A method for controlling the mixed combustion of multiple coal types in a fluidized bed boiler

Abstract

The present invention relates to the technical field of combustion control, and specifically discloses a method for controlling the mixed combustion of multiple coal types in a fluidized bed boiler, comprising data acquisition and state monitoring, analysis of combustion efficiency and temperature distribution, intelligent control of airflow velocity and bed height, feedback adjustment of secondary phase volume fraction, output and feedback control, and introducing penalties for insufficient combustion areas and overheating areas in a reward function to improve the uniformity and efficiency of combustion; through real-time monitoring of particle state, reinforcement learning control and secondary phase volume fraction feedback, the bed height and airflow velocity of the fluidized bed boiler are dynamically adjusted to ensure uniform combustion of anthracite and lignite and improve combustion efficiency.

Description

Technical Field

[0001] The invention relates to the technical field of combustion control, and in particular to a method for controlling the mixed combustion of multiple coal types in a fluidized bed boiler. Background Art

[0002] Fluidized bed boilers are boiler systems that can efficiently burn various types of coal by placing solid fuel in a fluidized bed, mixing it with air and suspending it in the airflow. In the fluidized bed, solid particles are suspended by the airflow (usually air), forming a so-called "fluidized" state, which can enhance the contact area and heat transfer effect between the fuel and the air, thereby improving combustion efficiency and fuel utilization. In a fluidized bed boiler burning lignite and anthracite, if the bed height is too low, the anthracite particles will be deposited at the bottom of the bed due to their high density, resulting in incomplete combustion and excessively high local temperatures in the bed. Lignite has a high volatile content and burns quickly, which can easily lead to local temperature increases and uneven combustion. If the bed height is too high, the airflow velocity is insufficient, the fuel particles are not suspended enough, especially the heavier anthracite particles, which will be deposited in the bottom area of ​​the bed, making it difficult to achieve uniform combustion, and the excessively high airflow velocity will bring the lignite particles out of the bed. Therefore, how to adjust the bed height in real time according to the different characteristics of the coal types in the fuel group to ensure that the combustion of anthracite and lignite is sufficiently uniform, especially when the fuel group and the fluidized bed load change, how to adjust the bed height and airflow velocity in time according to the suspension state and combustion state of the particles in the bed has become a problem that needs to be solved urgently. Summary of the invention

[0003] In order to overcome the above-mentioned defects of the prior art, the present invention provides a method for controlling the mixed combustion of multiple coal types in a fluidized bed boiler. By real-time monitoring of particle status, reinforcement learning control and secondary phase volume fraction feedback, the bed height and air flow velocity of the fluidized bed boiler are dynamically adjusted to ensure uniform combustion of anthracite and lignite and improve combustion efficiency.

[0004] Fluidized bed boilers can effectively control the generation of pollutants due to their large combustion space and low combustion temperature. In actual applications, due to the different characteristics of different coal types (density, volatile matter, ash content and particle size, etc.), the mixed combustion of multiple coal types poses a technical challenge to the combustion control of fluidized bed boilers. Coal types with higher density are usually deposited at the bottom of the fluidized bed, while coal types with lower density are easily carried to the upper part of the fluidized bed by the airflow. This uneven distribution will lead to incomplete combustion and reduced energy utilization. Larger coal particles require higher airflow speeds to be suspended, while smaller coal particles are more likely to float. It is easy to rise to the upper part of the boiler or be discharged with the flue gas flow. The difference in the distribution of combustion particle size will affect the stability of the bed and the combustion reaction rate. High-volatile coal types burn faster and release a lot of heat, while low-volatile coal types burn slower and have a lower calorific value. This difference in combustion rate will cause an uneven temperature field inside the boiler, thus affecting the combustion efficiency. The combustion of multiple coals requires ensuring the uniform distribution and full mixing of different coal types to achieve an overall stable combustion reaction. It is necessary to control the airflow and fuel distribution inside the boiler so that each type of coal can be reasonably suspended, distributed and fully burned with the air.

[0005] To achieve the above object, the present invention provides the following technical solutions:

[0006] A fluidized bed boiler multi-coal mixed combustion control method comprises the following steps:

[0007] Step 1, data acquisition and status monitoring: Based on the intelligent sensor network, the density, volume, suspension state, bed temperature, and airflow velocity of anthracite and lignite in the fluidized bed are monitored in real time. The secondary phase volume fraction equation is used to represent the distribution of different particles in the bed. The dynamic distribution of various types of particles in the bed is obtained through real-time monitoring and sensors to determine the change of the secondary phase volume fraction.

[0008] Step 2: Analyze combustion efficiency and temperature distribution: Analyze combustion efficiency by detecting the spatial distribution of bed temperature, combining temperature gradient and particle volume fraction, and identify overheating and incomplete combustion areas;

[0009] Step 3: Intelligently control airflow velocity and bed height: Use an intelligent algorithm control model based on reinforcement learning to continuously adjust bed height and airflow velocity based on historical data and current combustion status;

[0010] Step 4, feedback adjustment of secondary phase volume fraction: by solving the secondary phase volume fraction equation, the suspension state of different particles and the bed change trend are obtained, so as to adjust the feedback parameters of air flow velocity and bed height. In combination with the different density and combustion speed of particles, the bed height is adjusted dynamically in real time based on the preset adaptive height adjustment formula to ensure uniform combustion of anthracite and lignite.

