PID controller, dynamically adjusting the heat medium flow rate based on the agglomerate thickness gradient.

[0054] The sootblowing system comprises multiple sets of rotatable high-pressure nozzles, evenly spaced along the pipeline axis. A gear transmission mechanism delivers circumferentially sweeping spray. High-pressure steam is generated by a steam generator driven by the thermal energy of the heat storage system. The spray pressure is adaptively adjusted based on the resistance to agglomerate removal. Once the sootblowing command is triggered, the nozzles pulse the target area along a preset trajectory, releasing agglomerated debris and entering the ash collection system along the flue gas flow.

[0055] The ash collection system uses a Venturi negative pressure generator to create a directional airflow channel. Under negative pressure, vaporized ammonium bisulfate vapor and stripped solid particles enter a cyclone separator. The inner wall of the cyclone separator is coated with a PTFE anti-corrosion coating, and centrifugal force is used to achieve gas-solid separation. The separated gas is purified in an activated carbon adsorption tower before being discharged into the flue, while the solid particles fall into a closed ash hopper. A pneumatic ash discharge valve is installed at the bottom of the hopper, and the discharge frequency is dynamically adjusted based on data from the hopper weight sensor. The ash is then transported to the power plant's ash removal system via a screw conveyor.

[0056] System operating data is stored in the industrial controller's circular buffer, including tube wall temperature curves, agglomerate thickness variation data, and sootblowing operation logs. This data is uploaded to the power plant's DCS system via the OPC-UA protocol and correlated with boiler operating parameters for analysis, providing fundamental data support for research on the ammonium bisulfate formation mechanism. Under abnormal operating conditions, emergency bypass mode is activated, cutting off heat input to the thermal storage system and initiating a compressed air purge circuit to prevent further blockage.

[0057] Explanation of each model in the technical solution of the present invention: The present invention integrates three types of machine learning models, XGBoost, LightGBM, and deep neural network (DNN), into the AI prediction module. These three models work together to achieve high-precision prediction of the air preheater temperature field evolution trend and load fluctuation. The specific functions and collaborative mechanisms are as follows: XGBoost model: Extraction of correlation features between coal quality and temperature; XGBoost (Extreme Gradient Boosting) is an ensemble learning algorithm based on decision trees. Its core is to gradually correct the prediction error by combining multiple weak decision trees, and finally form a strong prediction model. In this paper, the XGBoost model is mainly used to analyze the correlation between coal quality parameters and flue gas temperature fluctuations: Input data: coal element content (such as sulfur, ash, volatile matter) and combustion characteristic parameters (such as volatile matter release rate, fixed carbon reaction activity).

[0058] Processing logic: Use feature selection algorithms (such as the Gini coefficient) to identify coal quality parameters that have a significant impact on temperature (for example, high-sulfur coal may aggravate low-temperature corrosion and affect local temperature distribution), generate a correlation weight matrix between coal combustion characteristics and temperature fluctuations, and quantify the degree of influence of different coal quality conditions on the temperature field.

[0059] Output function: Provide key features of coal quality for subsequent models, and assist in predicting the changing trend of flue gas temperature under different coal types.

[0060] LightGBM model: thermal inertia feature extraction for load mutation scenarios; LightGBM (Lightweight Gradient Boosting Machine) is an efficient algorithm based on the gradient boosting framework. Through histogram optimization and unilateral gradient sampling technology, it significantly improves computational efficiency while maintaining high accuracy. In this paper, the LightGBM model focuses on the dynamic change analysis of load instructions: Input data: dispatching load instructions (target load value, regulation rate) and historical load fluctuation curve.

[0061] Processing logic: Use time series decomposition algorithms (such as STL decomposition) to separate long-term trends, seasonal fluctuations, and random noise in load instructions, focusing on extracting thermal inertia characteristics (i.e., the delayed characteristics of the temperature field response after load changes) in scenarios with sudden load changes (such as sudden increase / decrease in unit load).

[0062] Output function: Identify the "mutation point" of load fluctuation and the corresponding thermal inertia parameters (such as temperature response lag time and fluctuation amplitude), providing a time-dimensional decision-making basis for the heat scheduling of the thermal storage system.

