Energy-saving control method and system for sintering fan system based on multi-objective optimization algorithm

Abstract

The application discloses a sintering fan system energy-saving control method and system based on a multi-objective optimization algorithm and belongs to the technical field of intelligent control of industrial processes. A multimodal time sequence characteristic model is constructed through an infrared thermal imager, a laser waste gas analyzer and process parameters to predict a sintering end position; millimeter wave radar and acoustic emission sensor signals are fused, and infrared temperature calibration is combined to inversely calculate a vertical combustion speed in real time; the end position deviation, the vertical combustion speed deviation and the total energy consumption of the fan are taken as optimization targets, the NSGA-II algorithm is adopted to solve the optimal frequency combination of the main exhaust fan and the desulfurization fan, and the fan rotating speed is dynamically adjusted through an industrial bus. The application realizes the collaborative optimization of sintering quality and energy efficiency, solves the problems of insufficient state sensing, regulation lag, and difficulty in considering energy saving and quality in traditional control, and can improve control precision and significantly reduce system energy consumption.

Description

Technical Field

[0001] This invention relates to the field of intelligent control technology for industrial processes, and in particular to an energy-saving control method and system for sintering blower systems based on multi-objective optimization algorithms. Background Technology

[0002] In the steel sintering process, the main exhaust fan and desulfurization fan are among the most energy-intensive pieces of equipment, and their operating status directly affects the quality of sinter and the system's energy efficiency. Currently, sintering fan control generally relies on operator experience or an open-loop adjustment strategy based on a fixed air volume-trolley speed curve. The sintering endpoint is often indirectly determined by the inflection point of the exhaust gas temperature in the wind box, while the vertical combustion speed cannot be obtained in real time due to the lack of effective online measurement methods. This control method results in severe lag in fan adjustment, making it difficult to balance the stability of the sintering endpoint with the rationality of the combustion process: excessive ventilation is often used to avoid under-burning, resulting in energy waste; while forcibly maintaining the endpoint position when the permeability of the material layer deteriorates can easily lead to combustion zone breakage and increased return ore rate. Therefore, existing technologies face core bottlenecks such as insufficient perception of sintering status, a single control objective, and difficulty in coordinating energy saving and quality.

[0003] To address the high energy consumption of sintering blowers, existing technologies have attempted optimization from an energy recovery perspective. For example, CN116105506B discloses a steam-electric dual-drive system and method for energy recovery in sintering systems. This system recovers high-temperature flue gas from the sintering ring cooler using a waste heat boiler to generate steam, which drives a steam turbine and an electric motor coaxially to operate the sintering blower. Based on real-time monitoring of steam flow and pressure, it dynamically switches between steam-driven and electric modes to achieve energy savings. While this solution effectively reduces the grid load, its control logic relies entirely on waste heat steam parameters and preset thresholds. It does not perceive the combustion state of the sintering process itself, nor does it establish a correlation mechanism with blower regulation, lacking process state awareness. Therefore, under conditions such as raw material fluctuations and changes in permeability, it cannot proactively adjust the blower operation strategy to balance sinter quality and energy efficiency, potentially leading to over-burning, under-burning, or abnormal combustion zones, failing to achieve true synergistic optimization of sintering process quality and energy efficiency. Summary of the Invention

[0004] In view of this, the present invention proposes an energy-saving control method and system for sintering blower system based on multi-objective optimization algorithm. By integrating multi-modal sensing, state inversion and multi-objective optimization algorithm to solve the optimal frequency combination control scheme, the synergistic optimization of sintering process quality and energy efficiency is achieved.

