A high-efficiency stable and synergistic combustion system for blending raw coal with high-calorific-value waste residue

Abstract

The present application relates to the technical field of thermal energy and power engineering, and discloses a kind of high heat value waste residue efficient stable collaborative combustion system of raw coal blending, comprising: waste residue crushing screening pretreatment module, multi-source online detection module, waste residue adaptability self-learning module, weighing closed-loop blending module, grading air distribution combustion optimization module, abnormal early warning and self-healing module of furnace condition, local central collaborative control module;Through multi-source online detection module, the calorific value, ash content, volatile matter, sulfur content, moisture content and other key combustion characteristic parameters of raw coal and waste residue are detected in real time, fuel characteristic data set is generated, the dynamic change of fuel quality can be mastered in real time, data basis is provided for subsequent blending decision, avoid the problem of blending ratio error caused by fuel quality fluctuation, ensure the accuracy of blending process.

Description

Technical Field

[0001] This invention relates to the field of thermal energy and power engineering technology, specifically to a high-efficiency and stable synergistic combustion system for raw coal blended with high-calorific-value waste residue. Background Technology

[0002] my country's coal resources are unevenly distributed and diverse. To ensure production or improve operational efficiency, thermal power plants often passively or actively blend coals not designed for their operations. Coal blending technology has become an important means to improve the safety, environmental friendliness, and economy of thermal power generating units. The primary principle of coal blending is to ensure the combustion performance of the blended coals during the blending process, that is, to ensure the combustion stability of the mixed coal within the boiler.

[0003] Currently, due to the dynamic fluctuations in fuel quality during the blending and combustion of raw coal and high-calorific-value waste residue, existing systems lack the ability to monitor the combustion characteristics of raw coal and waste residue in real time. This makes it impossible to grasp the dynamic changes in fuel quality in real time, resulting in the blending ratio decision relying on static experience data. This can easily lead to inaccurate blending ratios due to fuel quality fluctuations, and the inability to guarantee the accuracy of the blending process and the consistency of the mixed fuel quality.

[0004] Therefore, a high-efficiency and stable synergistic combustion system for raw coal blended with high-calorific-value waste residue is proposed to solve the above problems. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention provides a high-efficiency and stable synergistic combustion system for raw coal blended with high-calorific-value waste residue, which solves the problems mentioned in the background technology of not being able to guarantee the accuracy of the blending process and the consistency of the mixed fuel quality.

[0006] To achieve the above objectives, the present invention provides the following technical solution: a high-efficiency and stable synergistic combustion system for raw coal blended with high-calorific-value waste residue, the system comprising: The waste residue crushing and screening pretreatment module is used to crush and screen high-calorific-value waste residue and output pretreated waste residue that meets the particle size requirements. The multi-source online detection module is used to detect key combustion characteristic parameters of raw coal and pretreated waste residue in real time and generate fuel characteristic datasets. The waste residue adaptive self-learning module receives fuel characteristic datasets and blending execution feedback, updates deep learning model parameters through transfer learning algorithms, and outputs an adaptive blending strategy. The weighing closed-loop blending module receives adaptive blending strategies and fuel characteristic datasets, dynamically calculates the optimal blending ratio through a built-in deep learning algorithm, and executes closed-loop control to dynamically adjust the instantaneous feed flow of raw coal and waste residue, and outputs mixed fuel. The graded air distribution combustion optimization module receives mixed fuel and acquires its furnace input characteristic data, including calorific value, volatile matter content and fineness parameters. It adjusts the air supply parameters of the combustion zone in real time, stabilizes the temperature field and flow field inside the furnace, and outputs a stable combustion state indication signal. The furnace abnormality early warning and self-healing module receives stable combustion state indication signals and furnace safety parameters, including furnace wall temperature distribution, furnace internal pressure and negative pressure fluctuation characteristics. It generates early warning signals and emergency protection commands through an abnormal pattern recognition algorithm and feeds them back to the local central collaborative control module. The local central coordination control module receives feedback data from early warning signals and stable combustion status indication signals. By coordinating the operating parameters of each module, it generates global optimization control commands and distributes them to each module.

[0007] Preferably, the waste residue crushing and screening pretreatment module includes: The multi-stage crushing unit sequentially performs coarse and fine crushing on the high-calorific-value waste residue, reducing the particle size of the waste residue to a preset particle size distribution range. The vibrating screening unit classifies the crushed waste residue by particle size, screens out the fine material that meets the requirements for entering the mill, and returns the unqualified coarse material to the crushing unit for further processing. The iron and impurity separation unit automatically removes metal impurities from the waste residue during the conveying process, filters non-combustible particles using a screening device, and separates and discharges non-combustible impurities.

[0008] Preferably, the multi-source online detection module includes: The near-infrared spectroscopy analysis unit scans the surface spectral characteristics of raw coal and pretreated waste residue in real time, and retrieves the contents of calorific value, ash content and volatile matter. The X-ray fluorescence analysis unit simultaneously detects the content of sulfur, chlorine, and alkali metal elements in raw coal and pretreated waste residue, generating data on the concentration of harmful elements. The microwave moisture measurement unit uses the microwave penetration characteristics to measure the total moisture content of raw coal and pretreated waste residue in real time. The total moisture content, the content values ​​inverted by the near-infrared spectroscopy analysis unit, and the harmful element concentration data generated by the X-ray fluorescence analysis unit are used as key combustion characteristic parameters of raw coal and pretreated waste residue, and are integrated into a fuel characteristic dataset containing timestamps and material batch numbers.

