Multivariable energy efficiency optimization control system for heating furnace

Abstract

The invention relates to the technical field of control, and particularly discloses a multivariable energy efficiency optimization control system for a heating furnace, which is used for solving the problems of local overheating, non-uniform temperature and difficulty in accurate positioning and compensation of heat loss in the operation of the existing cracking heating furnace. Comprising a parameter detection module, a multivariable coupling modeling and simulation module, an optimization control module and an execution and feedback module. According to the method, dynamic digital twinning is constructed through multi-modal online sensing and data assimilation, a Pareto frontier solution is generated based on model prediction control and improved NSGA-II parallel optimization, and the weight is adaptively adjusted; and when the hot spot / cold spot is triggered, a quadric surface fitting compensation strategy is implemented and issued for execution, so that high-precision simulation prediction, precise closed-loop control and real-time online energy efficiency optimization are realized.

Description

Technical Field

[0001] The present invention relates to the field of control technology, and in particular to a multivariable energy efficiency optimization control system for a heating furnace. Background Art

[0002] The Chinese invention patent with publication number CN118500140A discloses a heating furnace energy-saving monitoring and prediction method and system based on digital twin technology. Although the overall thermal efficiency and energy consumption of the cracking heating furnace using petroleum coke or natural gas as fuel under different working conditions are simulated, analyzed and predicted by constructing a multi-zone digital twin simulation model and combining historical and real-time operation data, the monitoring and evaluation of single indicators such as exhaust temperature, furnace surface temperature, flue gas oxygen content, excess air coefficient and overall thermal efficiency are achieved, but due to the long rotating drum structure of the cracking heating furnace, the material is circulated and retained in the furnace body for multiple times, the difference in heat capacity and thermal inertia of the refractory lining of the zoned heating section causes the local temperature response to lag, the interaction between the furnace body rotation and the gas-solid two-phase flow in the furnace affects the time-varying heat transfer coefficient, and the fuel injection and combustion-supporting air volume are required. Taking into account the combustion characteristics of volatile matter and carbonaceous fuels, the distribution of burners in multiple combustion zones resulting in large fluctuations in local flame temperature, the wear of the furnace head and furnace tail seals causing flue gas leakage and increased uncertainty in heat loss, and the heat transfer efficiency of the waste heat recovery heat exchanger being affected by the nonlinear fluctuations of smoke dust content and fouling conditions, the multiple characteristics are superimposed. As a result, key variables such as fuel flow rate, combustion air flow rate, zone temperature gradient, flue gas flow rate and oxygen content show strongly coupled dynamic changes. Even if the overall thermal efficiency meets the standard, local overheating or cold spots, uneven heating of the side walls, load fluctuations in the heat exchange section and heat transfer losses are difficult to accurately locate and correct in time, resulting in the inability to further improve equipment life, operational stability and energy-saving benefits. Therefore, it is urgent to propose a multivariable energy efficiency optimization control system for the cracking heating furnace based on the above-mentioned multivariable coupling characteristics of the cracking heating furnace. Summary of the Invention

[0003] The technical problem to be solved by the present invention is to provide a multivariable energy efficiency optimization control system for a heating furnace. Through multi-modal online sensing and multi-physics field digital twins with second-level data assimilation, model predictive control combined with NSGA-II multi-objective optimization and hot / cold spot dynamic compensation closed loop, high-precision simulation prediction, precise control and stable energy saving are achieved.

[0004] To achieve the above object, the present invention provides the following technical solutions: A multivariable energy efficiency optimization control system for a heating furnace includes a parameter detection module, which includes a flue gas parameter detection module, a furnace body temperature detection module, a fuel flow detection module, a combustion air flow detection module, a seal wear monitoring module and a waste heat recovery monitoring module, which are respectively used to collect the exhaust gas flow rate and oxygen content, the surface temperature of each partition of the furnace body, the fuel injection amount, the combustion air amount, the seal wear of the furnace head and furnace tail and the heat transfer efficiency of the heat exchanger in real time. The parameter detection module is sequentially connected to a multivariable coupling modeling and simulation module, an optimization control module and an execution and feedback module. The multivariable coupling modeling and simulation module is constructed based on the acquired parameters, covering material circulation retention, furnace body thermal inertia, two-phase flow A dynamic digital twin model of multi-physics field coupling of interaction, combustion characteristics, sealing and heat transfer efficiency is established, and historical operation data and real-time operation data are combined to perform multi-operating condition simulation prediction. The optimization control module solves the objective function of the simulation results of the multi-variable coupling modeling and simulation module and the set overall thermal efficiency, local temperature gradient and heat loss indicators, and generates fuel and combustion air zoning distribution strategy and waste heat recovery reversing valve opening adjustment strategy, as well as dynamic and precise compensation control strategy for layout hot spots and cold spots. The execution and feedback module drives the burner group fuel distribution device, combustion air volume adjustment device and waste heat recovery reversing valve according to the optimization control decision of the optimization control module, and updates the digital twin model in real time.

[0005] As a further solution of the present invention, the flue gas parameter detection module adopts multi-point ultrasonic wind speed sensors arranged in the flue ducts of different zones of the cracking heating furnace. By measuring the time difference of sound wave propagation in the flue gas medium, the instantaneous flow velocity of each zone is calculated in real time, and the influence of dust disturbance is eliminated by combining a multi-point static pressure difference compensation algorithm based on online sampling of smoke dust concentration. The oxygen content detection adopts a Raman spectrum sensor set before and after the oxygen meter, supplemented by a dual-mode fusion calibration method based on an electrochemical oxygen sensor, through real-time calibration of the Raman signal drift and the temperature response compensation of the electrochemical sensor.

[0006] It should be noted that the multi-point ultrasonic wind speed sensors used in the present invention, which are arranged in the flues of different zones of the cracking heating furnace, calculate the instantaneous flow velocity of each zone in real time by measuring the time difference of sound wave propagation in the flue gas medium, and combine with the multi-point static pressure difference compensation algorithm based on online sampling of smoke dust concentration to eliminate the error caused by dust disturbance. Compared with the traditional single-point detection method, the response speed and measurement accuracy are improved, and the flue gas flow parameters are provided for the multivariable coupling model; at the same time, Raman spectroscopy sensors are respectively arranged before and after the oxygen meter, and supplemented by the dual-mode fusion calibration method of the electrochemical oxygen sensor. By real-time calibration of the Raman signal drift with temperature and the temperature response deviation of the electrochemical sensor, the oxygen content detection accuracy and anti-drift ability in high temperature and high dust environment are improved, and the response time is shortened, which provides an effective confidence boundary condition for simulation prediction, improves the robustness of system operation and fault warning capability, enhances the accuracy, stability and dynamic response performance of flue gas parameter detection, and provides a data basis for the multivariable coupling simulation and optimization control of the present invention.