[0005] This invention provides an energy-saving control method for a sintering blower system based on a multi-objective optimization algorithm, comprising the following steps: S1. Acquire multi-source real-time data of the sintering process, including the surface temperature field distribution of the material layer collected by an infrared thermal imager, the time-series data of exhaust gas composition collected by a gas analyzer, and the operating parameters of the sintering machine. S2. Based on the multi-source real-time data, construct a multimodal time-series feature model to predict the sintering endpoint position; S3. The vertical combustion speed is retrieved in real time by fusing the measurement signals of millimeter-wave radar and acoustic emission sensors and combining them with infrared temperature calibration. S4. Obtain the endpoint position deviation based on the predicted sintering endpoint position, calculate the vertical combustion speed deviation based on the inverted vertical combustion speed and the preset set combustion speed, and take the endpoint position deviation, vertical combustion speed deviation and the total energy consumption of the fan as a multi-objective optimization objective, and use the NSGA-II algorithm to solve the optimal frequency combination of the main exhaust fan and the desulfurization fan. S5. The optimal frequency combination is sent to the corresponding frequency converter via the industrial communication bus to dynamically adjust the fan speed.

[0006] Furthermore, the temperature field distribution on the surface of the material layer is dynamically aligned with the time-series data of the exhaust gas composition in the time dimension, and spatial mapping is performed based on the speed of the sintering trolley to construct a spatiotemporally consistent multimodal input tensor.

[0007] Furthermore, the millimeter-wave radar and acoustic emission sensor respectively characterize the electromagnetic properties and mechanical fracture behavior of the combustion zone. The signals from the two are fused through a cross-attention mechanism to generate a confidence spectrum of the combustion front, which is then combined with infrared temperature calibration to invert the vertical combustion velocity.

[0008] Furthermore, the infrared temperature calibration serves as a physical anchor point to establish a spatial correspondence between the extreme points of the radar dielectric constant gradient and the abrupt change points of acoustic emission energy, thereby calibrating the absolute position of the combustion front.

[0009] Furthermore, the multimodal time-series feature model incorporates historical labeled data of the sintering endpoint during the training phase and employs an adversarial verification mechanism to filter historical samples that are consistent with the current operating condition distribution, thereby suppressing model domain shift caused by fluctuations in raw material composition.

[0010] Furthermore, in the NSGA-II optimization process, the vertical combustion rate deviation and the endpoint position deviation are coupled and weighted, and the weights are dynamically adjusted according to the current material layer permeability index, so that the combustion rate stability is prioritized under low permeability conditions.

[0011] Furthermore, before executing S5 to dynamically adjust the fan speed, the optimal frequency combination is subjected to fuzzy discretization processing to limit the fan frequency change amplitude within adjacent control cycles to not exceed a preset threshold, so as to suppress system oscillation caused by optimization solution jumps.

[0012] Furthermore, when the predicted sintering endpoint position deviation exceeds the threshold and the vertical combustion speed continues to deviate, the dual-fan coordinated adjustment mode is triggered: the main exhaust fan adjusts the total air volume, the desulfurization fan finely adjusts the negative pressure gradient, and controls the air pressure distribution and combustion front advancement rate within the material layer.

[0013] Furthermore, the set combustion speed is obtained by retrieving cases from the historical successful operating condition database that are most similar to the current mixture composition, moisture, and particle size distribution, and extracting their corresponding vertical combustion speed as a dynamic set value.

[0014] This invention also proposes an energy-saving control system for a sintering fan system based on a multi-objective optimization algorithm, comprising: a data acquisition module for real-time acquisition of the surface temperature field distribution of the material layer collected by an infrared thermal imager, the time-series data of exhaust gas composition output by an exhaust gas composition analyzer, and operating parameters from the sintering machine DCS system; a prediction module for constructing a multi-modal time-series feature model based on the data from the data acquisition module to predict the current sintering endpoint position; a combustion rate inversion module for fusing measurement signals from a fusion millimeter-wave radar and an acoustic emission sensor, and combining them with infrared temperature calibration information to invert the vertical combustion rate in real time; a multi-objective optimization decision module for constructing a multi-objective optimization problem based on the endpoint position deviation output by the sintering endpoint prediction module, the vertical combustion rate deviation output by the vertical combustion rate inversion module, and the real-time energy consumption of the main exhaust fan and the desulfurization fan, and using the NSGA-II algorithm to solve for the optimal frequency combination of the fans; and an execution control module for sending the optimal frequency combination to the frequency converters corresponding to the main exhaust fan and the desulfurization fan via an industrial communication bus to dynamically adjust the fan speed.