[0009] Preferably, the waste residue adaptive self-learning module includes: The rapid calibration unit performs offline small-sample analysis on a new batch of waste residue to obtain its basic calorific value, ash fusion characteristics and combustion kinetic parameters, and forms an initial characteristic label. The transfer learning adaptation unit uses the initial feature labels as prior knowledge to fine-tune the parameters of the existing deep learning prediction model and records the combustion behavior feature mapping relationship of the new batch of waste residue. The online verification and correction unit compares the actual combustion effect with the model prediction value in the initial stage of adding a new batch of waste residue for co-firing. Based on the deviation between the actual combustion effect and the model prediction value, it corrects the internal weights of the model through backpropagation and updates the parameter set of the deep learning prediction model. The strategy output unit receives the updated deep learning prediction model parameter set, combines it with the fuel characteristic dataset, and determines the actual blending ratio deviation based on the instantaneous feed flow rate of raw coal and waste residue fed back by the weighing closed-loop blending module, and calculates and generates an adaptive blending strategy.

[0010] Preferably, the online verification and correction unit includes: The deviation calculation subunit compares the model-predicted blending ratio with the actual blending ratio during the waste residue co-firing process in real time, and calculates the absolute error value between the two. The weighted backpropagation subunit triggers the model training mechanism when the absolute error value exceeds the preset confidence interval threshold. It then uses actual running data to perform gradient descent correction on the weights of the underlying neurons of the deep learning prediction model and updates the parameter set of the deep learning prediction model.

[0011] Preferably, the weighing closed-loop blending module includes: The deep learning prediction unit receives an adaptive blending strategy and a fuel characteristic dataset. Taking the fuel characteristic dataset as input, it introduces the adaptive blending strategy as a dynamic constraint for multi-objective optimization. The multi-objective optimization function is based on clinker heat consumption, pollutant emission, and combustion stability values. The optimal blending ratio under the current operating conditions is iteratively solved through a locally deployed lightweight neural network model. The double-helix quantitative feeding unit controls the instantaneous feeding flow rates of raw coal and pretreated waste residue according to the optimal blending ratio. The belt weighing closed-loop control unit collects the actual weight of the material on the belt in real time and compares the difference with the target feed flow rate. The PID controller dynamically adjusts the speed of the twin-screw feeder to adjust the blending error to the preset accuracy range and outputs mixed fuel.

[0012] Preferably, the staged air distribution combustion optimization module includes: The primary air regulating unit receives the furnace input characteristic data of the mixed fuel, extracts the volatile matter content and fineness parameters of the fuel, determines the fuel ignition characteristics and conveying requirements, adjusts the injection speed and momentum of the primary air, determines the combustion stability of the primary air zone, and outputs a combustion state indication signal as a component of the stable combustion state indication signal. The secondary air regulation unit receives temperature field distribution data and oxygen content signals in the furnace, calculates the required aerodynamic field shape in each zone, adjusts the swirl intensity and axial ratio of the secondary air, monitors the combustion conditions in the secondary air zone, and feeds back combustion adjustment signals to correct the stable combustion state indication signals. The tertiary air conditioning unit monitors the concentration of residual combustibles and gas composition in the burnout zone, introduces tertiary air into the burnout zone to replenish the oxygen required for the oxidation of residual combustibles, determines the burnout effect, and outputs a combustion completion signal as the basis for terminating the stable combustion state indication signal.

[0013] Preferably, the furnace condition anomaly early warning and self-healing module includes: The coking and scaling monitoring unit identifies potential coking and scaling areas by analyzing the uniformity of temperature distribution and negative pressure fluctuation characteristics on the furnace wall, generates early warning signals and sends them to the early warning classification and linkage response unit, and generates coking removal operation suggestions at the same time. The deflagration and flashback protection unit monitors the rate of pressure and temperature change at the burner inlet in real time. When abnormal fluctuations are detected, it immediately cuts off the fuel supply and opens the inert gas purging pipeline. The early warning classification and linkage response unit determines the severity of the abnormal event based on the abnormal parameters contained in the early warning signal, classifies the early warning signal into different levels, and sends the corresponding linkage response request to the local central coordination control module. The local central coordination control module then coordinates other modules to perform load reduction, fuel switching, and shutdown maintenance operations. The self-healing recovery unit reads the preset slow-start control parameters after the abnormal event is resolved, and performs a step-by-step increase operation of fuel supply flow rate, primary air volume, secondary air volume and furnace temperature.

[0014] Preferably, the early warning classification and linkage response unit includes: The signal priority determination subunit assigns different warning signals execution priorities based on the degree of impact of abnormal events on equipment safety and system stability. The multi-module coordinated braking subunit, when receiving the highest level warning signal, simultaneously sends a reduction command to the weighing closed-loop blending module, a stop command to the staged air distribution and combustion optimization module, and locks the operation of the waste residue crushing, screening and pretreatment module, forcibly cutting off the hazard source.

[0015] Preferably, the local central collaborative control module includes a data aggregation and status awareness unit, a collaborative optimization decision-making unit, and an instruction distribution and execution monitoring unit. The data aggregation and status sensing unit collects real-time data on the equipment operation status of the waste residue crushing and screening pretreatment module, the fuel characteristics of the multi-source online detection module, the execution accuracy of the weighing closed-loop blending module, and the furnace temperature and flue gas composition data of the staged air distribution and combustion optimization module, thus constructing a digital twin mirror. The collaborative optimization decision-making unit, based on the digital twin image, uses the built-in expert rule base and reinforcement learning algorithm to comprehensively evaluate the overall energy efficiency and emission level of the current system, and generate collaborative control target values ​​for each module. The instruction distribution and execution monitoring unit converts the coordinated control target value into standardized control instructions, distributes them to the corresponding modules, and continuously monitors the execution effect of the instructions to form a closed-loop control circuit.

[0016] Compared with the prior art, the present invention has the following beneficial effects: 1. The multi-source online detection module monitors key combustion characteristic parameters such as calorific value, ash content, volatile matter, sulfur content, and moisture content of raw coal and pretreated waste residue in real time, generating a fuel characteristic dataset. This allows for real-time monitoring of dynamic changes in fuel quality, providing a data foundation for subsequent blending decisions and avoiding inaccurate blending ratios due to fuel quality fluctuations, thus ensuring the accuracy of the blending process. The weighing closed-loop blending module receives the fuel characteristic dataset, uses deep learning algorithms to dynamically calculate the optimal blending ratio, and executes closed-loop control to dynamically adjust the instantaneous feed flow rate. This enables real-time correction of blending deviations and reduces the risk of combustion instability caused by static proportions.