[0007] As a further solution of the present invention, the furnace body temperature detection module adopts a composite measurement of a high-temperature short-wave infrared thermal imager arranged in each heating zone on the outer surface of the cracking heating furnace and a fiber Bragg grating array temperature sensor buried under the refractory lining. The infrared and FBG sensor data acquisition is synchronously triggered by a rotary position encoder, and the infrared radiation drift and FBG temperature drift are corrected online based on the established furnace body thermal inertia compensation model and multi-point data fusion algorithm, and the surface temperature of each zone is continuously monitored in real time at a frequency of ≥20Hz.

[0008] It should be noted that the furnace body temperature detection module adopted in the present invention arranges high-temperature short-wave infrared thermal imagers in each heating zone on the outer surface of the cracking heating furnace body, and buries fiber Bragg grating array temperature sensors under the refractory lining for mutual complementary measurement, and also combines the rotary position encoder for synchronous triggering and acquisition, so as to online correct the infrared radiation drift and FBG temperature drift based on the furnace body thermal inertia compensation model and multi-point data fusion algorithm, and realize real-time continuous monitoring of the surface temperature of each zone, effectively overcoming the temperature measurement blind spots and error drift caused by high-temperature dust obstruction, rapid rotation of the furnace body and local thermal inertia lag, improving the temporal and spatial consistency and accuracy of the temperature measurement data, providing real boundary conditions with good time resolution and spatial resolution for the multi-physics field coupling simulation model, ensuring the dynamic response capability and error convergence speed of the simulation prediction, and providing data support for the optimization control module to compensate for local hot spots and cold spots, thereby improving the refinement level and operation stability of the energy efficiency optimization control of the heating furnace.

[0009] As a further solution of the present invention, the fuel flow detection module adopts a dual-mode measurement architecture of a mass flow sensor and an ultrasonic time-difference volume flow sensor arranged in series in the fuel pipeline at the front end of each burner, and a microwave dielectric analyzer is set between the two sensors to collect fuel density and viscosity data in real time. The collected mass flow and volume flow signals are dynamically calibrated and fused based on an improved gain adaptive Kalman filter algorithm and an online rheological model to compensate for measurement errors caused by changes in volatile matter and temperature and pressure fluctuations.

[0010] It should be noted that the fuel flow detection module adopted in the present invention arranges a Coriolis mass flow sensor and an ultrasonic volume flow sensor in series at the front end of each burner, and introduces a microwave dielectric analyzer to obtain fuel density and viscosity data in real time. The mass and volume flow signals and physical property parameters are synchronously input into the improved gain adaptive Kalman filter, and the noise covariance used is dynamically calibrated based on the online rheological model, thereby realizing the fusion and compensation of dual-mode signals under the conditions of fuel volatility and temperature and pressure fluctuations. It can capture the instantaneous signal distortion caused by sudden changes in fuel physical properties in real time, and maintain measurement consistency when switching between multiple operating conditions. It provides fuel flow boundary conditions for multivariable coupled digital twin simulation, improves the real-time, stability and fault warning capabilities of optimization control, and provides data support for dynamic energy efficiency regulation of the combustion system.

[0011] As a further solution of the present invention, the fuel flow detection module synchronously inputs the collected Coriolis mass flow signal and ultrasonic volume flow signal, along with the fuel density and viscosity parameters acquired in real time by the microwave dielectric analyzer, into a dual-channel extended Kalman filter. The process noise covariance of the first channel is dynamically updated based on an online regression reconstruction model of changes in fuel volatile content and historical operating deviations. The measurement noise covariance of the second channel is adjusted in real time through temperature-pressure hysteresis characteristic identification and an adaptive recursive least squares algorithm. After filter fusion, a secondary correction based on residual orthogonal projection is performed to compensate for flow measurement errors caused by changes in fuel volatility and temperature and pressure fluctuations.

[0012] It should be noted that the fuel flow detection module adopted by the present invention synchronously inputs the collected Coriolis mass flow signal and ultrasonic volume flow signal and the fuel density and viscosity parameters obtained in real time by the microwave dielectric analyzer into a dual-channel extended Kalman filter, wherein the process noise covariance of the first channel is dynamically updated based on the online regression reconstruction model of the fuel volatile content change and the historical operation deviation, and the measurement noise covariance of the second channel is adjusted in real time through the temperature-pressure hysteresis characteristic identification and adaptive recursive least squares algorithm, and after filtering and fusion, it is quadratically corrected based on the residual orthogonal projection, thereby realizing online detection of mass and volume flow measurement under extreme working conditions such as fuel volatile component fluctuations, rapid temperature changes and pressure pulsations, providing fuel flow boundary conditions with excellent confidence for multivariable coupled digital twin simulation, improving the response speed, stability and fault warning capability of the optimization control module to the dynamic adjustment of the combustion system, and enhancing the energy efficiency optimization control benefit of the present invention under complex working conditions.

[0013] As a further solution of the present invention, the combustion air flow detection module arranges a vortex flowmeter and a multi-point static pressure sensor at the inlet of the combustion air duct and the branch pipes of each combustion zone, and sets temperature and humidity sensors upstream and downstream of the flowmeter to collect environmental parameters. Based on the collected static pressure difference, vortex pulse frequency and temperature and humidity data, a nonlinear adaptive observer algorithm combined with a CFD simulation pre-calibration curve is used to estimate the velocity profile distortion coefficient in real time and compensate for the vortex measurement value.

[0014] It should be noted that the combustion air flow detection module of the present invention arranges high-temperature resistant vortex flowmeters and multi-point static pressure sensors at the inlet of the combustion air duct and the branch pipes of each combustion zone, and adds temperature and humidity sensors upstream and downstream of the flowmeter to synchronously collect environmental parameters. By combining the static pressure difference, vortex pulse frequency and temperature and humidity data input with the nonlinear adaptive observer algorithm of the CFD simulation pre-calibration curve, the velocity profile distortion coefficient is identified and estimated in real time, and the vortex measurement value is dynamically corrected. Compared with the vortex measurement method in the prior art that relies on a single calibration coefficient and cannot respond to changes in working conditions, the flow measurement accuracy under high dust, high temperature and flow pulsation conditions is improved, and the error caused by profile unevenness is eliminated through online profile correction, providing consistent and confident air volume boundary conditions for subsequent multivariable coupling simulation models, thereby enhancing the reliability and response speed of optimized control decisions.

[0015] As a further solution of the present invention, the seal wear monitoring module includes: a fiber Bragg grating strain sensor array distributed circumferentially on the outer edges of the sealing rings at the head and tail of the cracking heating furnace, for real-time collection of tiny radial deformations of the sealing rings; a high-frequency ultrasonic transducer is embedded at the contact surface of the sealing ring to detect ultrasonic signals generated by wear and gas leakage; an industrial infrared thermal imager is arranged on the outside of the sealing part to capture the local temperature field mutation signal caused by leaking flue gas; the three-way sensor signals are processed online by a multi-modal deep learning algorithm that integrates multi-scale wavelet packet decomposition and bidirectional long short-term memory network to detect the seal wear amount and leakage flow rate in real time, and the detection results are fed back to the optimization control decision module in real time for dynamic compensation.