[0063] Deep Neural Network (DNN): Modeling and prediction of spatiotemporal characteristics of temperature fields; Deep neural networks (such as temporal convolutional networks with encoder-decoder architectures) have powerful nonlinear modeling capabilities and are suitable for processing high-dimensional, unstructured spatiotemporal sequence data. In this invention, the DNN model is responsible for the final prediction of the temperature field: Input data: coal quality association weight matrix output by XGBoost, load thermal inertia characteristics extracted by LightGBM, and flue gas temperature spatial distribution data (such as thermocouples and infrared temperature measurement data in different areas of the air preheater).

[0064] Processing logic: The encoder extracts the spatial distribution features of the temperature field (such as the location of high-temperature areas and temperature gradients) through convolutional layers, and captures the long-term dependencies of the temperature sequence (such as the impact of temperature changes in the previous hour on the next 30 minutes) through temporal convolutional layers (including dilated causal convolutions). The decoder weights key features such as load mutation points and coal quality sensitive areas based on the attention mechanism, and outputs a temperature field prediction map within the future time window (including the temperature values and gradient distribution of each region).

[0065] Output function: Generate short-term (next 5 minutes) temperature gradient change trend and medium-term (next 30 minutes) load fluctuation pattern prediction results, which are used to trigger the dynamic charging and discharging strategy of the thermal storage system (short-term data) and drive the medium circulation pre-adjustment of the temperature compensation module (medium-term data).

[0066] Synergy is achieved through "feature concatenation and spatiotemporal fusion": XGBoost and LightGBM extract key features from the coal quality and load dimensions, respectively, and then fuse them into a multidimensional feature vector through a feature concatenation layer. This vector is input into the DNN's temporal convolution model and, combined with the spatial distribution data of the temperature field, ultimately generates high-resolution temperature field and load forecast results. This design leverages the efficient feature extraction capabilities of traditional gradient boosting algorithms for structured data (coal quality and load instructions) and the nonlinear modeling advantages of deep neural networks for unstructured spatiotemporal data (temperature field distribution), jointly supporting dynamic decision-making in the thermal storage system control module and the temperature compensation control module.

[0067] The present invention solves the problems of temperature field instability and thermal stress exceeding the limit in the air preheater under load fluctuations by constructing a closed-loop control system with multi-module collaboration. The data acquisition module obtains the flue gas temperature, coal quality parameters and load instructions in real time, and transmits them to the AI prediction module through the industrial bus. The AI prediction module uses XGBoost, LightGBM and deep neural network models to extract spatiotemporal features of the temperature field evolution trend and generate high-resolution load prediction results. The prediction data is divided into short-term and medium-term categories. Short-term data triggers the heat storage strategy of the heat storage system control module, and medium-term data drives the medium circulation pre-adjustment of the temperature compensation control module, forming a feedforward control mechanism to offset the thermal inertia delay effect.

[0068] The thermal storage system control module dynamically switches heat storage and release modes based on load conditions. Under high load, the phase-change thermal storage unit absorbs waste heat from flue gas and main steam. Under low load, it distributes heat to the warm air heating system and rapid start-stop system. Based on a finite element thermomechanical coupling model, the thermal stress analysis module converts the difference between the AI-predicted temperature and the measured temperature of the heat transfer element into a three-dimensional stress field distribution, generating hierarchical control instructions. Upon receiving these instructions, the thermal storage system control module dynamically adjusts the heat distribution ratio among multiple units. The temperature compensation control module simultaneously limits the adjustment range of the medium circulation speed, achieving dynamic matching of the heat output rate with the stress threshold of the heat transfer element.

[0069] The system utilizes hard real-time Industrial Ethernet to establish a closed-loop verification mechanism. The thermal storage system control module, temperature compensation control module, and thermal stress analysis module utilize the IEEE 1588 clock protocol to maintain command synchronization. When the thermal stress analysis module detects an area of stress concentration, it triggers the emergency cooling system to inject cryogenic nitrogen for forced cooling. It also adjusts the spatiotemporal coupling between the medium circulation rate and the thermal storage heat output. This synergistic effect of multiple modules achieves dynamic temperature balancing while maintaining controllable thermal stress in the heat transfer element, overcoming the control lag bottleneck inherent in existing systems due to a lack of thermal inertia regulation.