[0017] 2. The weighing closed-loop blending module receives adaptive blending strategies and fuel characteristic datasets, dynamically calculates the optimal blending ratio using deep learning algorithms, and adjusts the instantaneous feed flow of raw coal and waste residue in real time through the belt weighing closed-loop control unit to adjust the blending error to a preset accuracy range, thus achieving closed-loop blending control and ensuring the consistency of mixed fuel quality. The staged air distribution combustion optimization module receives the furnace input characteristic data of mixed fuels, and the primary air adjustment unit, secondary air adjustment unit, and tertiary air adjustment unit adjust the air supply parameters of each zone respectively. This can optimize the aerodynamic field of the combustion zone in real time according to fuel characteristics, maintain the stability of the furnace temperature field and flow field, and reduce the impact of combustion fluctuations on the system.

[0018] 3. In this invention, the furnace condition abnormality early warning and self-healing module receives stable combustion state indication signals and furnace safety parameters, the coking and scaling monitoring unit identifies potential coking areas, the deflagration and flashback protection unit cuts off fuel supply and opens inert gas purging pipeline when abnormal fluctuations are detected, and the self-healing recovery unit gradually restores fuel supply and combustion parameters after the abnormality is resolved. This enables early identification and automatic recovery of furnace condition abnormalities, improving the safety and continuity of system operation. Attached Figure Description

[0019] Figure 1 This is a structural diagram of a high-efficiency and stable synergistic combustion system for raw coal blended with high-calorific-value waste residue according to the present invention; Figure 2 This is a flowchart illustrating the operation steps of a high-efficiency and stable synergistic combustion system for raw coal blended with high-calorific-value waste residue, as described in this invention. Detailed Implementation

[0020] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0021] Example Please see Figures 1-2 A high-efficiency and stable co-combustion system for raw coal blended with high-calorific-value waste residue, comprising: The waste residue crushing and screening pretreatment module is used to crush and screen high-calorific-value waste residue and output pretreated waste residue that meets the particle size requirements. The multi-source online detection module is used to detect key combustion characteristic parameters of raw coal and pretreated waste residue in real time and generate fuel characteristic datasets. The waste residue adaptive self-learning module receives fuel characteristic datasets and blending execution feedback, updates deep learning model parameters through transfer learning algorithms, and outputs an adaptive blending strategy. The weighing closed-loop blending module receives adaptive blending strategies and fuel characteristic datasets, dynamically calculates the optimal blending ratio through a built-in deep learning algorithm, and executes closed-loop control to dynamically adjust the instantaneous feed flow of raw coal and waste residue, and outputs mixed fuel. The staged air distribution combustion optimization module receives mixed fuel and acquires its furnace input characteristic data, including calorific value, volatile matter content and fineness parameters. It adjusts the air supply parameters of the combustion zone in real time, stabilizes the temperature field and flow field inside the furnace, and outputs a stable combustion state indication signal. The furnace anomaly early warning and self-healing module receives stable combustion status indication signals and furnace safety parameters, including furnace wall temperature distribution, internal furnace pressure, and negative pressure fluctuation characteristics. It generates early warning signals and emergency protection commands through an anomaly pattern recognition algorithm and feeds them back to the local central collaborative control module. The specific implementation steps of the anomaly pattern recognition algorithm include: Collect normal and abnormal operating condition data from historical operating data in advance, and mark various abnormal events, including sensor signal characteristics corresponding to coking, scaling, deflagration, and flashback; A classification model based on decision trees is constructed. The classification model uses the uniformity of furnace wall temperature distribution, negative pressure fluctuation amplitude, burner inlet pressure change rate and temperature change rate as node splitting features. In each non-leaf node, one feature and its splitting value are selected to divide the current dataset into left and right subsets. This splitting process is executed recursively until the samples in each leaf node belong to the same category. Thus, a classification rule with a tree-like hierarchical structure is constructed. The real-time collected furnace safety parameters are input into the classification model, and the classification model outputs the anomaly category of the current operating condition and its confidence level. When the confidence level exceeds the preset threshold, it is determined that the corresponding abnormal event has occurred, and a corresponding early warning signal and emergency protection instruction are generated. The local central coordination control module receives feedback data from early warning signals and stable combustion status indication signals. By coordinating the operating parameters of each module, it generates global optimization control commands and distributes them to each module.

[0022] Specifically, the waste residue crushing, screening, and pretreatment module includes: The multi-stage crushing unit sequentially performs coarse and fine crushing on the high-calorific-value waste residue, reducing the particle size of the waste residue to a preset particle size distribution range. The vibrating screening unit classifies the crushed waste residue by particle size, screens out the fine material that meets the requirements for entering the mill, and returns the unqualified coarse material to the crushing unit for further processing. The iron and impurity separation unit automatically removes metal impurities from the waste residue during the conveying process, filters non-combustible particles using a screening device, and separates and discharges non-combustible impurities.

[0023] The multi-source online detection module includes: The near-infrared spectroscopy analysis unit scans the surface spectral characteristics of raw coal and pretreated waste residue in real time, and retrieves the contents of calorific value, ash content and volatile matter. The X-ray fluorescence analysis unit simultaneously detects the content of sulfur, chlorine, and alkali metal elements in raw coal and pretreated waste residue, generating data on the concentration of harmful elements. The microwave moisture measurement unit uses the microwave penetration characteristics to measure the total moisture content of raw coal and pretreated waste residue in real time. The total moisture content, the content values ​​inverted by the near-infrared spectroscopy analysis unit, and the harmful element concentration data generated by the X-ray fluorescence analysis unit are used as key combustion characteristic parameters of raw coal and pretreated waste residue, and integrated into a fuel characteristic dataset containing timestamps and material batch numbers.