[0016] It should be noted that the method adopted by the present invention is based on three-way multimodal sensing of fiber Bragg grating strain array, high-frequency ultrasonic transducer and industrial infrared thermal imager, combined with multi-scale wavelet packet decomposition and bidirectional LSTM deep learning fusion algorithm to perform real-time online monitoring of small radial deformation of sealing ring, ultrasonic leakage signal and temperature field mutation. In the closed-loop adaptive execution link of multivariable coupling simulation model update and optimization control decision of the present invention, by providing seal wear and leakage flow rate boundary conditions, it can not only quickly warn of abnormal sealing status and make compensation decisions in time, but also enable the digital twin model to dynamically self-calibrate, suppress smoke leakage and unpredictable heat loss caused by seal wear, improve the operating stability, response speed and comprehensive energy efficiency optimization level of the heating furnace system, and achieve long-term operation reliability improvement and extended downtime cycle through continuous online monitoring.

[0017] As a further solution of the present invention, the waste heat recovery monitoring module adopts a multi-parameter coupled online detection method, which specifically includes: arranging K-type thermocouple arrays at the inlet and outlet flues of the heat exchanger for real-time measurement of the inlet and outlet flue gas temperatures; arranging multi-point thin-film heat flux meters in the heat exchanger tube bundle for real-time collection of the heat flux per unit area of ​​the tube wall; installing an acoustic imaging sensor array on the outside of the tube bundle for online identification of scaling location and thickness distribution; the waste heat recovery monitoring module dynamically calculates the heat transfer coefficient based on the bidirectional heat balance equation and the adaptive wall thermal resistance model, and combines the wall thermal resistance change rate threshold to trigger the pipe cleaning cycle prediction algorithm to locate and quantitatively evaluate the decrease in heat transfer efficiency caused by the initial scaling and uneven distribution.

[0018] It should be noted that the waste heat recovery monitoring module of the present invention arranges K-type thermocouple arrays, multi-point thin-film heat flux meters and acoustic imaging sensor arrays in the inlet and outlet flues and tube bundles of the heat exchanger respectively, combines the bidirectional heat balance equation with the adaptive wall thermal resistance model to dynamically calculate the heat transfer coefficient in real time, and triggers the pipe cleaning cycle prediction algorithm with the wall thermal resistance change rate, thereby realizing the rapid positioning, accurate quantitative evaluation and active cleaning warning of initial trace fouling and local non-uniform fouling inside the heat exchanger. Compared with the limitations of the existing technology that cannot achieve high-sensitivity online detection of early fouling and difficulty in accurately predicting the cleaning cycle, the present invention realizes the improvement of dynamic and accurate detection accuracy of heat exchange efficiency and accuracy of early warning of fouling, improves the operating stability and long-term heat transfer performance of the heat exchanger, effectively reduces the frequency of maintenance shutdowns and improves the overall energy-saving effect of the system.

[0019] As a further solution of the present invention, the multivariable coupling modeling and simulation module adopts a hierarchical dynamic adaptive modeling method, including: a multi-physics field simulation engine based on finite element heat conduction, CFD flow, discrete element material circulation and combustion chemical reaction coupling; the simulation engine integrates a real-time online data assimilation algorithm, injects monitoring data into the simulation model at a frequency of seconds, and self-calibrates thermal inertia, material residence time and two-phase flow boundary conditions through a model error update mechanism driven by feedback residuals, gradually converging the simulation error to within ±0.5%; the simulation calculation runs on a GPU cluster parallel architecture.

[0020] It should be noted that the multivariable coupling modeling and simulation module of the present invention adopts a hierarchical dynamic adaptive modeling method of finite element heat conduction, CFD flow, discrete element material circulation and combustion chemical reaction multi-physical field coupling, and injects monitoring data in seconds with a real-time online data assimilation algorithm. It realizes dynamic self-calibration of thermal inertia, material residence time and two-phase flow boundary conditions through feedback residual drive, so that the simulation error converges to the set range in real time, and greatly reduces the simulation time through GPU cluster parallel computing, overcoming the high-precision and strong real-time simulation requirements that cannot be achieved by a single modeling method in the existing technology, improving the simulation prediction accuracy and real-time response performance, and providing a model foundation and rapid optimization capability for dynamic and precise control.

[0021] As a further solution of the present invention, the optimization control module includes: a multi-objective optimization engine based on model predictive control and an improved non-dominated sorting genetic algorithm, which is used to jointly solve the three objectives of maximizing overall thermal efficiency, homogenizing local temperature gradients, and minimizing heat loss; the engine automatically adjusts the weights of each objective through online sensitivity analysis, and generates a Pareto frontier solution set in parallel within the feasible domain of each control variable; when a local hot spot or cold spot deviates from a threshold and is triggered, the optimization control module automatically calls a local compensation sub-algorithm based on quadratic surface fitting to update the control strategy, and sends the final optimal control instruction to the execution and feedback module in real time.

[0022] It should be noted that the optimization control module of the present invention adopts a multi-objective optimization engine based on model predictive control and improved non-dominated sorting genetic algorithm, automatically adjusts the target weights online and generates the Pareto frontier solution set in parallel, and calls the local compensation sub-algorithm of quadratic surface fitting in real time to accurately adjust the control strategy when hot spots or cold spots appear. Compared with the limitations of single-objective optimization and lack of adaptive local compensation in existing technologies, it realizes the simultaneous optimization of overall and local energy efficiency indicators, improves the accuracy of optimization decisions, real-time response speed and energy efficiency stability.

[0023] Technical effect of a multivariable energy efficiency optimization control system for a heating furnace: This technical solution constructs a dynamic digital twin model coupled with multiple physical fields, and collects multimodal monitoring data of flue gas flow rate and oxygen content, furnace surface temperature, fuel and air flow, seal wear and waste heat recovery efficiency in real time, and uses adaptive data fusion and dynamic calibration algorithms to perform real-time simulation prediction and closed-loop self-calibration optimization control of the heating furnace's multiple variables, so that the overall thermal efficiency and local temperature distribution under complex working conditions of the cracking heating furnace are simultaneously optimized, eliminating the strong coupling disturbances caused by fuel component fluctuations, temperature and pressure disturbances, seal leakage and heat exchange fouling, and improving the system's operation robustness, accuracy and rapid response performance, so as to achieve the effect of coordinated improvement of energy efficiency indicators, equipment life and operation stability, improve overall energy efficiency, reduce local temperature difference fluctuations, effectively extend maintenance cycles and reduce unplanned downtime. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] Figure 1 This is a block diagram of a multivariable energy efficiency optimization control system for a heating furnace; Figure 2 Flowchart for solving technical problems for the system; Figure 3 Comparison chart of flow velocity simulation cloud map before and after compensation for CFD simulation; Figure 4 This is the signal processing flow chart of the seal wear monitoring module. DETAILED DESCRIPTION

[0025] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments 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 efforts are within the scope of protection of the present invention.