[0015] The present invention has the following advantages over the prior art: By acquiring high-resolution material surface temperature fields using infrared thermal imagers and integrating parameters such as exhaust gas composition and trolley speed to construct a multimodal time-series characteristic model, the accuracy of sintering endpoint prediction is significantly improved. Simultaneously, for the first time, millimeter-wave radar and acoustic emission sensors are fused to measure the electromagnetic properties and mechanical fracture behavior of the combustion zone. Combined with infrared temperature calibration to invert the vertical combustion velocity, this solves the long-standing problem of the inability to obtain vertical combustion velocity online. Through the fusion and inversion of infrared thermal imagers, millimeter-wave radar, and acoustic emission sensors, the system's perception of the sintering state can be directly quantified, improving control accuracy by more than 15% compared to traditional methods, and significantly enhancing the accuracy and stability of sintering process control.

[0016] Furthermore, traditional control methods only focus on the endpoint position, often leading to high energy consumption or combustion zone disruption due to excessive ventilation. This invention incorporates endpoint deviation, vertical combustion velocity deviation, and total fan energy consumption into the NSGA-II multi-objective optimization framework, and introduces a permeability index to dynamically adjust the objective weights. Under low permeability conditions, it prioritizes stable combustion velocity to prevent sinter strength degradation. This mechanism ensures that fan regulation always serves a healthy combustion state, achieving synergistic optimization of quality and energy efficiency.

[0017] Meanwhile, by solving for the Pareto optimal frequency combination of the main exhaust fan and the desulfurization fan, and using fuzzy discretization to suppress command jumps, the system achieves smooth and efficient adjustment of the fan speed while meeting process requirements. Practical applications show that this invention can reduce the overall power consumption of the fan system, while reducing mechanical stress and electrical shocks caused by frequent and large-amplitude frequency adjustments, and extending the life of the frequency converter and motor.

[0018] Finally, the combustion rate is dynamically generated based on the physical properties of the mixture from a historical successful operating condition database. The multimodal model uses an adversarial verification mechanism to suppress the impact of raw material fluctuations, and the control commands are issued in milliseconds via the Profibus-DP bus, ensuring the robustness and real-time performance of the solution in complex industrial sites, and possessing strong adaptability and engineering implementation capabilities. Attached Figure Description

[0019] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0020] Figure 1 This is a system overall architecture diagram according to an embodiment of the present invention; Figure 2 This is a flowchart of the algorithm in an embodiment of the present invention. Detailed Implementation

[0021] The technical solutions of the present invention will be clearly and completely described below with reference to the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0022] like Figure 1 As shown, the overall composition and data flow architecture of the intelligent control system for sintering blowers involved in this invention are illustrated. The system consists of four core layers: Sensor layer: includes infrared thermal imager, laser gas analyzer and acoustic emission sensor, which are used to collect material surface temperature field, exhaust gas composition and combustion zone mechanical rupture signal, respectively; Data acquisition and control system: gathers raw data from various sensors into the data preprocessing module to perform operations such as time alignment, spatial mapping and noise filtering; Edge computing server: A high-performance computing unit deployed on-site, containing two key models: a multi-dimensional coupled analysis engine, used to fuse information from multiple sources such as temperature, gas, and acoustics to determine the current sintering state; and a fan frequency optimization model, which uses the NSGA-II algorithm to solve for the optimal combination of operating frequencies of the main exhaust fan and the desulfurization fan. Control execution unit: Through frequency converter control commands, the optimization results are sent to the main exhaust fan and desulfurization fan to achieve closed-loop dynamic adjustment.