[0024] The waste residue adaptive self-learning module includes: The rapid calibration unit performs offline small-sample analysis on a new batch of waste residue to obtain its basic calorific value, ash fusion characteristics and combustion kinetic parameters, and forms an initial characteristic label. The transfer learning adaptation unit uses the initial feature labels as prior knowledge to fine-tune the parameters of the existing deep learning prediction model and records the feature mapping relationship of the combustion behavior of the new batch of waste residue. The specific steps include: The initial feature labels generated by the fast calibration unit are used as source domain data, and the parameters of the existing deep learning prediction model are used as initial parameters of the target domain. Freeze the weight parameters of the first three fully connected layers in the existing deep learning prediction model, and only allow the weight parameters of the last two fully connected layers to be updated during training; Using offline analysis data of small samples from the new batch of waste residue as training samples, and the cross-entropy loss function as the training objective, the learning rate was set to one-tenth of the original learning rate; the expression for the cross-entropy loss function is: ; in, The total number of training samples, For the first The true label of each sample The probability value predicted by the model; Iterative training continues until the cross-entropy loss function value stabilizes, and the mapping relationship of the combustion behavior characteristics of the new waste residue is recorded, i.e., the updated weight parameters of the last two layers. The online verification and correction unit compares the actual combustion effect with the model prediction value in the initial stage of the system's input of a new batch of waste residue for co-firing. Based on the deviation between the actual combustion effect and the model prediction value, it corrects the internal weights of the model through backpropagation and updates the parameter set of the deep learning prediction model. The strategy output unit receives the updated deep learning prediction model parameter set, combines it with the fuel characteristic dataset, and determines the actual blending ratio deviation based on the instantaneous feed flow rate of raw coal and waste residue fed back by the weighing closed-loop blending module, and calculates and generates an adaptive blending strategy.

[0025] The online verification and correction unit includes: The deviation calculation subunit compares the model-predicted blending ratio with the actual blending ratio during the waste residue co-firing process in real time, and calculates the absolute error value between the two. Specific steps include: First, obtain the predicted blending ratio vector at the current time point output by the existing deep learning prediction model, denoted as... ; Secondly, the instantaneous mass flow rate of the waste residue falling into the mixing silo is collected in real time using a high-precision weighing sensor, and combined with the feeder speed feedback, the actual blending ratio vector is calculated, denoted as... ; Finally, the absolute error between the two is calculated using the Euclidean distance formula, which is as follows: ; in, This is the absolute error value. The number of feature dimensions used in the comparison. For the prediction vector of the th dimensional components, For the actual vector of the th Dimensional components; The weighted backpropagation subunit triggers the model training mechanism when the absolute error value exceeds a preset confidence interval threshold. It then uses actual runtime data to perform gradient descent correction on the weights of the underlying neurons in the deep learning prediction model, updating the parameter set of the deep learning prediction model. Specific steps include: When the calculated absolute error value Exceeding the preset confidence interval threshold When the time is right, the model training mechanism is triggered, and the actual running data of the current batch is extracted as the training sample set; Calculate the gradient of the loss function with respect to the weights of the neural network output layer, and use the chain rule to propagate the error layer by layer forward. The formula for calculating the error term of the intermediate layer is as follows: ; in, For the first Layer error term, For the first The weight matrix of the layer, This is the error term for the next layer. For the first The derivative of the layer activation function, Represents element-wise multiplication of matrices; Based on the calculated gradients of each layer, the weights of the bottom-level neurons are updated along the negative gradient direction. The formula for calculating the update amount is as follows: ; in, For the first The amount of layer weight update, For learning rate, For the first The output feature matrix of the layer, Indicates matrix transpose; Repeat the above error backpropagation and weight update process until the error value converges below the threshold, thus completing the correction of the model parameter set.

[0026] The weighing closed-loop blending module includes: The deep learning prediction unit receives an adaptive blending strategy and a fuel characteristic dataset. Using the fuel characteristic dataset as input, it introduces the adaptive blending strategy as a dynamic constraint for multi-objective optimization. The multi-objective optimization function uses clinker heat consumption, pollutant emissions, and combustion stability as values. Through a locally deployed lightweight neural network model, it iteratively solves for the optimal blending ratio under the current operating conditions. Specifically, this is achieved through the following iterative steps: The first step is to receive the fuel characteristic dataset and the adaptive blending strategy, using the calorific value, ash content, volatile matter, sulfur content, moisture content and harmful element concentration parameters in the fuel characteristic dataset as the input feature vector. The second step is to convert the adaptive blending strategy into constrained weight coefficients, which are then superimposed on the input feature vector to form a weighted input matrix. The third step involves inputting the weighted input matrix into a locally deployed lightweight neural network model. Through forward propagation calculations using multiple fully connected layers, the candidate blending ratios under the current operating conditions are output. The forward propagation calculation process is represented by the following formula: ; in, Indicates the first The output vector of the hidden layer. Indicates the first The weight matrix of the layer, Indicates the first The layer's bias vector, Represents a non-linear activation function. The input feature vector; The fourth step is to substitute the candidate blending ratios into the preset multi-objective optimization function to calculate the corresponding heat consumption, emission, and stability values. Fifth step: If the optimization function value does not reach the convergence condition, the connection weights of neurons in each layer of the neural network are adjusted through the backpropagation algorithm, and steps three to five are repeated until the optimization function value converges, and the optimal blending ratio under the current working condition is output. The process of constructing a multi-objective optimization function includes: The clinker heat consumption value, pollutant emission value, and combustion stability value were set as three independent optimization targets. A baseline reference value is set for each optimization objective. The actual output value is compared with the baseline reference value to obtain the degree of deviation from each objective. The formula for calculating the degree of deviation is as follows: ; in, Indicates the first The degree of deviation from the optimization objective, This represents the actual output value. This represents the baseline reference value for the corresponding target; Different weighting coefficients are assigned to the degree of deviation of each target, with the weight of heat consumption value being higher than that of emission value, and the weight of emission value being higher than that of stability value. The weighted deviations of each objective are summed to obtain a comprehensive optimization evaluation value, and minimizing this evaluation value is taken as the optimization direction; the expression for the comprehensive optimization evaluation value is: ; in, Indicates the first The weight coefficients of each optimization objective, and satisfying , To comprehensively optimize the evaluation value; The double-helix quantitative feeding unit controls the instantaneous feeding flow rate of raw coal and pretreated waste residue according to the optimal blending ratio. The belt weighing closed-loop control unit collects the actual weight of the material on the belt in real time and compares it with the target feed flow rate. It then dynamically adjusts the speed of the twin-screw feeder via a PID controller to regulate the blending error within a preset accuracy range before outputting the mixed fuel. Specific steps include: The actual weight of the material on the conveyor belt is collected in real time, and the difference between this weight and the target feed flow rate is calculated as an error signal. The formula for calculating the error signal is: ; in, For a moment The error signal, Feed flow rate to target, This represents the actual weight of the collected materials. The error signal is input to the PID controller, and the controller output adjustment is applied to the rotor of the twin-helix feeder; the output expression of the PID controller is: ; in, For the controller output value, This is the proportionality coefficient. The integral coefficient is... These are the differential coefficients. Deviation signal The first derivative with respect to time, Indicates from the start time to Accumulated deviations over time For integration variables; The rotational speed of the twin-helix feeder is dynamically adjusted according to the controller output value, so that the actual feed flow rate approaches the target value and the blending error is adjusted to the preset accuracy range.