[0011] Step 5, output and feedback control: Based on the actual operation data, use the preset optimization control function to optimize the control process, and output the optimal values ​​of bed height and airflow velocity through multiple iterations and feedback optimization.

[0012] As a further solution of the present invention, in step 1, the intelligent sensor network includes a sub-node responsible for monitoring the specified physical quantity and a main node responsible for data aggregation, analysis and decision-making. The main node performs anomaly detection on the data of the sub-node based on the Gaussian mixture model. When an anomaly is detected, the main node sends an adaptive adjustment signal based on the data fed back by the sub-node to automatically adjust the working state of the sensors of each node. The communication between the main node and the sub-node is realized using the Internet of Things technology. The frequency of data collection by the sub-node is based on an adaptive filter algorithm to predict the change trend of the bed layer according to the trend of historical data. When a sudden change in the data of the sub-node is detected, the acquisition frequency of the sub-node is automatically increased from the normal acquisition frequency of 100Hz to the strain acquisition frequency of 500Hz.

[0013] As a further solution of the present invention, in step 1, the secondary phase volume fraction equation includes an anthracite secondary phase volume fraction equation and a lignite secondary phase volume fraction equation, and the form thereof is as follows:

[0014] Anthracite secondary phase volume fraction equation: ;

[0015] Where: is the volume fraction of the secondary phase of anthracite particles, is the time variable, is the velocity field of anthracite particles, , is the air flow velocity, is the density of anthracite particles, is the density of the airflow, is the acceleration due to gravity, is the radius of anthracite particles, is the gas dynamic viscosity, is the generation term of anthracite particles, , is the combustion rate constant of anthracite, is the bed temperature, is the ignition temperature of anthracite, is a step function, when Exceed When It starts to take effect, otherwise it is 0. is the deposition term of anthracite particles, , is the height of the anthracite bed;

[0016] The secondary phase volume fraction equation of lignite: ;

[0017] Where: is the volume fraction of the secondary phase in lignite particles, is the velocity field of lignite particles, , is the density of lignite particles, is the radius of the lignite particle, is the generation term of lignite particles, ,in is the combustion rate constant of lignite, is the bed temperature, is the combustion starting temperature of lignite, is the consumption item of lignite particles, ,in is the consumption rate constant of lignite combustion.

[0018] As a further solution of the present invention, the specific execution process of step 2 includes:

[0019] Step 21, spatial interpolation of temperature distribution: using the collected temperature data, adopting the inverse distance weighted method to construct the temperature field in the bed, constructing a three-dimensional temperature distribution map of the bed, and calculating the temperature gradient;

[0020] Step 22, analyzing the combustion efficiency: using the average temperature of the temperature field in the fluidized bed and the combustion reaction rate to calculate the combustion efficiency, obtaining the distribution of anthracite and lignite in the bed through the secondary phase volume fraction equation, and further refining the analysis of the combustion efficiency in combination with the combustion efficiency formula;

[0021] Step 23, identifying overheating and incomplete combustion areas: by calculating the temperature at each location, identify the area that exceeds the preset normal combustion temperature, use temperature gradient analysis to further identify the boundary of the overheating area, identify the area with a temperature lower than the fuel ignition temperature as the incomplete combustion area, and combine the particle volume fraction equation to identify the area where the anthracite volume fraction is higher than the preset value and the temperature is lower than the ignition temperature as the area where the anthracite is not fully burned and has serious deposition.

[0022] As a further solution of the present invention, in step 3, the specific implementation process of intelligently controlling the air flow velocity and the bed height includes:

[0023] Step 31, initializing parameters: setting the initial value of the airflow velocity and the initial value of the bed height based on the historical best operating conditions, setting the learning rate, exploration rate and discount factor of the reinforcement learning, setting the operating range of the airflow velocity and the bed height, setting the action adjustment amplitude of the airflow velocity and the bed height, and setting the triggering condition of the adjustment action;

[0024] Step 32, define a reinforcement learning model: define combustion state parameters in the state space including temperature distribution, air flow velocity, bed height, combustion efficiency, volume fraction of anthracite and lignite particles, and construct a reward function by combining combustion efficiency, temperature gradient, uniformity of particle distribution, and volume penalty of three-dimensional overheating and insufficient combustion areas. At the same time, while optimizing the combustion efficiency in the Q-value update formula, introduce a three-dimensional volume penalty term to reduce the overheating or insufficient combustion areas.

[0025] Step 33, execute intelligent control decision: according to the current state, - The greedy strategy selects the optimal action, adjusts the airflow speed and bed height in real time, waits until the combustion state is balanced after executing the action, collects a new round of data, updates the combustion state, and recalculates all items in the reward function. It uses the latest reward value and the next state to update the Q value, continuously iterates, executes actions, obtains feedback, and updates the Q value. Through multiple rounds of learning, the optimal airflow speed and bed height adjustment strategy is gradually found.

[0026] As a further solution of the present invention, in step 32, the reward function formula involved is:

[0027]

[0028] Where: is the state space at the current time, For real-time combustion efficiency, For ideal combustion efficiency, ,in is the spatial coordinate of the temperature measurement point, is the action space at the current moment, ,in , , , The limit is set according to actual needs. is the reward function value corresponding to the current state space and action space, is the bed temperature distribution in the fluidized bed boiler, is the temperature gradient at different locations in the bed, , , , They are direction, Direction and The unit vector of the direction, is the difference in volume fraction between anthracite and lignite particles, is the volume of the overheated area, is the volume of the incomplete combustion area, , , , They are the weight coefficients of combustion efficiency term, temperature gradient term, particle distribution difference term, overheating and incomplete combustion area penalty term, which are set according to actual needs.