[0026] Example 1

[0027] The present invention proposes a multivariable energy efficiency optimization control system for a heating furnace, including a parameter detection module, the parameter detection module including a flue gas parameter detection module, a furnace body temperature detection module, a fuel flow detection module, a combustion air flow detection module, a seal wear monitoring module and a waste heat recovery monitoring module, which are respectively used for real-time collection of exhaust gas flow rate and oxygen content, surface temperature of each partition of the furnace body, fuel injection amount, combustion air amount, seal wear of the head and tail of the furnace and heat transfer efficiency of the heat exchanger. The parameter detection module is sequentially connected to a multivariable coupling modeling and simulation module, an optimization control module and an execution and feedback module. The multivariable coupling modeling and simulation module is constructed based on the acquired parameters, covering material circulation retention, furnace body thermal inertia, A dynamic digital twin model of multi-physics field coupling of two-phase flow interaction, combustion characteristics, sealing and heat transfer efficiency is built, and historical and real-time operation data are combined to perform multi-condition simulation prediction. The optimization control module solves the objective function of the simulation results of the multi-variable coupling modeling and simulation module and the set overall thermal efficiency, local temperature gradient and heat loss indicators, and generates fuel and combustion air zoning distribution strategies and waste heat recovery reversing valve opening adjustment strategies, as well as dynamic and precise compensation control strategies for layout hot spots and cold spots. The execution and feedback modules drive the burner group fuel distribution device, combustion air volume adjustment device and waste heat recovery reversing valve according to the optimization control decision of the optimization control module, and update the digital twin model in real time. Figure 1 In the process, the raw materials are transported by the feeder, and after the burner is ignited, high-temperature flue gas is generated and enters the cracking heating furnace body and recovers heat energy through the waste heat recovery heat exchanger. The system monitors the flue gas flow rate and oxygen content through composite sensing, infrared and FBG monitor the partition temperature, and the fuel dual sensor calibration obtains the flow rate. The combustion air flow rate is corrected through vortex street, static pressure and CFD curve algorithm, and the multi-mode fusion algorithm is used to detect seal wear and leakage. Thermocouple, heat flow meter and acoustic imaging data are collected in the heat exchanger and tube bundle, and the heat exchange efficiency and pipe cleaning timing are evaluated based on the thermal balance and adaptive wall resistance model. The parameter detection module performs signal filtering and correction. The multi-physics field coupling digital twin module is then input. This module generates thermal efficiency, partition temperature and heat loss predictions through multi-physics field simulation and online data assimilation. The optimization control module combines the improved NSGA-II in the MPC framework to generate Pareto frontier solutions in parallel and adaptively adjust the weights. When the predicted temperature exceeds the limit, the quadratic surface fitting compensation strategy is called to generate fuel, combustion air volume and reversing valve control solutions, and the execution and feedback module is sent to the DCS / PLC closed-loop implementation, thereby achieving high precision, high efficiency and stable energy saving of the cracking heater under dynamic conditions, and improving the overall energy efficiency level.

[0028] like Figure 2 As shown, in order to solve the technical problem raised by the background technology of the present invention, the process of solving the technical problem by the system proposed by the present invention is as follows: Step 1, real-time parameter detection: The flue gas parameter detection module, furnace body temperature detection module, fuel flow detection module, combustion air flow detection module, furnace head and furnace tail seal wear monitoring module, and waste heat recovery efficiency monitoring module collect online exhaust gas flow rate and oxygen content, surface temperature of each furnace partition, gas injection amount, combustion air flow, seal wear degree, and heat transfer efficiency of the waste heat recovery heat exchanger; Step 2: Multivariable Coupling Modeling and Simulation: Based on the real-time and historical data collected in Step 1, the multivariable coupling modeling and simulation module constructs a multi-physics digital twin simulation module, including multiple material circulation retention, thermal inertia differences among furnace sections, gas-solid two-phase flow coupling within the rotating furnace, combustion dynamics, seal leakage nonlinearity, and heat transfer efficiency fluctuations. Dynamic simulation is performed under different fuel ratios, combustion air volumes, and load conditions to predict overall thermal efficiency, local temperature differences, and heat loss distribution. Step 3: Generate optimal control decisions: The optimization control module constructs a multi-objective optimization problem based on simulation results and preset objectives (maximizing overall thermal efficiency, homogenizing local temperature gradients, and minimizing heat loss). It solves the optimal strategy for the fuel-to-combustion air distribution ratio in each combustion zone, the opening of the preheat recovery reversing valve, and hot / cold spot compensation control. It employs a model predictive control (MPC) framework combined with an improved non-dominated sorting genetic algorithm (NSGA-II) to generate Pareto optimal solution sets in parallel. Online sensitivity analysis is used to dynamically adjust the weights of each objective, and a quadratic surface fitting local sub-algorithm is used for rapid compensation under unexpected operating conditions. Step 4, strategy execution and closed-loop feedback: The execution and feedback module drives the fuel distribution device of the burner group, issues control instructions through the proportional / servo valve, variable frequency fan and PLC-RTU or DCS system to achieve fine adjustment of the fuel injection amount in each zone, controls the combustion-supporting fan or bypass valve to adjust the combustion-supporting air flow in each zone, operates the preheating recovery reversing valve to achieve dynamic switching between waste heat recovery and direct exhaust gas discharge, implements local temperature compensation control in hot / cold spot areas, and feeds back the execution results and real-time response data to the digital twin model to complete adaptive iterative updates.

[0029] In order to clearly and specifically illustrate the implementation of the technical solution of the present invention, the following test cases are used for illustration.

[0030] First, the cracking furnace was preheated and operated continuously for 4 hours until the temperature fluctuation of each zone was within ±2°C before the test began. The ambient temperature was 25°C and the relative humidity was 45%. Three-phase AC power was used with a voltage of 380V and a frequency of 50Hz. The historical operating data of the last two weeks (including load, fuel / air volume, temperature, sealing status, and heat exchange efficiency) was loaded into the system proposed in the present invention, and a complete processing flow was performed. In the test, test group 1 used natural gas as fuel with a standard calorific value of 38MJ / m³, test group 2 used petroleum coke as fuel with a calorific value of 29MJ / m³, a sulfur content of 0.8% by mass, and a volatile matter content of 12% by mass. Test group 3 used a mixed combustion mode of natural gas and petroleum coke with a mass ratio of 60:40. The tests were carried out separately. The statistical table of the test indicators is shown in Table 1: Table 1 Test index comparison table

[0031] The above data show that under 50%, 75% and 100% load and fuel conditions, the multi-physics field coupled digital twin model combined with MPC+NSGA-II multi-objective optimization and hot / cold spot dynamic compensation control can improve the overall thermal efficiency by 68 percentage points compared with the pre-optimization level, the maximum local temperature gradient decreases by 15-30°C, the heat loss per unit area is reduced by 4-10kW / m², the model simulation error converges to ±2.5%, the response time is shortened to 3 seconds, and the anti-disturbance capability is improved by 40%. This shows that the technical solution of the present invention has synergistically improved the three major innovations of accurate prediction, optimal decision-making and rapid compensation, thereby improving the overall thermal efficiency, balancing the temperature distribution and reducing heat loss, reducing the energy consumption of the cracking heater, enhancing equipment stability and extending equipment life, and enhancing the equipment's anti-disturbance capability.