[0027] The staged air distribution and combustion optimization module includes: The primary air regulating unit receives the furnace feed characteristic data of the mixed fuel, extracts the volatile matter content and fineness parameters of the fuel, determines the fuel ignition characteristics and conveying requirements, adjusts the primary air injection velocity and carried momentum, and determines the combustion stability of the primary air zone, outputting a combustion state indication signal as a component of the stable combustion state indication signal; the specific steps include: First, based on the extracted volatile matter content and fineness parameters of the mixed fuel, the ignition tendency index of the fuel is evaluated using a linear weighted function. The calculation formula is as follows: ; in, The fire tendency index value, This refers to the percentage of volatile matter content in the fuel. This represents the percentage of fuel residue on a fine sieve. and These are the influence weighting coefficients for volatile matter and fineness, respectively. Based on the magnitude of the ignition tendency index value, the level of fuel transportation demand is classified. When the index value is greater than the preset threshold, it is determined that the fuel requires high air pressure to meet the transportation and enhanced ignition requirements; otherwise, it is determined to be a normal transportation requirement. Secondly, the primary air injection velocity is calculated simultaneously based on the judgment results. The injection velocity has a positive correlation with the ignition tendency index, and its calculation expression is as follows: ; in, For the target jet speed, Based on the injection speed, For speed adjustment gain coefficient, This is the fire tendency index value; Get the current primary wind jet speed With density and the cross-sectional area of ​​the primary air duct. According to the momentum theorem in fluid mechanics, the air volume of a single injection is calculated using the following formula: ; in, The momentum flux of the air volume; Finally, the calculated air volume momentum flux is compared with the physical momentum of the fuel particles, and the injection speed is dynamically adjusted to keep the total momentum of the air-powder mixture within the preset ideal range, so as to ensure the suspension and transport state of the fuel particles. The secondary air regulation unit receives temperature field distribution data and oxygen content signals within the furnace, calculates the required aerodynamic field shape for each zone, adjusts the swirl intensity and axial ratio of the secondary air, monitors the combustion conditions in the secondary air zone, and feeds back combustion adjustment signals to correct the stable combustion state indication signals. Specific steps include: Receive temperature field distribution data at different heights and radii of the furnace, and fit the temperature gradient field of the furnace cross section; Based on the characteristics of the temperature gradient field, the combustion zone of the furnace is divided into multiple computational grid cells. Within each computational grid cell, the required local excess air coefficient is calculated based on the ratio of the local temperature value to the average temperature value. The calculation formula is as follows: ; in, The coefficient for local excess air. This is the actual temperature value of the unit. The reference temperature under ideal complete combustion conditions; The local excess air coefficient of each computational grid cell is mapped to the corresponding aerodynamic field morphology features, including the central jet length and the size of the recirculation zone; Based on the aerodynamic field morphology characteristics, the target value of the swirl number of the secondary wind is determined; by adjusting the blade angle of the secondary wind nozzle, the rotational and axial components of the airflow are changed, thereby altering the swirl intensity. The adjustment amount for the swirl intensity is calculated using the following formula: ; in, This refers to the adjustment angular displacement of the hydrocyclone blades. For the target swirl number, This represents the currently measured swirl number. The sensitivity coefficient of the actuator; Simultaneously adjust the distribution ratio of secondary air in the central and peripheral areas of the furnace according to the axial ratio requirements to maintain a stable combustion center; The tertiary air conditioning unit monitors the concentration of residual combustibles and gas composition in the burnout zone, introduces tertiary air into the burnout zone to replenish the oxygen required for the oxidation of residual combustibles, determines the burnout effect, and outputs a combustion completion signal as the basis for terminating the stable combustion state indication signal.