[0029] As a further solution of the present invention, in step 32, the Q value update formula involved is:

[0030]

[0031] Where: For the current state Next action The cumulative expected reward update value that can be obtained, is the old Q value, is the learning rate, For instant rewards, is the discount factor, For the next time step Status Choose the best action The maximum Q value that can be obtained.

[0032] As a further solution of the present invention, in the Q value update formula involved in step 32, Through the reinforcement learning model - Greedy strategy selects actions based on probability Select the action with the largest current Q value as the optimal action. , is the action in the action space, where is the exploration rate, with probability Randomly select actions to try and explore new strategy spaces. ,Based on the initially set exploration rate, the system randomly selects actions and explores different optimization strategies.,As the system gradually learns a better optimization control strategy, the,exploration rate is gradually reduced with a preset step size.

[0033] As a further solution of the present invention, in step 4, the specific implementation process of the feedback adjustment of the secondary phase volume fraction is:

[0034] Step 41, solving the secondary phase volume fraction equation: using the finite difference method to solve the anthracite secondary phase volume fraction equation and the lignite secondary phase volume fraction equation, using the time step Iterative solution to predict the anthracite particle deposition area and lignite particle suspension state at future time points;

[0035] Step 42, analyzing the bed change trend: based on the anthracite particle deposition area and the lignite particle suspension state obtained in step 41, the bed change trend identification analysis is performed. When the anthracite particle deposition area is larger than a preset value, the bed height is reduced and the particle combustion is incomplete. When the number of suspended particles increases beyond a preset value, the airflow velocity is too high or the bed height is inappropriate.

[0036] Step 43, adjusting the feedback control parameters: according to the change trend of the secondary phase volume fraction, when the deposition of anthracite particles increases beyond a preset value, the air flow velocity is increased and the bed height is increased; when the number of suspended lignite particles increases beyond a preset value, the air flow velocity is reduced; when the number of lignite suspended particles is lower than a preset value, the bed height is reduced, and the bed height and air flow velocity are adjusted in real time;

[0037] Step 44, iterative optimization of feedback regulation: adjust the parameters of air flow velocity and bed height based on feedback, dynamically optimize the adjustment range in combination with actual operation data, wait for a preset stable time after each adjustment and then make adjustments, adapt the learning rate and adjustment range, and optimize the control effect through multiple iterations and feedback;

[0038] Step 45, verification of adjustment results: after adjusting the gas velocity and the bed height, the adjustment effect is verified by solving the secondary phase volume fraction equation again, and the combustion efficiency and temperature distribution are monitored in real time.

[0039] Compared with the prior art, the technical effects of the present invention are as follows: the present invention can grasp the suspension state, density, volume distribution, bed temperature and airflow velocity of anthracite and lignite particles in the bed in real time through data collection and state monitoring, ensure the accuracy and timeliness of the data, identify insufficient combustion and overheating areas by analyzing the combustion efficiency and temperature distribution, and provide a basis for subsequent adjustment. The reinforcement learning intelligent control algorithm is introduced to dynamically adjust the bed height and airflow velocity according to historical data and current status, and can adapt to changes in different loads and fuel groups. Through feedback adjustment of the secondary phase volume fraction, the distribution and suspension state of the particles are monitored and solved in real time to ensure dynamic optimization adjustment of the bed height and improve the uniformity and efficiency of combustion. The feedback control mechanism outputs the optimal bed height and airflow velocity values ​​through multiple optimization iterations, thereby achieving more efficient and uniform combustion, reducing the problem of uneven combustion caused by anthracite deposition and excessive combustion of lignite, significantly improving fuel utilization and system efficiency, and can effectively solve the technical problem of real-time adjustment of bed height and airflow velocity when mixed combustion of lignite and anthracite in fluidized bed boilers. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] Figure 1 The figure is a flow chart of the method of the present invention. DETAILED DESCRIPTION

[0041] The following will be combined with the drawings in the present invention to clearly and completely describe the technical solutions in the present invention. Obviously, the described content is only a part of the present invention rather than the whole. Based on the content of the present invention, all other technical solutions obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0042] like Figure 1 As shown, a method for controlling the mixed combustion of multiple coal types in a fluidized bed boiler proposed by the present invention comprises the following steps:

[0043] Step 1, data acquisition and status monitoring: Based on the intelligent sensor network, the density, volume, suspension state, bed temperature, and airflow velocity of anthracite and lignite in the fluidized bed are monitored in real time. The secondary phase volume fraction equation is used to represent the distribution of different particles in the bed. The dynamic distribution of various types of particles in the bed is obtained through real-time monitoring and sensors to determine the change of the secondary phase volume fraction.