[0023] In one implementation, such as Figure 2 As shown, this invention provides an energy-saving control method for a sintering blower system based on a multi-objective optimization algorithm, comprising the following steps: S1. Acquire multi-source real-time data of the sintering process, including the surface temperature field distribution of the material layer collected by an infrared thermal imager, the time series data of the exhaust gas composition collected by a gas analyzer, and the operating parameters of the sintering machine; the surface temperature field distribution of the material layer and the time series data of the exhaust gas composition are dynamically aligned in the time dimension and spatially mapped based on the speed of the sintering trolley to construct a spatiotemporally consistent multimodal input tensor.

[0024] S2. Based on the multi-source real-time data, a multimodal time-series feature model is constructed to predict the sintering endpoint position. The multimodal time-series feature model introduces historical labeled data of the sintering endpoint during the training phase and uses an adversarial verification mechanism to filter historical samples that are consistent with the current working condition distribution, thereby suppressing model domain shift caused by fluctuations in raw material composition.

[0025] In a preferred embodiment of the present invention, the construction of the multimodal temporal feature model includes the following steps: (1) Multi-source heterogeneous data acquisition and standardization An infrared thermal imager is installed above the sintering machine's air box, continuously acquiring the surface temperature field of the material layer at a sampling frequency of no less than 5Hz. Each frame has a resolution of 1024×768 pixels, corresponding to an area of ​​approximately 3m in width and 4m in length of the material surface, with a measurement point density exceeding 1000. The raw images are filtered by median filtering to remove dust interference, and motion compensation is performed based on the trolley encoder signal to eliminate image blurring caused by trolley movement.

[0026] The laser gas analyzer is deployed in the main flue and outputs at a frequency of 1Hz. , , Concentration time-series data. Since its sampling rate is lower than that of infrared data, the original concentration sequence is smoothed using the sliding window averaging method, and then upsampled to 5Hz by linear interpolation to synchronize it with the infrared frame rate.

[0027] Sintering machine operating parameters, including trolley speed The material layer thickness H and the material level in the mixing silo are provided by the DCS system at a frequency of 1Hz, which is also interpolated to 5Hz and used as an auxiliary channel input model.

[0028] (2) Dynamic alignment and spatial mapping in the time dimension Time alignment refers to using the timestamps of each frame of an infrared thermal imager. Based on this, the interpolated exhaust gas composition data will be used as a benchmark. and running parameters Corresponding to the same moment, they form synchronized data tuples. ,in Here is the temperature field matrix.

[0029] Spatial mapping is necessary because the exhaust gas composition reflects the gases produced in the downstream combustion zone, while infrared observation reflects the current state of the material surface; therefore, a spatial lag relationship needs to be established. Let L be the distance L from the combustion zone to the downstream of the infrared field of view, then the exhaust gas composition... Actual correspondence The material layer located in the infrared field of view at any given time, where the delay time The unit is seconds. Therefore, Map back Temperature field at time Construct a causally consistent spatiotemporal input tensor.

[0030] Finally, the data from the most recent 60 seconds (300 frames) is extracted using a sliding window to form a three-dimensional input tensor: Where C represents the number of channels for waste gas and process parameters, including , , , H, etc.

[0031] (3) Multimodal feature extraction and fusion Construct a hybrid neural network architecture to process data from different modalities: Spatial feature extraction: extracting temperature field sequences The input is a spatiotemporal convolutional network, ST-ConvNet, consisting of 3 layers of 3D convolutions with a kernel size of 3×3×3 and a stride of 1, used to capture the evolution patterns of temperature gradients in time and space, and outputting a high-dimensional spatiotemporal feature map. .

[0032] Temporal feature extraction: This involves combining waste gas components with process parameter sequences. Input a bidirectional LSTM network with 128 hidden units, extract its long-term dependencies and temporal trend features, and output a temporal context vector. .