[0028] The furnace condition anomaly early warning and self-healing module includes: The coking and scaling monitoring unit analyzes the uniformity of temperature distribution and negative pressure fluctuation characteristics on the furnace wall to identify potential coking and scaling areas, generates early warning signals, and sends them to the early warning classification and linkage response unit. Simultaneously, it generates coking removal operation suggestions, the specific steps of which include: Multiple temperature measurement points are arranged along the height of the furnace to collect temperature values ​​at each point in real time, and the temperature distribution uniformity index is calculated. The formula for calculating the uniformity index is: ; in, The standard deviation of the temperature at each measuring point. This is the average temperature at each measuring point. This is an index of temperature distribution uniformity, with a value ranging from 0 to 1. The lower the value, the more uneven the temperature distribution. Synchronously acquire the furnace negative pressure signal and calculate the variance of the negative pressure fluctuation. The expression is as follows: ; in, The variance of the negative pressure fluctuation. The number of sampling points. For the first The negative pressure value at each sampling point The average negative pressure during the sampling period; When the uniformity index is lower than the first preset threshold and the negative pressure fluctuation variance is higher than the second preset threshold, it is determined that there is a risk of coking and skin formation in the corresponding area, and a coking removal operation suggestion is generated. The deflagration and flashback protection unit monitors the rate of pressure and temperature change at the burner inlet in real time. When abnormal fluctuations are detected, it immediately cuts off the fuel supply and opens the inert gas purging pipeline. The early warning classification and linkage response unit determines the severity of the abnormal event based on the abnormal parameters contained in the early warning signal, classifies the early warning signal into different levels, and sends the corresponding linkage response request to the local central coordination control module. The local central coordination control module then coordinates other modules to perform load reduction, fuel switching, and shutdown maintenance operations. The self-healing recovery unit reads the preset slow-start control parameters after the abnormal event is resolved, and performs a step-by-step increase operation of fuel supply flow rate, primary air volume, secondary air volume and furnace temperature.

[0029] The early warning classification and joint response unit includes: The signal priority determination subunit assigns different warning signals execution priorities based on the degree of impact of abnormal events on equipment safety and system stability. The multi-module coordinated braking subunit, upon receiving the highest-level warning signal, simultaneously sends a reduction command to the weighing closed-loop blending module, a stop-air command to the staged air distribution and combustion optimization module, and locks the operation of the waste residue crushing, screening, and pretreatment module, forcibly cutting off the hazard source. Specific steps include: Upon receiving the highest-level warning signal, calculate the response time required for emergency braking based on the current operating parameters of each module. This response time consists of communication delay and actuator action time, and its estimation formula is as follows: ; in, For response time, Communication delay for signal transmission, The mechanical action time of the actuator; A reduction command is simultaneously sent to the weighing closed-loop blending module to reduce the feed flow rate. It decreases linearly to zero over time; the slope of the decrease is determined by the following formula: ; in, This is the current feed rate. Reduce the flow rate; Send a stop command to the staged air distribution and combustion optimization module, and set the blower damper opening to [value missing]. Shut down to zero within a specified time; Locking the operation of the waste residue crushing, screening, and pretreatment module prevents subsequent materials from entering the system, thus forcibly cutting off the source of danger.

[0030] The local central collaborative control module includes: The data aggregation and status sensing unit collects real-time data on the equipment operation status of the waste residue crushing and screening pretreatment module, the fuel characteristics of the multi-source online detection module, the execution accuracy of the weighing closed-loop blending module, and the furnace temperature and flue gas composition data of the staged air distribution and combustion optimization module, thus constructing a digital twin mirror. The collaborative optimization decision-making unit, based on a digital twin mirror image, utilizes a built-in expert rule base and reinforcement learning algorithm to comprehensively evaluate the overall energy efficiency and emission levels of the current system, generating collaborative control target values ​​for each module. The construction content of the expert rule base and the decision-making process of the reinforcement learning algorithm include: The expert rule base stores multiple experience rules. Each rule contains a condition part and an action part. The condition part describes a specific combination of operating condition parameters, and the action part specifies the adjustment amount of the corresponding module control parameters. The reinforcement learning algorithm uses the system state in the digital twin mirror as the environment state, the adjustable parameters of each module as the action space, and a weighted score combining energy efficiency and emission levels as the reward signal; the formula for calculating the reward signal is: ; in, As a reward signal, To normalize the energy efficiency savings rate, To normalize pollutant emission rates, To normalize the combustion stability index, The corresponding weight coefficients and ; The algorithm iteratively updates the state-action value function using a Q-learning framework, selecting the action that maximizes the expected cumulative reward as the target value for collaborative regulation at each decision time; the Q-value update formula is: ; in, For state-action value functions, This is the current state. For the current action, For learning rate, As a discount factor, For the next state, The executable actions for the next state; In the early stages of training, the algorithm uses the actions provided by the expert rule base as the initial strategy, gradually explores better actions, and finally converges to the globally optimal control strategy. The instruction distribution and execution monitoring unit converts the collaborative control target value into standardized control instructions, distributes them to the corresponding modules, and continuously monitors the execution effect of the instructions, forming a closed-loop control circuit.

[0031] The operating steps of this efficient and stable co-combustion system for raw coal blended with high-calorific-value waste residue are as follows: Step 1: Waste residue pretreatment and multi-source online detection: High-calorific-value waste residue first enters the waste residue crushing and screening pretreatment module. After multi-stage crushing and vibrating screening, pretreated waste residue that meets the particle size requirements is output. At the same time, the multi-source online detection module uses near-infrared spectroscopy analysis unit, X-ray fluorescence analysis unit, and microwave moisture measurement unit to detect key combustion characteristic parameters such as calorific value, ash content, volatile matter, sulfur content, and moisture of raw coal and pretreated waste residue in real time, integrating them into a fuel characteristic dataset.

[0032] Step 2: Waste residue adaptive self-learning and blending strategy generation: The waste residue adaptive self-learning module receives fuel characteristic datasets and blending execution feedback. The rapid calibration unit performs small-sample offline analysis on new batches of waste residue to form initial characteristic labels; the transfer learning adaptation unit uses these labels as prior knowledge to fine-tune the parameters of the deep learning prediction model; the online verification and correction unit compares the actual combustion effect with the model prediction value and corrects the model weights through backpropagation of the deviation; the strategy output unit combines the updated model parameters with the actual blending ratio deviation to output an adaptive blending strategy.

[0033] Step 3: Weighing, closed-loop blending, and mixed fuel output: The weighing closed-loop blending module receives an adaptive blending strategy and a fuel characteristic dataset. The deep learning prediction unit uses the fuel characteristic dataset as input and the blending strategy as a dynamic constraint, iteratively solving for the optimal blending ratio using a lightweight neural network model. The double-helix quantitative feeding unit controls the instantaneous feed flow rates of raw coal and waste residue according to this ratio. The belt weighing closed-loop control unit adjusts the feeder speed in real time via a PID controller, adjusting the blending error to a preset accuracy range before outputting the mixed fuel.