[0044] Step 2: Analyze combustion efficiency and temperature distribution: Analyze combustion efficiency by detecting the spatial distribution of bed temperature, combining temperature gradient and particle volume fraction, and identify overheating and incomplete combustion areas;

[0045] Step 3: Intelligently control airflow velocity and bed height: Use an intelligent algorithm control model based on reinforcement learning to continuously adjust bed height and airflow velocity based on historical data and current combustion status;

[0046] Step 4, feedback adjustment of secondary phase volume fraction: by solving the secondary phase volume fraction equation, the suspension state of different particles and the bed change trend are obtained, so as to adjust the feedback parameters of air flow velocity and bed height. In combination with the different density and combustion speed of particles, the bed height is adjusted dynamically in real time based on the preset adaptive height adjustment formula to ensure uniform combustion of anthracite and lignite.

[0047] Step 5, output and feedback control: Based on the actual operation data, use the preset optimization control function to optimize the control process, and output the optimal values ​​of bed height and airflow velocity through multiple iterations and feedback optimization.

[0048] The present invention can grasp the suspension state, density, volume distribution, bed temperature and airflow velocity of anthracite and lignite particles in the bed in real time through data collection and state monitoring, ensure the accuracy and timeliness of data, identify insufficient combustion and overheating areas by analyzing combustion efficiency and temperature distribution, and provide a basis for subsequent adjustment. The reinforcement learning intelligent control algorithm is introduced to dynamically adjust the bed height and airflow velocity according to historical data and current state, and can adapt to changes in different loads and fuel groups. Through feedback adjustment of the secondary phase volume fraction, the distribution and suspension state of particles are monitored and solved in real time to ensure dynamic optimization and adjustment of the bed height, and improve the uniformity and efficiency of combustion. The feedback control mechanism outputs the optimal bed height and airflow velocity values ​​through multiple optimization iterations, thereby achieving more efficient and uniform combustion, reducing the problem of uneven combustion caused by anthracite deposition and excessive combustion of lignite, and significantly improving fuel utilization and system efficiency. It can effectively solve the technical problem of real-time adjustment of bed height and airflow velocity when mixed combustion of lignite and anthracite in fluidized bed boilers.

[0049] It should be noted that in step 1, the intelligent sensor network includes sub-nodes responsible for monitoring designated physical quantities and main nodes responsible for data aggregation, analysis and decision-making. The main node performs anomaly detection on the data of the sub-nodes based on the Gaussian mixture model. When an anomaly is detected, the main node sends an adaptive adjustment signal based on the data fed back by the sub-nodes to automatically adjust the working state of the sensors at each node. The communication between the main node and the sub-nodes is achieved using the Internet of Things technology. The frequency of data collection by the sub-nodes is based on the adaptive filter algorithm to predict the change trend of the bed layer according to the trend of historical data. When a sudden change in the data of the sub-node is detected, the acquisition frequency of the sub-node is automatically increased from the normal acquisition frequency of 100 Hz to the strain acquisition frequency of 500 Hz.

[0050] The data of the sub-nodes are pre-processed by multi-sensor data fusion technology to reduce the impact of single sensor failure or data deviation on the overall performance of the system and ensure data stability and consistency. The Gaussian mixture model is not only used for anomaly detection in the main node, but also for dynamically adjusting the threshold according to the statistical characteristics of different monitored physical quantities of each sub-node to ensure more accurate detection of abnormal data. The adaptive filter algorithm of the sub-node is not only based on historical data trends, but also combines real-time feedback signals to adjust the weight of the prediction model to improve prediction accuracy and reduce oversampling and delay. The system is equipped with a sub-node fault detection mechanism. When a hardware or communication failure occurs in a sensor node, the system automatically starts redundant sensors or adjusts the acquisition frequency of adjacent nodes to supplement the data loss. Based on the anomaly detection results, the main node not only adjusts the working status of the sub-node sensor, but also optimizes the overall operation strategy of the system through decision logic based on machine learning algorithms, such as adjusting the bed height and airflow velocity to improve combustion efficiency. The main node and sub-nodes are coordinated through a high-precision time synchronization mechanism to ensure the timeliness and consistency of data acquisition and feedback, and avoid data delays or failures caused by asynchrony. The IoT protocols used for communication adopt low-latency, high-bandwidth standards such as LoRa or NB-IoT to ensure data transmission efficiency and stability between sub-nodes and the main node in large-scale monitoring scenarios.

[0051] It should be noted that in step 1, the secondary phase volume fraction equation includes the anthracite secondary phase volume fraction equation and the lignite secondary phase volume fraction equation, and the form thereof is as follows:

[0052] Anthracite secondary phase volume fraction equation: ;

[0053] Where: is the volume fraction of the secondary phase of anthracite particles, is the time variable, is the velocity field of anthracite particles, , is the air flow velocity, is the density of anthracite particles, is the density of the airflow, is the acceleration due to gravity, is the radius of anthracite particles, is the gas dynamic viscosity, is the generation term of anthracite particles, , is the combustion rate constant of anthracite, is the bed temperature, is the ignition temperature of anthracite, is a step function, when Exceed When It starts to take effect, otherwise it is 0. is the deposition term of anthracite particles, , is the height of the anthracite bed;

[0054] The secondary phase volume fraction equation of lignite: ;

[0055] Where: is the volume fraction of the secondary phase in lignite particles, is the velocity field of lignite particles, , is the density of lignite particles, is the radius of the lignite particle, is the generation term of lignite particles, ,in is the combustion rate constant of lignite, is the bed temperature, is the combustion starting temperature of lignite, is the consumption item of lignite particles, ,in is the consumption rate constant of lignite combustion.