Claims

1. The AI-based boiler air preheater temperature uniformity control system is characterized by: Including the heat storage system end, the self-circulation system end, the heat collection system end, the ash collection system end, the ash removal system end, the ammonium bisulfate capture system end, the warm air heating system end and the optimized boiler rapid start and stop system end, the heat storage system end establishes communication connections with the self-circulation system end, the heat collection system end, the ammonium bisulfate capture system end, the warm air heating system end and the optimized boiler rapid start and stop system end, the heat storage system end includes; Data acquisition module, used to collect real-time flue gas temperature, coal quality information and dispatching load instructions from the flue gas side temperature measurement point, boiler coal quality monitoring equipment and dispatching load instruction receiving end; An AI prediction module is in communication with the data acquisition module, receives the real-time flue gas temperature, coal quality information, and dispatch load instructions, predicts the flue gas temperature and load using a preset machine learning model, generates prediction results, and outputs them to the heat storage system control module and the temperature compensation control module; The thermal stress analysis module communicates with the AI prediction module, receives the prediction results and the actual operating parameters of the heat transfer element, and calculates the thermal stress distribution through the finite element thermal-mechanical coupling model; A heat storage system control module is in communication with the AI prediction module and the thermal stress analysis module. It controls the heat input and output of the heat storage system based on the prediction results and the load status of the unit. It stores flue gas heat and residual steam heat under high load conditions, and outputs heat to improve boiler efficiency or assist in unit startup under low load conditions. It also fine-tunes heat distribution based on feedback from the thermal stress analysis module. A temperature compensation control module is in communication with the AI prediction module, receives the prediction results and the temperature monitoring data from the circulation system end, and controls the temperature uniformity of the flue gas side and the exhaust temperature by adjusting the forced circulation speed of the medium; The thermal stress analysis module is in communication with the AI prediction module, the heat storage system control module, and the temperature compensation control module. It analyzes the thermal stress distribution based on the prediction results and the actual operating parameters of the heat transfer element, and feeds the analysis results back to the heat storage system control module and the temperature compensation control module to adjust the heat output and the medium circulation speed.

2. The AI-based boiler air preheater temperature uniformity control system according to claim 1 is characterized in that: The data acquisition module is configured as follows: Temperature data is collected from the temperature measuring points on the flue gas side, the coal element content and combustion characteristic parameters are obtained through coal quality monitoring equipment, and load adjustment instructions are obtained in real time from the dispatch instruction receiving end.

3. The AI-based boiler air preheater temperature uniformity control system according to claim 1 is characterized in that: The machine learning models preset in the AI prediction module include XGBoost, LightGBM and deep neural networks; The flue gas temperature change trend is predicted based on coal quality information and load instructions, and high-resolution load prediction results are generated. The prediction results are transmitted to the heat storage system control module for heat storage strategy adjustment and to the temperature compensation control module for medium circulation speed optimization.

4. The AI-based boiler air preheater temperature uniformity control system according to claim 1 is characterized in that: The thermal storage system control module is configured as follows: Under high load conditions, the flue gas heat and residual steam heat are input into the heat storage system for storage. Under low load conditions, the heat is output to the warm air heating system to increase the inlet air temperature, or to the optimized boiler rapid start-stop system to accelerate the start-up of the unit. Based on the stress distribution data of heat transfer elements fed back by the thermal stress analysis module, the heat distribution ratio among multiple units is dynamically adjusted.

5. The AI-based boiler air preheater temperature uniformity control system according to claim 1 is characterized in that: The temperature compensation control module is configured to: Based on the flue gas temperature distribution data fed back from the circulation system, the medium circulation speed in the high-temperature zone is controlled to accelerate heat absorption. The stress threshold data of the thermal stress analysis module is simultaneously received to limit the adjustment range of the medium circulation speed to avoid excessive thermal stress.

6. The AI-based boiler air preheater temperature uniformity control system according to claim 1 is characterized in that: The thermal stress analysis module is configured to: Based on the difference between the predicted temperature and the actual operating temperature by the AI prediction module, the instantaneous thermal stress distribution of the heat transfer element is calculated, and a stress adjustment instruction is generated and transmitted to the heat storage system control module to reduce heat output, or transmitted to the temperature compensation control module to reduce the medium circulation speed.

7. The AI-based boiler air preheater temperature uniformity control system according to claim 1 is characterized in that: The ammonium bisulfate capture system includes: The corrosion-resistant independent circulation pipeline maintains the temperature in the low-temperature section of the air preheater within the ammonium bisulfate condensation range. The heat is input through the heat storage system to gasify the condensed ammonium bisulfate. The soot blowing device is used to remove the lumps on the pipe wall, and the gaseous ammonium bisulfate is transported to the ash removal system through the ash collection system.

Images