[0034] Step 4: Staged air distribution combustion optimization and stable combustion: The staged air distribution and combustion optimization module receives data on the characteristics of the mixed fuel entering the furnace. The primary air regulating unit adjusts the injection velocity and momentum based on the volatile matter content and fineness, determines the combustion stability of the primary air zone, and outputs a combustion status indication signal. The secondary air regulating unit adjusts the swirl intensity and axial ratio based on the furnace temperature field and oxygen content signals, and feeds back combustion adjustment signals. The tertiary air regulating unit replenishes oxygen in the burnout zone, determines the burnout effect, outputs a combustion completion signal, maintains the stability of the furnace temperature field and flow field, and outputs a stable combustion status indication signal.

[0035] Step 5: Early Warning and Self-Healing Recovery of Abnormal Furnace Conditions The furnace condition anomaly early warning and self-healing module receives stable combustion status indication signals and furnace safety parameters. The coking and scaling monitoring unit identifies potential coking areas through temperature uniformity and negative pressure fluctuations, generating early warning signals and coking removal operation suggestions; the deflagration and flashback protection unit cuts off fuel supply and initiates inert gas purging when abnormal fluctuations are detected; the early warning classification and linkage response unit sends linkage response requests to the local central collaborative control module according to the early warning signal level; and the self-healing recovery unit gradually restores operating parameters according to a preset procedure after the anomaly is resolved.

[0036] Step Six: Local Centralized Coordination Control and Global Optimization: The data aggregation and status awareness unit of the local central collaborative control module constructs a digital twin mirror image. The collaborative optimization decision-making unit uses an expert rule base and reinforcement learning algorithm to generate collaborative control target values. The instruction distribution and execution monitoring unit converts the target values ​​into control instructions and sends them to each module, forming a closed-loop control circuit.

[0037] It should be noted that, although embodiments of the present invention have been shown and described herein, those skilled in the art will understand that various changes, modifications, substitutions and variations can be made to these embodiments without departing from the principles and spirit of the present invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A high-efficiency and stable co-combustion system for raw coal blended with high-calorific-value waste residue, characterized in that: The system includes: The waste residue crushing and screening pretreatment module is used to crush and screen high-calorific-value waste residue and output pretreated waste residue that meets the particle size requirements. The multi-source online detection module is used to detect key combustion characteristic parameters of raw coal and pretreated waste residue in real time and generate fuel characteristic datasets. The waste residue adaptive self-learning module receives fuel characteristic datasets and blending execution feedback, updates deep learning model parameters through transfer learning algorithms, and outputs an adaptive blending strategy. The weighing closed-loop blending module receives adaptive blending strategies and fuel characteristic datasets, dynamically calculates the optimal blending ratio through a built-in deep learning algorithm, and executes closed-loop control to dynamically adjust the instantaneous feed flow of raw coal and waste residue, and outputs mixed fuel. The graded air distribution combustion optimization module receives mixed fuel and acquires its furnace input characteristic data, including calorific value, volatile matter content and fineness parameters. It adjusts the air supply parameters of the combustion zone in real time, stabilizes the temperature field and flow field inside the furnace, and outputs a stable combustion state indication signal. The furnace abnormality early warning and self-healing module receives stable combustion state indication signals and furnace safety parameters, including furnace wall temperature distribution, furnace internal pressure and negative pressure fluctuation characteristics. It generates early warning signals and emergency protection commands through an abnormal pattern recognition algorithm and feeds them back to the local central collaborative control module. The local central coordination control module receives feedback data from early warning signals and stable combustion status indication signals. By coordinating the operating parameters of each module, it generates global optimization control commands and distributes them to each module.

2. The efficient and stable co-combustion system for raw coal blended with high-calorific-value waste residue according to claim 1, characterized in that: The waste residue crushing, screening, and pretreatment module includes: The multi-stage crushing unit sequentially performs coarse and fine crushing on the high-calorific-value waste residue, reducing the particle size of the waste residue to a preset particle size distribution range. The vibrating screening unit classifies the crushed waste residue by particle size, screens out the fine material that meets the requirements for entering the mill, and returns the unqualified coarse material to the crushing unit for further processing. The iron and impurity separation unit automatically removes metal impurities from the waste residue during the conveying process, filters non-combustible particles using a screening device, and separates and discharges non-combustible impurities.

3. The efficient and stable co-combustion system for raw coal blended with high-calorific-value waste residue according to claim 1, characterized in that: The multi-source online detection module includes: The near-infrared spectroscopy analysis unit scans the surface spectral characteristics of raw coal and pretreated waste residue in real time, and retrieves the contents of calorific value, ash content and volatile matter. The X-ray fluorescence analysis unit simultaneously detects the content of sulfur, chlorine, and alkali metal elements in raw coal and pretreated waste residue, generating data on the concentration of harmful elements. The microwave moisture measurement unit uses the microwave penetration characteristics to measure the total moisture content of raw coal and pretreated waste residue in real time. The total moisture content, the content values ​​inverted by the near-infrared spectroscopy analysis unit, and the harmful element concentration data generated by the X-ray fluorescence analysis unit are used as key combustion characteristic parameters of raw coal and pretreated waste residue, and integrated into a fuel characteristic dataset containing timestamps and material batch numbers.