[0056] The problem of uneven combustion in fluidized bed boilers was solved by constructing secondary phase volume fraction equations for anthracite and lignite respectively. The anthracite equation and lignite equation take into account their own generation terms and deposition / consumption terms respectively, and can accurately describe the dynamic distribution and combustion process of different coal particles in the bed. The velocity field expression in the anthracite equation combines factors such as particle density, air flow velocity and bed temperature, so that the combustion rate can be dynamically adjusted according to actual conditions, ensuring that the generation terms are reached after the anthracite ignition temperature is reached. The lignite equation uses its consumption term The combustion process and consumption of lignite are accurately described by the combustion rate constant. The combustion state of anthracite and lignite in the fluidized bed is monitored and adjusted by the secondary phase volume fraction equation, making the temperature and particle distribution in each area of ​​the bed more uniform, reducing the uneven combustion caused by anthracite deposition and over-rapid combustion of lignite, and significantly improving the combustion efficiency and system stability.

[0057] It should be noted that the specific execution process of step 2 includes:

[0058] Step 21, spatial interpolation of temperature distribution: using the collected temperature data, adopting the inverse distance weighted method to construct the temperature field in the bed, constructing a three-dimensional temperature distribution map of the bed, and calculating the temperature gradient;

[0059] Step 22, analyzing the combustion efficiency: using the average temperature of the temperature field in the fluidized bed and the combustion reaction rate to calculate the combustion efficiency, obtaining the distribution of anthracite and lignite in the bed through the secondary phase volume fraction equation, and further refining the analysis of the combustion efficiency in combination with the combustion efficiency formula;

[0060] Step 23, identifying overheating and incomplete combustion areas: by calculating the temperature at each location, identify the area that exceeds the preset normal combustion temperature, use temperature gradient analysis to further identify the boundary of the overheating area, identify the area with a temperature lower than the fuel ignition temperature as the incomplete combustion area, and combine the particle volume fraction equation to identify the area where the anthracite volume fraction is higher than the preset value and the temperature is lower than the ignition temperature as the area where the anthracite is not fully burned and has serious deposition.

[0061] The collected temperature data is spatially interpolated by the inverse distance weighted method to construct a three-dimensional temperature distribution map in the bed, making the construction of the temperature field more accurate and clearly showing the temperature gradient changes. This provides a basis for identifying overheated and incompletely burned areas. Based on the average temperature and combustion reaction rate of the temperature field, combined with the secondary phase volume fraction equation of anthracite and lignite, the combustion efficiency of different areas is accurately calculated. This process makes the analysis of combustion efficiency more detailed, so that the combustion process can be dynamically adjusted to improve fuel utilization. By calculating the temperature gradient, the overheated areas in the bed that exceed the normal combustion temperature and the incompletely burned areas that do not reach the fuel ignition temperature can be accurately identified. At the same time, combined with the secondary phase volume fraction equation, the areas where anthracite is not fully burned and has severe deposition are further accurately identified.

[0062] It should be noted that in step 3, the specific implementation process of intelligently controlling the air flow velocity and bed height includes:

[0063] Step 31, initializing parameters: setting the initial value of the airflow velocity and the initial value of the bed height based on the historical best operating conditions, setting the learning rate, exploration rate and discount factor of the reinforcement learning, setting the operating range of the airflow velocity and the bed height, setting the action adjustment amplitude of the airflow velocity and the bed height, and setting the triggering condition of the adjustment action;

[0064] Step 32, define a reinforcement learning model: define combustion state parameters in the state space including temperature distribution, air flow velocity, bed height, combustion efficiency, volume fraction of anthracite and lignite particles, and construct a reward function by combining combustion efficiency, temperature gradient, uniformity of particle distribution, and volume penalty of three-dimensional overheating and insufficient combustion areas. At the same time, while optimizing the combustion efficiency in the Q-value update formula, introduce a three-dimensional volume penalty term to reduce the overheating or insufficient combustion areas.

[0065] Step 33, execute intelligent control decision: according to the current state, - The greedy strategy selects the optimal action, adjusts the airflow speed and bed height in real time, waits until the combustion state is balanced after executing the action, collects a new round of data, updates the combustion state, and recalculates all items in the reward function. It uses the latest reward value and the next state to update the Q value, continuously iterates, executes actions, obtains feedback, and updates the Q value. Through multiple rounds of learning, the optimal airflow speed and bed height adjustment strategy is gradually found.

[0066] Based on real-time monitoring and reinforcement learning algorithms, the airflow velocity and bed height can be dynamically adjusted to ensure the optimal suspension state of different types of coal during the combustion process and improve combustion uniformity. The parameters are initialized using the historical best operating conditions to accelerate the learning process. Through continuous iteration and Q value updates, the optimal control strategy is gradually found after multiple rounds of learning, reducing the complexity of manual debugging. The reward function not only optimizes the combustion efficiency, but also introduces a three-dimensional volume penalty term to reduce overheating and insufficient combustion areas, thereby improving the stability of the overall system and fuel utilization.

[0067] It should be noted that in step 32, the reward function formula involved is:

[0068]

[0069] Where: is the state space at the current time, For real-time combustion efficiency, For ideal combustion efficiency, ,in is the spatial coordinate of the temperature measurement point, is the action space at the current moment, ,in , , , The limit is set according to actual needs. is the reward function value corresponding to the current state space and action space, is the bed temperature distribution in the fluidized bed boiler, is the temperature gradient at different locations in the bed, , , , They are direction, Direction and The unit vector of the direction, is the difference in volume fraction between anthracite and lignite particles, is the volume of the overheated area, is the volume of the incomplete combustion area, , , , They are the weight coefficients of combustion efficiency term, temperature gradient term, particle distribution difference term, overheating and incomplete combustion area penalty term, which are set according to actual needs.