[0033] To achieve feature-weighted fusion, a cross-attention mechanism is introduced: Flatten into a sequence , As ; Calculate attention weights , obtain fusion features It retains the spatial structure information of the temperature field and incorporates the chemical reaction semantics of the exhaust gas components.

[0034] (4) Prediction of the sintering endpoint Fusion features Input a fully connected regression head, and output the predicted value of the sintering endpoint position within the next 30 seconds. Unit: m, relative to the starting point of the sintering machine. The model training uses historical operating data, labeled with manual annotations or the endpoint position derived from the inflection point of the bellows temperature. An adversarial verification mechanism is introduced to screen samples with consistent distribution and suppress performance degradation caused by raw material fluctuations.

[0035] S3. The vertical combustion velocity is inverted in real time by fusing the measurement signals of millimeter-wave radar and acoustic emission sensor and combining them with infrared temperature calibration. The millimeter-wave radar and acoustic emission sensor respectively characterize the electromagnetic properties and mechanical fracture behavior of the combustion zone. The two signals are fused by a cross-attention mechanism to generate a confidence spectrum of the combustion front, and then combined with infrared temperature calibration to invert the vertical combustion velocity.

[0036] Fusion features Normalization was performed to generate a combustion front confidence map. At each time t, apply the Softmax function to the depth dimension z. The location with the highest confidence level in the graph represents the current combustion front estimate. .

[0037] Millimeter-wave radar and fiber optic acoustic sensors, along with an infrared thermal sensing module, are arranged on both sides of the sintering machine to realize a three-dimensional model of the combustion state and combustion front. Based on the established model, the vertical combustion velocity is calculated. The specific parameters and methods are shown in the table below: The infrared temperature calibration serves as a physical anchor point, establishing a spatial correspondence between the extreme points of the radar dielectric constant gradient and the abrupt change points of acoustic emission energy, and calibrating the absolute position of the combustion front.

[0038] The confidence level graphs described above only provide relative positions and require absolute calibration. Therefore, a coaxially mounted infrared temperature calibration module is introduced, with a temperature measurement range of 600-1500℃ and an accuracy of ±1℃.

[0039] The highest temperature point of the material surface measured by infrared Typically lags behind the combustion front by approximately ; Establish a temperature-depth mapping relationship using historical calibration data: when At that time, the corresponding combustion front depth ; this physical anchor point The absolute scale used for calibrating the confidence profile: if Then, a rigid translation is performed on the entire graph, so that... Simultaneously, the bias terms of the fusion network are updated to achieve online self-correction.

[0040] To obtain the absolute position of the combustion front in a continuous time series Then, calculate its moving speed: Unit: mm / min. A sliding window linear regression with a window length of 10 s was used to suppress transient jitter and output a smooth vertical burning speed. .

[0041] S4. Based on the predicted sintering endpoint position, the endpoint position deviation is obtained. The vertical combustion speed deviation is calculated based on the inverted vertical combustion speed and the preset combustion speed. The endpoint position deviation, vertical combustion speed deviation, and total fan energy consumption are used as multi-objective optimization targets. The NSGA-II algorithm is employed to solve for the optimal frequency combination of the main exhaust fan and the desulfurization fan. During the NSGA-II optimization process, the vertical combustion speed deviation and the endpoint position deviation are coupled and weighted. The weights are dynamically adjusted according to the current material layer permeability index, ensuring stable combustion speed under low permeability conditions. When the predicted sintering endpoint position deviation exceeds a threshold and the vertical combustion speed continues to deviate, a dual-fan coordinated adjustment mode is triggered: the main exhaust fan adjusts the total air volume, and the desulfurization fan fine-tunes the negative pressure gradient, regulating the air pressure distribution and combustion front advance rate within the material layer.