4. The efficient and stable co-combustion system for raw coal blended with high-calorific-value waste residue according to claim 1, characterized in that: The waste residue adaptive self-learning module includes: The rapid calibration unit performs offline small-sample analysis on a new batch of waste residue to obtain its basic calorific value, ash fusion characteristics and combustion kinetic parameters, and forms an initial characteristic label. The transfer learning adaptation unit uses the initial feature labels as prior knowledge to fine-tune the parameters of the existing deep learning prediction model and records the combustion behavior feature mapping relationship of the new batch of waste residue. The online verification and correction unit compares the actual combustion effect with the model prediction value in the initial stage of adding a new batch of waste residue for co-firing. Based on the deviation between the actual combustion effect and the model prediction value, it corrects the internal weights of the model through backpropagation and updates the parameter set of the deep learning prediction model. The strategy output unit receives the updated deep learning prediction model parameter set, combines it with the fuel characteristic dataset, and determines the actual blending ratio deviation based on the instantaneous feed flow rate of raw coal and waste residue fed back by the weighing closed-loop blending module, and calculates and generates an adaptive blending strategy.

5. The efficient and stable co-combustion system for raw coal blended with high-calorific-value waste residue according to claim 4, characterized in that: The online verification and correction unit includes: The deviation calculation subunit compares the model-predicted blending ratio with the actual blending ratio during the waste residue co-firing process in real time, and calculates the absolute error value between the two. The weighted backpropagation subunit triggers the model training mechanism when the absolute error value exceeds the preset confidence interval threshold. It then uses actual running data to perform gradient descent correction on the weights of the underlying neurons of the deep learning prediction model and updates the parameter set of the deep learning prediction model.

6. The efficient and stable co-combustion system for raw coal blended with high-calorific-value waste residue according to claim 1, characterized in that: The weighing closed-loop blending module includes: The deep learning prediction unit receives an adaptive blending strategy and a fuel characteristic dataset. Taking the fuel characteristic dataset as input, it introduces the adaptive blending strategy as a dynamic constraint for multi-objective optimization. The multi-objective optimization function is based on clinker heat consumption, pollutant emission, and combustion stability values. The optimal blending ratio under the current operating conditions is iteratively solved through a locally deployed lightweight neural network model. The double-helix quantitative feeding unit controls the instantaneous feeding flow rates of raw coal and pretreated waste residue according to the optimal blending ratio. The belt weighing closed-loop control unit collects the actual weight of the material on the belt in real time and compares the difference with the target feed flow rate. The PID controller dynamically adjusts the speed of the twin-screw feeder to adjust the blending error to the preset accuracy range and outputs mixed fuel.

7. The efficient and stable co-combustion system for raw coal blended with high-calorific-value waste residue according to claim 1, characterized in that: The graded air distribution and combustion optimization module includes: The primary air regulating unit receives the furnace input characteristic data of the mixed fuel, extracts the volatile matter content and fineness parameters of the fuel, determines the fuel ignition characteristics and conveying requirements, adjusts the injection speed and momentum of the primary air, determines the combustion stability of the primary air zone, and outputs a combustion state indication signal as a component of the stable combustion state indication signal. The secondary air regulation unit receives temperature field distribution data and oxygen content signals in the furnace, calculates the required aerodynamic field shape in each zone, adjusts the swirl intensity and axial ratio of the secondary air, monitors the combustion conditions in the secondary air zone, and feeds back combustion adjustment signals to correct the stable combustion state indication signals. The tertiary air conditioning unit monitors the concentration of residual combustibles and gas composition in the burnout zone, introduces tertiary air into the burnout zone to replenish the oxygen required for the oxidation of residual combustibles, determines the burnout effect, and outputs a combustion completion signal as the basis for terminating the stable combustion state indication signal.

8. The efficient and stable co-combustion system for raw coal blended with high-calorific-value waste residue according to claim 1, characterized in that: The furnace condition anomaly early warning and self-healing module includes: The coking and scaling monitoring unit identifies potential coking and scaling areas by analyzing the uniformity of temperature distribution and negative pressure fluctuation characteristics on the furnace wall, generates early warning signals and sends them to the early warning classification and linkage response unit, and generates coking removal operation suggestions at the same time. The deflagration and flashback protection unit monitors the rate of pressure and temperature change at the burner inlet in real time. When abnormal fluctuations are detected, it immediately cuts off the fuel supply and opens the inert gas purging pipeline. The early warning classification and linkage response unit determines the severity of the abnormal event based on the abnormal parameters contained in the early warning signal, classifies the early warning signal into different levels, and sends the corresponding linkage response request to the local central coordination control module. The local central coordination control module then coordinates other modules to perform load reduction, fuel switching, and shutdown maintenance operations. The self-healing recovery unit reads the preset slow-start control parameters after the abnormal event is resolved, and performs a step-by-step increase operation of fuel supply flow rate, primary air volume, secondary air volume and furnace temperature.

9. The efficient and stable co-combustion system for raw coal blended with high-calorific-value waste residue according to claim 8, characterized in that: The early warning classification and coordinated response unit includes: The signal priority determination subunit assigns different warning signals execution priorities based on the degree of impact of abnormal events on equipment safety and system stability. The multi-module coordinated braking subunit, when receiving the highest level warning signal, simultaneously sends a reduction command to the weighing closed-loop blending module, a stop command to the staged air distribution and combustion optimization module, and locks the operation of the waste residue crushing, screening and pretreatment module, forcibly cutting off the hazard source.

10. The efficient and stable co-combustion system for raw coal blended with high-calorific-value waste residue according to claim 1, characterized in that: The local central collaborative control module includes: The data aggregation and status sensing unit collects real-time data on the equipment operation status of the waste residue crushing and screening pretreatment module, the fuel characteristics of the multi-source online detection module, the execution accuracy of the weighing closed-loop blending module, and the furnace temperature and flue gas composition data of the staged air distribution and combustion optimization module, thus constructing a digital twin mirror. The collaborative optimization decision-making unit, based on the digital twin image, uses the built-in expert rule base and reinforcement learning algorithm to comprehensively evaluate the overall energy efficiency and emission level of the current system, and generate collaborative control target values ​​for each module. The instruction distribution and execution monitoring unit converts the coordinated control target value into standardized control instructions, distributes them to the corresponding modules, and continuously monitors the execution effect of the instructions to form a closed-loop control circuit.

Images