[0070] The reward function formula dynamically optimizes the airflow velocity and bed height of the fluidized bed boiler by introducing combustion efficiency, temperature gradient, particle distribution difference, and volume penalty for overheating and incomplete combustion areas. The formula comprehensively considers combustion efficiency, temperature uniformity, coal particle distribution difference, and overheating / incomplete combustion areas, ensuring the comprehensive optimization of the combustion process and avoiding deviations caused by a single optimization target. The reinforcement learning model continuously adjusts the airflow velocity and bed height according to the real-time combustion state, dynamically optimizes the combustion process, and improves the system adaptability. The penalty terms of temperature gradient and volume fraction difference reduce local overheating and incomplete combustion, making the combustion process more uniform and improving the overall combustion efficiency.

[0071] It should be noted that in step 32, the Q value update formula involved is:

[0072]

[0073] Where: For the current state Next action The cumulative expected reward update value that can be obtained, is the old Q value, is the learning rate, For instant rewards, is the discount factor, For the next time step Status Choose the best action The maximum Q value that can be obtained.

[0074] The Q-value update formula realizes dynamic learning and optimization of the fluidized bed boiler combustion control process by introducing immediate rewards and the maximum Q-value of the next state. The immediate reward term ensures that the system immediately feedbacks the combustion efficiency, temperature uniformity and other states after each step of action, and the maximum reward that may be obtained in the future is introduced into the Q-value update through the discount factor, which enhances the system's focus on long-term benefits and avoids short-sighted decisions. At the same time, the learning rate λ controls the speed of Q-value update, so that the system gradually tends to the optimal control strategy. Compared with the existing technology, this method can adaptively optimize the airflow velocity and bed height, and through continuous iterative Q-value updates, it can achieve efficient and stable combustion control under complex combustion conditions, and improve the system's self-learning ability and control accuracy.

[0075] It should be noted that in the Q value update formula involved in step 32, Through the reinforcement learning model - Greedy strategy selects actions based on probability Select the action with the largest current Q value as the optimal action. , is the action in the action space, where is the exploration rate, with probability Randomly select actions to try and explore new strategy spaces. ,Based on the initially set exploration rate, the system randomly selects actions and explores different optimization strategies.,As the system gradually learns a better optimization control strategy, the,exploration rate is gradually reduced with a preset step size.

[0076] By using - Greedy strategy, which achieves a balance between exploration and utilization in the reinforcement learning process, with probability Select the action with the largest current Q value to ensure that the system optimizes the airflow velocity and bed height under the learned optimal control strategy, thereby improving the combustion efficiency. Randomly selecting actions allows the system to explore new strategy spaces and avoid falling into local optimality. As the system gradually learns better control strategies, the exploration rate gradually decreases, making the system more dependent on the optimal strategy for decision-making. This gradual optimization of strategies is more flexible than existing technologies and can adapt to complex combustion environments and dynamically changing operating conditions, thereby achieving more precise control and improving the system's self-learning ability and overall operating efficiency.

[0077] It should be noted that in step 4, the specific implementation process of the feedback adjustment of the secondary phase volume fraction is:

[0078] Step 41, solving the secondary phase volume fraction equation: using the finite difference method to solve the anthracite secondary phase volume fraction equation and the lignite secondary phase volume fraction equation, using the time step Iterative solution to predict the anthracite particle deposition area and lignite particle suspension state at future time points;

[0079] Step 42, analyzing the bed change trend: based on the anthracite particle deposition area and the lignite particle suspension state obtained in step 41, the bed change trend identification analysis is performed. When the anthracite particle deposition area is larger than a preset value, the bed height is reduced and the particle combustion is incomplete. When the number of suspended particles increases beyond a preset value, the airflow velocity is too high or the bed height is inappropriate.

[0080] Step 43, adjusting the feedback control parameters: according to the change trend of the secondary phase volume fraction, when the deposition of anthracite particles increases beyond a preset value, the air flow velocity is increased and the bed height is increased; when the number of suspended lignite particles increases beyond a preset value, the air flow velocity is reduced; when the number of lignite suspended particles is lower than a preset value, the bed height is reduced, and the bed height and air flow velocity are adjusted in real time;

[0081] Step 44, iterative optimization of feedback regulation: adjust the parameters of air flow velocity and bed height based on feedback, dynamically optimize the adjustment range in combination with actual operation data, wait for a preset stable time after each adjustment and then make adjustments, adapt the learning rate and adjustment range, and optimize the control effect through multiple iterations and feedback;

[0082] Step 45, verification of adjustment results: after adjusting the gas velocity and the bed height, the adjustment effect is verified by solving the secondary phase volume fraction equation again, and the combustion efficiency and temperature distribution are monitored in real time.