[0042] In a preferred embodiment of the present invention, the steps for obtaining the optimal frequency combination of the main exhaust fan and the desulfurization fan are as follows: (1) Definition and normalization of multi-objectives Let the current control period be t, and define the following three optimization objectives: End point position deviation: ,in The predicted sintering endpoint location, in meters (m). To set the target position, take 88% of the sintering machine length.

[0043] Vertical combustion speed deviation: ,in The real-time vertical combustion velocity is the inverted value, in mm / min. The set combustion speed is dynamically generated from a database of historical successful operating conditions.

[0044] Total energy consumption of the fan: Based on an engineering model where wind turbine power is proportional to the cube of frequency, the inverter output frequency is used. , For input estimation.

[0045] Due to the large differences in the dimensions and orders of magnitude of the three, Min-Max normalization is used to normalize them to the [0,1] interval: ,in , These are extreme values ​​in historical operating data.

[0046] (2) Construction of dynamic coupling weighted and composite quality objectives To achieve adaptive optimization under operating conditions, and Integrate into a single quality objective : Calculation of the air permeability index of the material layer: Q is the main exhaust volume, in m³. 3 / min, Δp is the negative pressure in kPa, k is the calibration coefficient, and after normalization, P∈[0,1].

[0047] Dynamic weight function: Typical parameters: , , , .

[0048] Composite quality objectives: .

[0049] (3) NSGA-II optimization process and Pareto solution set selection Optimization variable: Main exhaust fan frequency Desulfurization fan frequency .

[0050] Constraints: Ensure minimum air volume; the negative pressure of the desulfurization system is [-12, -8] kPa.

[0051] Fitness function: Bi-objective .

[0052] Pareto solution set selection rule: Select the solution closest to the ideal point (0,0) from the non-dominated front. .

[0053] (4) Working condition diagnosis and control strategy triggering of multi-dimensional coupling analysis engine Operating conditions are classified based on preset rules: Under normal operating conditions, when At that time, the NSGA-II standard is optimized; under low air permeability conditions, and Enable high weight Prioritize stability Risk of losing control at the endpoint and If the process continues for ≥2 cycles, the dual-fan coordinated adjustment mode is triggered. The main exhaust fan adjusts its frequency significantly to quickly correct the total air volume; the desulfurization fan fine-tunes its frequency (±2Hz) to change the negative pressure gradient of the system, reconstruct the airflow distribution in the material layer, and guide the combustion front to return to the normal propulsion trajectory.

[0054] S5. The optimal frequency combination is sent to the corresponding strain gauge via the industrial communication bus to dynamically adjust the fan speed. Before executing S5 to dynamically adjust the fan speed, the optimal frequency combination is subjected to fuzzy discretization processing to limit the fan frequency change amplitude within adjacent control cycles to not exceed a preset threshold, so as to suppress system oscillation caused by optimization solution jumps.

[0055] The system adopts a fixed control cycle. The system performs closed-loop optimization and command updates every few seconds. This cycle comprehensively considers: the dynamic response time of the sintering process, the single-run solution time of the NSGA-II algorithm, and avoiding equipment fatigue caused by over-frequency adjustments. The dynamic response time of the sintering process refers to the combustion zone moving speed of approximately 15-25 mm / min, corresponding to a displacement of 7.5-12.5 mm in 30 seconds; the measured average dynamic response time of the sintering process is <8 seconds, completed on the edge server. At the beginning of each cycle, the system synchronously collects the latest sensor data, executes the S1-S4 process, and generates a new set of optimal frequency commands. .

[0056] To prevent frequency jumps in the NSGA-II optimization results due to sudden changes in operating conditions, a fuzzy discretization module is introduced to apply smoothing constraints to the original optimization solution: First, the maximum allowable variation is defined: Both the main exhaust and desulfurization fans are applicable; then the actual adjustment amount can be calculated. , The same applies to desulfurization fans; furthermore, the fuzzy logic is enhanced, and when the permeability index P < 0.4 or the negative pressure fluctuation rate > 10%, the threshold is further narrowed. Up to 3Hz.