[0032] Example 2

[0033] Different from the first embodiment, this embodiment specifically introduces the flue gas parameter detection module and the furnace temperature detection module.

[0034] The flue gas parameter detection module uses multi-point ultrasonic wind speed sensors arranged in the flue ducts of different zones of the cracking heating furnace. By measuring the time difference of sound wave propagation in the flue gas medium, the instantaneous flow velocity of each zone is calculated in real time, and the influence of dust disturbance is eliminated by combining a multi-point static pressure difference compensation algorithm based on online sampling of smoke dust concentration. The oxygen content detection adopts a Raman spectroscopy sensor set before and after the oxygen meter, supplemented by a dual-mode fusion calibration method based on an electrochemical oxygen sensor, through real-time calibration of the Raman signal drift and the temperature response compensation of the electrochemical sensor.

[0035] For example, at 9:00 a.m. on a certain day, ultrasonic wind speed sensors were arranged in the preheating zone (flue section A), middle section (flue section B), and tail section (flue section C) of a ceramic cracking heating furnace. Each group consisted of a transmitting probe and a receiving probe. By measuring the time difference Δt1 and Δt2 of the sound wave in the A→B and B→C paths, the instantaneous flow velocity v1=ΔL1 / Δt1 and v2=ΔL2 / Δt2 were calculated. Combined with the online smoke dust sampling, the dust concentration C powder = 15 mg / m 3 The static pressure differences ΔP1, ΔP2, and ΔP3 between the C powder and each point are input into a multi-point static pressure difference compensation algorithm based on Kalman filtering to correct the original wind speed to eliminate dust interference. At the same time, a Raman spectroscopy sensor is installed before and after the oxygen meter to collect the original spectra R1 and R2, respectively. The two Raman signals and the output voltage E of the electrochemical oxygen sensor are calibrated using a dual-mode fusion model constructed by multivariate linear regression and temperature response compensation algorithm. The measurement errors caused by Raman signal drift and electrochemical sensor temperature drift are corrected in real time, and the final output is the instantaneous oxygen content O2 = 20.8%. All collected and corrected data are transmitted to the host computer via industrial Ethernet at a frequency of 1s / time for edge computing and stored in a time series database for real-time visualization by the system, realizing a full-link online detection process of flue gas parameters combining ultrasonic-static pressure difference coupled noise elimination and Raman-electrochemical dual-mode fusion calibration.

[0036] The furnace body temperature detection module adopts a composite measurement method of high-temperature short-wave infrared thermal imagers arranged in each heating zone on the outer surface of the cracking heating furnace and fiber Bragg grating array temperature sensors buried under the refractory lining. The infrared and FBG sensor data acquisition is synchronously triggered by the rotary position encoder, and the infrared radiation drift and FBG temperature drift are corrected online based on the established furnace body thermal inertia compensation model and multi-point data fusion algorithm. The surface temperature of each zone is continuously monitored in real time at a frequency of ≥20Hz.

[0037] For example, at 10:00 am on a certain day, the following measurement process was implemented in the three heating zones (zone A, zone B, zone C) of a ceramic cracking heating furnace: (1) Sensor layout and synchronous triggering: A high-temperature short-wave infrared thermal imager (model IR-SW1024, resolution 1024×768, temperature measurement range 300-1200°C) was installed in each of the outer surface zones A, B, and C of the furnace. A 16-point fiber Bragg grating (FBG) temperature sensor array (central wavelength 1550nm, sensitivity 10pm / °C, spacing 50mm) was buried under the refractory lining of each zone. A rotary position encoder (10,000 pulses / rev) is installed on the furnace support shaft. It generates a trigger signal every 0.036° of rotation, which simultaneously triggers the thermal imager to capture a frame of infrared image and the FBG array to read all 16 wavelength data at once. The sampling frequency reaches 50Hz. (2) Online correction and compensation: Infrared radiation drift correction: Every 10 minutes, a 500°C standard blackbody reference target fixed to the furnace is used to collect infrared calibration points to correct the thermal imager's radiation drift caused by ambient temperature changes. FBG temperature drift correction: A set of room temperature reference FBGs (constant temperature 25±0.1°C) is set up outside the furnace room. Their wavelength drift is monitored in real time and the resulting drift is used as a baseline to eliminate the temperature drift error in each array reading. Thermal inertia compensation model: A first-order RC thermal inertia model is established for each heating zone, where the time constant in the model is set to 120s, 150s, and 180s based on the refractory lining and thermal conductivity of zone A, zone B, and zone C, respectively; (3) Multi-point data fusion algorithm: For the nth frame data, calculate the corrected infrared temperature; and the FBG average temperature, and adjust the fusion weight based on their respective historical residuals, and output the fusion temperature; (4) Real-time continuous monitoring: After the above correction and fusion, the system outputs the real-time surface temperature curves of zones A / B / C at a frequency of ≥20Hz and sends the results to the DCS via industrial Ethernet for real-time viewing by the optimization control module and operation and maintenance personnel.

[0038] The furnace temperature detection module coordinates the arrangement of short-wave infrared thermal imaging and FBG array sensors under the refractory lining to overcome the limitations of a single sensing method. A rotary position encoder is used to synchronously trigger the infrared image and FBG reading in real time, eliminating timing differences caused by rotation. A standard blackbody and room-temperature reference FBG in the furnace are used to achieve online compensation for infrared radiation drift and FBG temperature drift, eliminating the need for furnace shutdown and recalibration. Time-domain correction of measurement results is performed based on a segmented thermal inertia model to overcome the temperature hysteresis effect of high-temperature, large-volume furnaces. A weighted adaptive infrared-FBG data fusion algorithm is used to improve temperature measurement accuracy to ±1°C and maintain high-frequency monitoring of ≥20Hz.

[0039] Example 3

[0040] The difference between this embodiment and embodiments 1 and 2 is that this embodiment provides a detailed introduction and description of the fuel flow detection module.