[0083] The secondary phase volume fraction feedback regulation scheme solves the secondary phase volume fraction equation by finite difference method, accurately predicts the deposition area of ​​anthracite particles and the suspension state of lignite particles at future time points, and provides a scientific basis for real-time regulation of fluidized bed boilers. When analyzing the bed change trend, it can accurately identify the problem of excessive deposition of anthracite particles or unstable lignite suspension, and dynamically optimize the air flow velocity and bed height by adjusting the feedback control parameters to ensure a more uniform and efficient combustion process. Through iterative optimization of feedback regulation, the system can adaptively learn the optimal control strategy under different load conditions, and verify the regulation effect by solving the secondary phase volume fraction equation again after adjustment, ensuring that the adjustment measures are effective and gradually improving the control accuracy. Compared with the existing technology, this method significantly improves the combustion uniformity and efficiency of fluidized bed boilers, reduces the phenomenon of anthracite deposition and insufficient lignite combustion, has higher control flexibility and adaptability, and makes the system more stable and efficient under complex working conditions.

[0084] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art who is familiar with the present technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

[0085] Finally: The above description is only a specific implementation scheme 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 protection scope of the present invention.

Claims

1. A fluidized bed boiler multi-coal mixed combustion control method, characterized in that: The steps include: Step 1, data acquisition and status monitoring: Based on the intelligent sensor network, the density, volume, suspension state, bed temperature, and airflow velocity of anthracite and lignite in the fluidized bed are monitored in real time. The secondary phase volume fraction equation is used to represent the distribution of different particles in the bed. The dynamic distribution of various types of particles in the bed is obtained through real-time monitoring and sensors to determine the change of the secondary phase volume fraction. Step 2: Analyze combustion efficiency and temperature distribution: Analyze combustion efficiency by detecting the spatial distribution of bed temperature, combining temperature gradient and particle volume fraction, and identify overheating and incomplete combustion areas; Step 3: Intelligently control airflow velocity and bed height: Use an intelligent algorithm control model based on reinforcement learning to continuously adjust bed height and airflow velocity based on historical data and current combustion status; Step 4, feedback adjustment of secondary phase volume fraction: by solving the secondary phase volume fraction equation, the suspension state of different particles and the bed change trend are obtained, so as to adjust the feedback parameters of air flow velocity and bed height. In combination with the different density and combustion speed of particles, the bed height is adjusted dynamically in real time based on the preset adaptive height adjustment formula to ensure uniform combustion of anthracite and lignite. Step 5, output and feedback control: Based on the actual operation data, use the preset optimization control function to optimize the control process, and output the optimal values ​​of bed height and airflow velocity through multiple iterations and feedback optimization.

2. A fluidized bed boiler multi-coal mixed combustion control method according to claim 1, characterized in that: In step 1, the intelligent sensor network includes sub-nodes responsible for monitoring designated physical quantities and main nodes responsible for data aggregation, analysis and decision-making. The main node detects anomalies in the data of the sub-nodes based on the Gaussian mixture model. When an anomaly is detected, the main node sends an adaptive adjustment signal based on the data fed back by the sub-nodes to automatically adjust the working status of the sensors at each node. The communication between the main node and the sub-nodes is achieved using the Internet of Things technology. The frequency of data collection by the sub-nodes is based on the adaptive filter algorithm to predict the change trend of the bed layer according to the trend of historical data. When a sudden change in the sub-node data is detected, the sub-node collection frequency is automatically increased from the normal collection frequency of 100 Hz to the strain collection frequency of 500 Hz.

3. The method for controlling the mixed combustion of multiple coal types in a fluidized bed boiler according to claim 2, characterized in that: In step 1, the secondary phase volume fraction equation includes the anthracite secondary phase volume fraction equation and the lignite secondary phase volume fraction equation, which are in the following form: Anthracite secondary phase volume fraction equation: ; Where: is the volume fraction of the secondary phase of anthracite particles, is the time variable, is the velocity field of anthracite particles, , is the air flow velocity, is the density of anthracite particles, is the density of the airflow, is the acceleration due to gravity, is the radius of anthracite particles, is the gas dynamic viscosity, is the generation term of anthracite particles, , is the combustion rate constant of anthracite, is the bed temperature, is the ignition temperature of anthracite, is a step function, when Exceed When It starts to take effect, otherwise it is 0. is the deposition term of anthracite particles, , is the height of the anthracite bed; The secondary phase volume fraction equation of lignite: ; Where: is the volume fraction of the secondary phase in lignite particles, is the velocity field of lignite particles, , is the density of lignite particles, is the radius of the lignite particle, is the generation term of lignite particles, ,in is the combustion rate constant of lignite, is the bed temperature, is the combustion starting temperature of lignite, is the consumption item of lignite particles, ,in is the consumption rate constant of lignite combustion.

4. The method for controlling the mixed combustion of multiple coal types in a fluidized bed boiler according to claim 1, characterized in that: The specific implementation process of step 2 includes: Step 21, spatial interpolation of temperature distribution: using the collected temperature data, adopting the inverse distance weighted method to construct the temperature field in the bed, constructing a three-dimensional temperature distribution map of the bed, and calculating the temperature gradient; Step 22, analyzing the combustion efficiency: using the average temperature of the temperature field in the fluidized bed and the combustion reaction rate to calculate the combustion efficiency, obtaining the distribution of anthracite and lignite in the bed through the secondary phase volume fraction equation, and further refining the analysis of the combustion efficiency in combination with the combustion efficiency formula; Step 23, identifying overheating and incomplete combustion areas: by calculating the temperature at each location, identify the area that exceeds the preset normal combustion temperature, use temperature gradient analysis to further identify the boundary of the overheating area, identify the area with a temperature lower than the fuel ignition temperature as the incomplete combustion area, and combine the particle volume fraction equation to identify the area where the anthracite volume fraction is higher than the preset value and the temperature is lower than the ignition temperature as the area where the anthracite is not fully burned and has serious deposition.