[0057] The Profibus-DP industrial communication bus configuration uses the Profibus-DP (Decentralized Periphery) protocol, conforming to the IEC 61158 standard, and features high real-time performance and strong anti-interference capabilities. The network topology uses an edge computing server as the DP master, and the main exhaust fan inverter and desulfurization fan inverter as DP slaves, connected via shielded twisted-pair RS-485 cable with a baud rate of 1.5Mbps. Its data frame structure consists of an output of 2 bytes per unit and an input of 4 bytes per unit, including feedback current, actual frequency, and fault status. Communication performance includes a single-cycle polling time of <5ms and an end-to-end delay of <10ms from command issuance to inverter execution. It supports reconnection after disconnection and CRC verification to ensure command reliability.

[0058] Inverter receives frequency commands Then, following the built-in S-shaped acceleration and deceleration curve, with an acceleration time ≥15s, the motor speed is smoothly adjusted to avoid mechanical shock. At the same time, the actual operating frequency, three-phase current, DC bus voltage and other status data are transmitted back to the edge server via Profibus-DP. In the next control cycle, the system incorporates the actual response deviation into the NSGA-II fitness function correction term for correction, forming a closed-loop design of optimization, execution and correction.

[0059] This invention also proposes an energy-saving control system for a sintering fan system based on a multi-objective optimization algorithm, comprising: a data acquisition module for real-time acquisition of the surface temperature field distribution of the material layer collected by an infrared thermal imager, the time-series data of exhaust gas composition output by an exhaust gas composition analyzer, and operating parameters from the sintering machine DCS system; a prediction module for constructing a multi-modal time-series feature model based on the data from the data acquisition module to predict the current sintering endpoint position; a combustion rate inversion module for fusing measurement signals from a fusion millimeter-wave radar and an acoustic emission sensor, and combining them with infrared temperature calibration information to invert the vertical combustion rate in real time; a multi-objective optimization decision module for constructing a multi-objective optimization problem based on the endpoint position deviation output by the sintering endpoint prediction module, the vertical combustion rate deviation output by the vertical combustion rate inversion module, and the real-time energy consumption of the main exhaust fan and the desulfurization fan, and using the NSGA-II algorithm to solve for the optimal frequency combination of the fans; and an execution control module for sending the optimal frequency combination to the frequency converters corresponding to the main exhaust fan and the desulfurization fan via an industrial communication bus to dynamically adjust the fan speed.

[0060] In summary, this invention provides an energy-saving control system for sintering blower systems that integrates multi-source sensing, intelligent modeling, multi-objective optimization, and closed-loop execution. By fusing infrared, millimeter-wave radar, acoustic emission, and process parameters in real time, it accurately predicts the sintering endpoint and inversely calculates the vertical combustion velocity, thereby driving the NSGA-II algorithm to dynamically generate a blower control strategy that balances quality and energy efficiency. This invention is the first to introduce multi-physics field fusion sensing and adaptive optimization based on operating conditions into sintering blower control, realizing a paradigm shift from experience-based regulation to state-driven control.

[0061] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. An energy-saving control method for a sintering blower system based on a multi-objective optimization algorithm, characterized in that, Includes the following steps: S1. Acquire multi-source real-time data of the sintering process, including the surface temperature field distribution of the material layer collected by an infrared thermal imager, the time-series data of exhaust gas composition collected by a gas analyzer, and the operating parameters of the sintering machine. S2. Based on the multi-source real-time data, construct a multimodal time-series feature model to predict the sintering endpoint position; S3. The vertical combustion speed is retrieved in real time by fusing the measurement signals of millimeter-wave radar and acoustic emission sensors and combining them with infrared temperature calibration. S4. Obtain the endpoint position deviation based on the predicted sintering endpoint position, calculate the vertical combustion speed deviation based on the inverted vertical combustion speed and the preset set combustion speed, and take the endpoint position deviation, vertical combustion speed deviation and the total energy consumption of the fan as a multi-objective optimization objective, and use the NSGA-II algorithm to solve the optimal frequency combination of the main exhaust fan and the desulfurization fan. S5. The optimal frequency combination is sent to the corresponding frequency converter via the industrial communication bus to dynamically adjust the fan speed.