[0041] The fuel flow detection module adopts a dual-mode measurement architecture consisting of a mass flow sensor and an ultrasonic time-of-flight volume flow sensor arranged in series in the fuel pipeline at the front end of each burner. A microwave dielectric analyzer is set between the two sensors to collect fuel density and viscosity data in real time. The collected mass flow and volume flow signals are dynamically calibrated and fused based on an improved gain-adaptive Kalman filter algorithm and an online rheological model to compensate for measurement errors caused by changes in volatile matter and temperature and pressure fluctuations.

[0042] The fuel flow detection module synchronously inputs the collected Coriolis mass flow signal and ultrasonic volume flow signal, along with the fuel density and viscosity parameters acquired in real time by the microwave dielectric analyzer, into a dual-channel extended Kalman filter. The process noise covariance of the first channel is dynamically updated based on an online regression reconstruction model of changes in fuel volatile content and historical operating deviations. The measurement noise covariance of the second channel is adjusted in real time through temperature-pressure hysteresis characteristic identification and an adaptive recursive least squares algorithm. After filter fusion, a secondary correction based on residual orthogonal projection is performed to accurately compensate for flow measurement errors caused by changes in fuel volatility and temperature and pressure fluctuations.

[0043] Specifically, the fuel flow detection module deploys a Coriolis mass flow sensor and an ultrasonic time-difference volume flow sensor in series in the fuel pipeline at the front end of each burner, and adds a microwave dielectric analyzer between the two to obtain fuel density and viscosity data in real time. The collected mass flow, volume flow and medium parameters are synchronously input into a dual-channel extended Kalman filter. The process noise covariance of the first channel is dynamically updated based on an online regression reconstruction model of changes in fuel volatile content and historical operating deviations. The measurement noise covariance of the second channel is adjusted in real time through temperature-pressure hysteresis characteristic identification and adaptive recursive least squares algorithm. After filtering and fusion, it is corrected twice by residual orthogonal projection. The improved gain adaptive Kalman filter algorithm and online rheological model are used to realize real-time dynamic calibration and error compensation of fuel composition, viscoelasticity and operating condition fluctuations.

[0044] Example 4

[0045] Different from Example 1, Example 2 and Example 3, this embodiment provides a specific description of the combustion air flow detection module and the seal wear monitoring module.

[0046] The combustion air flow detection module arranges vortex flowmeters and multi-point static pressure sensors at the inlet of the combustion air duct and the branch pipes of each combustion zone, and sets temperature and humidity sensors upstream and downstream of the flowmeter to collect environmental parameters. Based on the collected static pressure difference, vortex pulse frequency and temperature and humidity data, a nonlinear adaptive observer algorithm combined with the CFD simulation pre-calibration curve is used to estimate the velocity profile distortion coefficient in real time and compensate the vortex measurement value.

[0047] The combustion air flow detection module arranges vortex flowmeters and multi-point static pressure sensors at the inlet of the combustion air duct and the branch pipelines of each combustion zone, and sets temperature and humidity sensors upstream and downstream of the vortex flowmeter to collect environmental parameters. Based on the non-uniform flow velocity profile curve pre-calibrated by CFD simulation combined with static pressure difference, vortex pulse frequency and temperature and humidity data, a nonlinear adaptive observer algorithm is used to estimate the flow velocity profile distortion coefficient online and compensate the vortex measurement value, and the correction result is updated in real time at a frequency of 1Hz to eliminate the measurement errors caused by secondary flow, geometric distortion of pipeline branches and fluctuations in ambient temperature and humidity, and solve the problems of low accuracy, response lag and poor stability of the traditional single vortex method in complex flow fields, improve the combustion air flow measurement accuracy, response time and system anti-disturbance capability, provide input for combustion control optimization, improve combustion efficiency and reduce emissions.

[0048] To illustrate the measurement improvement effect of the combustion air flow detection module at the combustion air duct inlet, a fluid dynamics CFD simulation was performed on the typical geometric cross-sections of the cracking furnace combustion air duct and each branch pipe, including the main air duct and several branches. Assuming that the air is incompressible and Newtonian, the steady-state Reynolds-averaged Navier-Stokes (RANS) equations are used, combined with the standard k-ε turbulence model to describe the turbulent turbulence. In ANSYS Fluent, a structured / unstructured grid of approximately 500,000 to 1 million units is generated for the cross-sectional area, ensuring that the grid at the shear layer and branch opening is encrypted to ensure the convergence of the results. The nominal average flow velocity at the inlet is set to 15m / s, the fixed static pressure at the outlet is 1atm, and the pipe wall is under a no-slip condition to avoid setting the roughness according to the actual pipe material. The isothermal assumption is used, a steady-state solver is used, and the pressure-velocity coupling uses the SIMPLE algorithm. The residual converges to 1×10 -6 The velocity distribution cloud map is then output. Before compensation, the velocity cloud map is directly simulated using the original inlet profile. After compensation, the velocity profile distortion coefficient calibrated by the nonlinear adaptive observer algorithm is applied to the inlet profile, and the CFD calculation with the same settings is performed again, as shown in the following example: Figure 3 As shown in the figure, by comparing the velocity cloud maps obtained twice under the same grid, boundary conditions and solution parameters, it can be clearly seen that the velocity profile is restored from distortion (high in the center, low at the edge, asymmetric) to a more symmetrical and uniform distribution. This is an intuitive verification that the technical solution can correct vortex measurement errors in real time and improve measurement accuracy and flow field consistency.

[0049] like Figure 4As shown, the seal wear monitoring module includes: a fiber Bragg grating strain sensor array distributed circumferentially on the outer edges of the sealing rings at the head and tail of the cracking heating furnace, which is used to collect small radial deformations of the sealing rings in real time; a high-frequency ultrasonic transducer is pre-embedded at the contact surface of the sealing ring to detect ultrasonic signals generated by wear and gas leakage; an industrial infrared thermal imager is set on the outside of the sealing part to capture the local temperature field mutation signal caused by the leakage of flue gas; the three-way sensor signals are processed online by a multi-modal deep learning algorithm that integrates multi-scale wavelet packet decomposition and bidirectional long short-term memory network to detect the seal wear amount and leakage flow rate in real time, and the detection results are fed back to the optimization control decision module in real time for dynamic compensation.

[0050] Through three-way sensor signal fusion, multi-scale wavelet packet decomposition and bidirectional LSTM online processing, the seal wear accuracy, leakage flow rate estimation error and response speed are improved, the heat loss caused by seal leakage is dynamically compensated, and the operating stability and equipment life of the heating furnace are improved.

[0051] Example 5

[0052] Different from Example 1, Example 2, Example 3 and Example 4, this example is a specific description of the waste heat recovery monitoring module.