5. The method for controlling the mixed combustion of multiple coal types in a fluidized bed boiler according to claim 4, characterized in that: In step 3, the specific implementation process of intelligently controlling the air flow velocity and bed height includes: Step 31, initializing parameters: setting the initial value of the airflow velocity and the initial value of the bed height based on the historical best operating conditions, setting the learning rate, exploration rate and discount factor of the reinforcement learning, setting the operating range of the airflow velocity and the bed height, setting the action adjustment amplitude of the airflow velocity and the bed height, and setting the triggering condition of the adjustment action; Step 32, define a reinforcement learning model: define combustion state parameters in the state space including temperature distribution, air flow velocity, bed height, combustion efficiency, volume fraction of anthracite and lignite particles, and construct a reward function by combining combustion efficiency, temperature gradient, uniformity of particle distribution, and volume penalty of three-dimensional overheating and insufficient combustion areas. At the same time, while optimizing the combustion efficiency in the Q-value update formula, introduce a three-dimensional volume penalty term to reduce the overheating or insufficient combustion areas. Step 33, execute intelligent control decision: according to the current state, - The greedy strategy selects the optimal action, adjusts the airflow speed and bed height in real time, waits until the combustion state is balanced after executing the action, collects a new round of data, updates the combustion state, and recalculates all items in the reward function. It uses the latest reward value and the next state to update the Q value, continuously iterates, executes actions, obtains feedback, and updates the Q value. Through multiple rounds of learning, the optimal airflow speed and bed height adjustment strategy is gradually found.

6. A fluidized bed boiler multi-coal mixed combustion control method according to claim 5, characterized in that: In step 32, the reward function formula involved is: Where: is the state space at the current time, For real-time combustion efficiency, For ideal combustion efficiency, ,in is the spatial coordinate of the temperature measurement point, is the action space at the current moment, ,in , , , The limit is set according to actual needs. is the reward function value corresponding to the current state space and action space, is the bed temperature distribution in the fluidized bed boiler, is the temperature gradient at different locations in the bed, , , , They are direction, Direction and The unit vector of the direction, is the difference in volume fraction between anthracite and lignite particles, is the volume of the overheated area, is the volume of the incomplete combustion area, , , , They are the weight coefficients of the combustion efficiency term, temperature gradient term, particle distribution difference term, overheating and incomplete combustion area penalty term, which are set according to actual needs.

7. The method for controlling the mixed combustion of multiple coal types in a fluidized bed boiler according to claim 5, characterized in that: In step 32, the Q value update formula involved is: Where: For the current state Next action The cumulative expected reward update value that can be obtained, is the old Q value, is the learning rate, For instant rewards, is the discount factor, For the next time step Status Choose the best action The maximum Q value that can be obtained.

8. The method for controlling the mixed combustion of multiple coal types in a fluidized bed boiler according to claim 7, characterized in that: In the Q value update formula involved in step 32, Through the reinforcement learning model - Greedy strategy selects actions based on probability Select the action with the largest current Q value as the optimal action. , is the action in the action space, where is the exploration rate, with probability Randomly select actions to try and explore new strategy spaces. ,Based on the initially set exploration rate, the system randomly selects actions and explores different optimization strategies.,As the system gradually learns a better optimization control strategy, the,exploration rate is gradually reduced with a preset step size.

9. A fluidized bed boiler multi-coal mixed combustion control method according to claim 8, characterized in that: In step 4, the specific implementation process of the feedback adjustment of the secondary phase volume fraction is: Step 41, solving the secondary phase volume fraction equation: using the finite difference method to solve the anthracite secondary phase volume fraction equation and the lignite secondary phase volume fraction equation, using the time step Iterative solution to predict the anthracite particle deposition area and lignite particle suspension state at future time points; Step 42, analyzing the bed change trend: based on the anthracite particle deposition area and the lignite particle suspension state obtained in step 41, the bed change trend identification analysis is performed. When the anthracite particle deposition area is larger than a preset value, the bed height is reduced and the particle combustion is incomplete. When the number of suspended particles increases beyond a preset value, the airflow velocity is too high or the bed height is inappropriate. Step 43, adjusting the feedback control parameters: according to the change trend of the secondary phase volume fraction, when the deposition of anthracite particles increases beyond a preset value, the air flow velocity is increased and the bed height is increased; when the number of suspended lignite particles increases beyond a preset value, the air flow velocity is reduced; when the number of lignite suspended particles is lower than a preset value, the bed height is reduced, and the bed height and air flow velocity are adjusted in real time; Step 44, iterative optimization of feedback regulation: adjust the parameters of air flow velocity and bed height based on feedback, dynamically optimize the adjustment range in combination with actual operation data, wait for a preset stable time after each adjustment and then make adjustments, adapt the learning rate and adjustment range, and optimize the control effect through multiple iterations and feedback; Step 45, verification of adjustment results: after adjusting the gas velocity and the bed height, the adjustment effect is verified by solving the secondary phase volume fraction equation again, and the combustion efficiency and temperature distribution are monitored in real time.

Images