2. The method according to claim 1, characterized in that, The surface temperature field distribution of the material layer is dynamically aligned with the time series data of the exhaust gas composition in the time dimension, and spatial mapping is performed based on the sintering trolley speed to construct a spatiotemporally consistent multimodal input tensor.

3. The method according to claim 1, characterized in that, The millimeter-wave radar and acoustic emission sensor characterize the electromagnetic properties and mechanical fracture behavior of the combustion zone, respectively. The signals from the two sensors are fused through a cross-attention mechanism to generate a confidence spectrum of the combustion front, which is then combined with infrared temperature calibration to retrieve the vertical combustion velocity.

4. The method according to claim 3, characterized in that, The infrared temperature calibration serves as a physical anchor point, establishing a spatial correspondence between the extreme points of the radar dielectric constant gradient and the abrupt change points of acoustic emission energy, and calibrating the absolute position of the combustion front.

5. The method according to claim 1, characterized in that, The multimodal time-series feature model incorporates historical labeled data of the sintering endpoint during the training phase and employs an adversarial verification mechanism to filter historical samples that are consistent with the current operating condition distribution, thereby suppressing model domain shift caused by fluctuations in raw material composition.

6. The method according to claim 1, characterized in that, In the NSGA-II optimization process, the deviation of vertical combustion rate and the deviation of endpoint position are coupled and weighted. The weights are dynamically adjusted according to the current permeability index of the material layer, so that the stability of combustion rate is prioritized under low permeability conditions.

7. The method according to claim 1, characterized in that, Before executing S5 to dynamically adjust the fan speed, the optimal frequency combination is subjected to fuzzy discretization processing to limit the fan frequency change amplitude within adjacent control cycles to not exceed a preset threshold, so as to suppress system oscillation caused by optimization solution jumps.

8. The method according to claim 1, characterized in that, When the predicted sintering endpoint position deviation exceeds the threshold and the vertical combustion speed continues to deviate, the dual-fan coordinated adjustment mode is triggered: the main exhaust fan adjusts the total air volume, the desulfurization fan finely adjusts the negative pressure gradient, and controls the air pressure distribution and combustion front advancement rate within the material layer.

9. The method according to claim 1, characterized in that, The set combustion speed is determined by retrieving cases from the historical successful operating condition database that are most similar to the current mixture composition, moisture, and particle size distribution, and extracting their corresponding vertical combustion speed as a dynamic set value.

10. An energy-saving control system for a sintering blower system based on a multi-objective optimization algorithm, characterized in that, include: The data acquisition module is used to acquire in real time the surface temperature field distribution of the material layer collected by the infrared thermal imager, the time series data of the exhaust gas composition output by the exhaust gas composition analyzer, and the operating parameters from the sintering machine DCS system. The prediction module is used to construct a multimodal time-series feature model based on real-time data from the data acquisition module to predict the current sintering endpoint position. The combustion rate inversion module is used to fuse the measurement signals from the fusion millimeter-wave radar and acoustic emission sensor, and combine them with infrared temperature calibration information to invert the vertical combustion rate in real time. The multi-objective optimization decision module constructs a multi-objective optimization problem based on the endpoint position deviation output by the sintering endpoint prediction module, the vertical combustion speed deviation output by the vertical combustion speed inversion module, and the real-time energy consumption of the main exhaust fan and the desulfurization fan, and uses the NSGA-II algorithm to solve for the optimal frequency combination of the fans. The execution control module is used to send the optimal frequency combination to the frequency converters corresponding to the main exhaust fan and the desulfurization fan via the industrial communication bus, so as to dynamically adjust the fan speed.

Images