[0053] The waste heat recovery monitoring module adopts a multi-parameter coupled online detection method, which specifically includes: arranging K-type thermocouple arrays at the inlet and outlet flues of the heat exchanger for real-time measurement of the inlet and outlet flue gas temperatures; arranging multi-point thin-film heat flux meters in the heat exchanger tube bundle for real-time collection of the heat flux per unit area of ​​the tube wall; installing an acoustic imaging sensor array on the outside of the tube bundle for online identification of scaling location and thickness distribution; the waste heat recovery monitoring module dynamically calculates the heat transfer coefficient based on the bidirectional heat balance equation and the adaptive wall thermal resistance model, and combines the wall thermal resistance change rate threshold to trigger the pipe cleaning cycle prediction algorithm to locate and quantitatively evaluate the decline in heat transfer efficiency caused by the initial stage of scaling and uneven distribution.

[0054] Specifically, the method for the waste heat recovery monitoring module to perform multi-parameter online monitoring includes: Step 1: Sensor placement and calibration: K-type thermocouple arrays are evenly spaced inside the heat exchanger inlet and outlet flues, with at least eight points on each side, and a calibration accuracy of ±0.5°C. Thin-film heat flux meters are evenly installed on the outer walls of key pipes within the tube bundle. An acoustic imaging sensor array is placed around the outside of the tube bundle to identify the thickness of the tube structure in real time through the ultrasonic reflection intensity profile. Step 2: Online data acquisition: Thermocouple, heat flow meter, and acoustic array signals are sampled and then processed at the edge, with signal preprocessing used to remove noise and environmental interference. Step 3, bidirectional heat balance calculation: perform bidirectional heat balance calculation on each tube bundle unit according to the bidirectional heat balance equation; Step 4, Adaptive wall thermal resistance model: Build a first-order adaptive model of the tube bundle wall thermal resistance, where the learning rate is automatically adjusted based on historical errors to track the increase in wall resistance caused by scaling in real time. Step 5: Scale location and thickness assessment: The acoustic imaging profile is mapped to the tube bundle grid to identify the local final scaled area and thickness distribution. The structural area and thickness distribution are used as inputs to the wall resistance model, and spatially weighted correction is performed on the local tube bundle wall thermal resistance at the corresponding location. Step 6: Predict and trigger the pigging cycle: Monitor the real-time calculated heat transfer coefficient of the tube wall, tube bundle resistance, local scale thickness, and pigging countdown time for visualization and output to the optimization control module for dynamic compensation measures such as adjusting preheated air distribution or increasing flushing pipeline flow.

[0055] Example 6

[0056] Different from Example 1, Example 2, Example 3, Example 4 and Example 5, this embodiment is an explanation of the multivariable coupling modeling and simulation module and the optimization control module.

[0057] The multivariable coupling modeling and simulation module adopts a hierarchical dynamic adaptive modeling approach, including: a multi-physics field simulation engine based on the coupling of finite element heat conduction, CFD flow, discrete element material circulation and combustion chemical reaction; the simulation engine integrates a real-time online data assimilation algorithm, injects monitoring data into the simulation model at a frequency of seconds, and self-calibrates thermal inertia, material residence time and two-phase flow boundary conditions through a model error update mechanism driven by feedback residuals, gradually converging the simulation error to within ±0.5%; the simulation calculation runs on a GPU cluster parallel architecture.

[0058] The optimization control module includes: a multi-objective optimization engine based on model predictive control and an improved non-dominated sorting genetic algorithm, which is used to jointly solve the three objectives of maximizing overall thermal efficiency, homogenizing local temperature gradients, and minimizing heat loss; the engine automatically adjusts the weights of each objective through online sensitivity analysis and generates a Pareto frontier solution set in parallel within the feasible domain of each control variable; when a local hot spot or cold spot deviates from the threshold and is triggered, the optimization control module automatically calls a local compensation sub-algorithm based on quadratic surface fitting to update the control strategy, and sends the final optimal control instructions to the execution and feedback module in real time.

[0059] The multivariable coupled modeling and simulation module first runs the finite element heat conduction, CFD flow, DEM material circulation, and chemical reaction sub-models in parallel within a unified framework, exchanging intermediate quantities such as temperature, flow rate, particle distribution, and reaction rate through the MPI interface. Every second, the system automatically collects measured data such as flue gas temperature, heat flux, and material concentration from the parameter detection module. Using a correction process driven by feedback residuals—comparing the differences between simulation outputs and monitored values—it iteratively updates the furnace thermal inertia coefficient, material residence time, and gas-solid boundary conditions online until the simulation results are highly consistent with the on-site operating conditions. All calculations are deployed on a GPU cluster to ensure that multi-condition simulations are completed within a limited time window, achieving a dynamic balance between simulation depth and online speed, and maintaining high model reliability over a long period without manual intervention. The optimization control module organically integrates model predictive control (MPC) with an improved non-dominated sorting genetic algorithm (NSGA-II): During each control cycle, the module first calls the multi-condition results predicted by the digital twin to construct objective functions for thermal efficiency, local temperature difference, and heat loss, and automatically assigns weights through online sensitivity analysis. It then searches in parallel for the Pareto frontier solution set within the feasible domain of a given fuel flow rate, combustion air volume, and reversing valve opening. When the predicted temperature in any section exceeds the preset threshold, the system immediately activates a local compensation sub-algorithm based on quadratic surface fitting, refines the control path near the current optimal solution, and forms a new fuel distribution and air volume adjustment scheme, which is then sent to the execution module via the OPC UA / DCS interface. This process is completed in a closed loop within the same control time slot, ensuring real-time connection between simulation decisions and on-site execution, achieving precise energy efficiency optimization in a high-frequency, multi-variable coupled environment.

[0060] The above description is merely a specific embodiment of the present application, but the scope of protection of the present application is not limited thereto. Any changes or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in this application should be included in the scope of protection of this application. Therefore, the scope of protection of this application should be based on the scope of protection of the claims.

[0061] Finally: The above description is only a preferred embodiment of the present invention and is 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 in the scope of protection of the present invention.

Claims

1. A multivariable energy efficiency optimization control system for a heating furnace, characterized in that: Including parameter detection module, the parameter detection module includes flue gas parameter detection module, furnace temperature detection module, fuel flow detection module, combustion air flow detection module, seal wear monitoring module and waste heat recovery monitoring module, which are respectively used to collect waste gas flow rate and oxygen content, surface temperature of each partition of the furnace, fuel injection amount, combustion air amount, furnace head and furnace tail seal wear and heat exchanger heat transfer efficiency in real time. The parameter detection module is sequentially connected to the multivariable coupling modeling and simulation module, the optimization control module and the execution and feedback module. The multivariable coupling modeling and simulation module is constructed based on the acquired parameters, covering material circulation retention, furnace thermal inertia, The multi-physics field coupled dynamic digital twin model of two-phase flow interaction, combustion characteristics, sealing and heat transfer efficiency combines historical and real-time operation data to perform multi-condition simulation prediction. The optimization control module combines the simulation prediction results and the set overall thermal efficiency, local temperature gradient and heat loss indicators to solve the objective function, generate the fuel and combustion air zoning distribution strategy and the waste heat recovery reversing valve opening adjustment strategy, and layout the hot and cold spot dynamic compensation control strategy. The execution and feedback module drives the burner group fuel distribution device, combustion air volume adjustment device and waste heat recovery reversing valve according to the optimization control decision, and updates the digital twin model in real time.

2. A multivariable energy efficiency optimization control system for a heating furnace according to claim 1, characterized in that: The flue gas parameter detection module uses multi-point ultrasonic wind speed sensors arranged in the flue ducts of different zones of the cracking heating furnace. By measuring the time difference of sound wave propagation in the flue gas medium, the instantaneous flow velocity of each zone is calculated in real time, and the influence of dust disturbance is eliminated by combining a multi-point static pressure difference compensation algorithm based on online sampling of smoke dust concentration. The oxygen content detection adopts a Raman spectroscopy sensor set before and after the oxygen meter, supplemented by a dual-mode fusion calibration method based on an electrochemical oxygen sensor, through real-time calibration of the Raman signal drift and the temperature response compensation of the electrochemical sensor.

3. The multivariable energy efficiency optimization control system for a heating furnace according to claim 1, characterized in that: The furnace body temperature detection module adopts a composite measurement method of high-temperature short-wave infrared thermal imagers arranged in each heating zone on the outer surface of the cracking heating furnace and fiber Bragg grating array temperature sensors buried under the refractory lining. The infrared and FBG sensor data acquisition is synchronously triggered by the rotary position encoder, and the infrared radiation drift and FBG temperature drift are corrected online based on the established furnace body thermal inertia compensation model and multi-point data fusion algorithm. The surface temperature of each zone is continuously monitored in real time at a frequency of ≥20Hz.

4. The multivariable energy efficiency optimization control system for a heating furnace according to claim 1, characterized in that: The fuel flow detection module adopts a dual-mode measurement architecture consisting of a mass flow sensor and an ultrasonic time-of-flight volume flow sensor arranged in series in the fuel pipeline at the front end of each burner. A microwave dielectric analyzer is set between the two sensors to collect fuel density and viscosity data in real time. The collected mass flow and volume flow signals are dynamically calibrated and fused based on an improved gain-adaptive Kalman filter algorithm and an online rheological model to compensate for measurement errors caused by changes in volatile matter and temperature and pressure fluctuations.

5. The multivariable energy efficiency optimization control system for a heating furnace according to claim 4, characterized in that: The fuel flow detection module synchronously inputs the collected Coriolis mass flow signal and ultrasonic volume flow signal, along with the fuel density and viscosity parameters acquired in real time by the microwave dielectric analyzer, into a dual-channel extended Kalman filter. The process noise covariance of the first channel is dynamically updated based on an online regression reconstruction model of changes in fuel volatile content and historical operating deviations. The measurement noise covariance of the second channel is adjusted in real time through temperature-pressure hysteresis characteristic identification and an adaptive recursive least squares algorithm. After filter fusion, a secondary correction based on residual orthogonal projection is performed to accurately compensate for flow measurement errors caused by changes in fuel volatility and temperature and pressure fluctuations.

6. The multivariable energy efficiency optimization control system for a heating furnace according to claim 1, characterized in that: The combustion air flow detection module arranges vortex flowmeters and multi-point static pressure sensors at the inlet of the combustion air duct and the branch pipes of each combustion zone, and sets temperature and humidity sensors upstream and downstream of the flowmeter to collect environmental parameters. Based on the collected static pressure difference, vortex pulse frequency and temperature and humidity data, a nonlinear adaptive observer algorithm combined with the CFD simulation pre-calibration curve is used to estimate the velocity profile distortion coefficient in real time and compensate the vortex measurement value.

7. The multivariable energy efficiency optimization control system for a heating furnace according to claim 1, characterized in that: The seal wear monitoring module includes: a fiber Bragg grating strain sensor array distributed circumferentially on the outer edges of the sealing rings at the head and tail of the cracking heating furnace, which is used to collect small radial deformations of the sealing rings in real time; a high-frequency ultrasonic transducer is embedded at the contact surface of the sealing ring to detect ultrasonic signals generated by wear and gas leakage; an industrial infrared thermal imager is set on the outside of the sealing part to capture the local temperature field mutation signal caused by leaking flue gas; the three-way sensor signals are processed online by a multi-modal deep learning algorithm that integrates multi-scale wavelet packet decomposition and bidirectional long and short-term memory network to detect the seal wear and leakage flow rate in real time, and the detection results are fed back to the optimization control decision module in real time for dynamic compensation.

8. The multivariable energy efficiency optimization control system for a heating furnace according to claim 1, characterized in that: The waste heat recovery monitoring module adopts a multi-parameter coupled online detection method, which specifically includes: arranging K-type thermocouple arrays at the inlet and outlet flues of the heat exchanger for real-time measurement of the inlet and outlet flue gas temperatures; arranging multi-point thin-film heat flux meters in the heat exchanger tube bundle for real-time collection of the heat flux per unit area of ​​the tube wall; installing an acoustic imaging sensor array on the outside of the tube bundle for online identification of scaling location and thickness distribution; the waste heat recovery monitoring module dynamically calculates the heat transfer coefficient based on the bidirectional heat balance equation and the adaptive wall thermal resistance model, and combines the wall thermal resistance change rate threshold to trigger the pipe cleaning cycle prediction algorithm to locate and quantitatively evaluate the decline in heat transfer efficiency caused by the initial stage of scaling and uneven distribution.

9. The multivariable energy efficiency optimization control system for a heating furnace according to claim 1, characterized in that: The multivariable coupling modeling and simulation module adopts a hierarchical dynamic adaptive modeling approach, including: a multi-physics field simulation engine based on finite element heat conduction, CFD flow, discrete element material circulation and combustion chemical reaction coupling; the simulation engine integrates a real-time online data assimilation algorithm, injects monitoring data into the simulation model at a frequency of seconds, and self-calibrates thermal inertia, material residence time and two-phase flow boundary conditions through a model error update mechanism driven by feedback residuals, gradually converging the simulation error to within ±0.5%; the simulation calculation runs on the parallel architecture of a GPU cluster.

10. The multivariable energy efficiency optimization control system for a heating furnace according to claim 1, characterized in that: The optimization control module includes: a multi-objective optimization engine based on model predictive control and an improved non-dominated sorting genetic algorithm, which is used to jointly solve the three objectives of maximizing overall thermal efficiency, homogenizing local temperature gradients, and minimizing heat loss; the engine automatically adjusts the weights of each objective through online sensitivity analysis, and generates a Pareto frontier solution set in parallel within the feasible domain of each control variable; when a local hot spot or cold spot deviates from the threshold and is triggered, the optimization control module automatically calls a local compensation sub-algorithm based on quadratic surface fitting to update the control strategy, and sends the final optimal control instructions to the execution and feedback module in real time.

Images