A method and system for controlling SO2 emission of a circulating fluidized bed unit

Abstract

This disclosure provides a method and system for controlling SO2 emissions from a circulating fluidized bed unit. By constructing a dynamic SO2 emission prediction model that includes economic objectives and adopting a hierarchical collaborative optimization control architecture, the method first uses an optimization algorithm to calculate the economically optimal SO2 concentration setpoint in real time. Then, it uses a dual-loop generalized predictive controller (GPC) for precise tracking control. At the same time, a fuzzy inference system is introduced to dynamically and adaptively adjust the weight of the control objective. This achieves stable, economical, and environmentally friendly synergistic optimization control of SO2 emissions under complex operating conditions such as deep peak shaving. It effectively solves the problems of slow response, unstable control, and high cost of traditional methods, and achieves the beneficial effects of significantly reducing desulfurization operating costs, enhancing emission concentration stability, and improving the system's adaptive capability.

Description

Technical Field

[0001] This invention relates to the field of pollutant control technology in coal-fired power generation, and in particular to a method and system for controlling SO2 emissions from a circulating fluidized bed unit. Background Technology

[0002] CFB units frequently need to participate in deep peak shaving, and drastic load changes cause fluctuations in the furnace environment, resulting in unsteady and violent oscillations in SO2 emission concentrations. This poses a severe challenge to the stability and economy of the environmental control system. Traditional control strategies (such as PID control and fixed-ratio feeding) are difficult to cope with such complex operating conditions, and have problems such as response lag, excessive addition of desulfurizing agents, high operating costs, and poor emission stability.

[0003] While generalized predictive control (GPC) has a certain degree of adaptability, its fixed setpoint and weighting coefficient strategy still falls short when facing multi-objective (economic, stability, and environmental) collaborative optimization problems. Therefore, it is crucial to develop an GPC control strategy that can adapt to changes in operating conditions and balance economic efficiency and stability. Summary of the Invention

[0004] A first aspect of this disclosure provides a method for controlling SO2 emissions from a circulating fluidized bed unit, comprising the following steps: S1: Construct a dynamic SO2 emission prediction model for circulating fluidized bed units. The model includes an SO2 generation sub-model, an in-furnace desulfurization sub-model, an out-of-furnace desulfurization sub-model, and a desulfurization economic cost sub-model. S2: Based on the SO2 dynamic emission prediction model and real-time operating data, the optimization algorithm is used to calculate the optimal setpoints for SO2 concentration in the furnace and SO2 concentration outside the furnace that minimize the total desulfurization cost in the future period during the current control cycle k. S3: Input the optimal setpoint of SO2 concentration in the furnace to the furnace generalized predictive controller, and input the optimal setpoint of SO2 concentration outside the furnace to the outside generalized predictive controller; S4: The in-furnace generalized predictive controller calculates the control increment of the limestone feed rate in the furnace based on the optimal setpoint of the in-furnace desulfurization SO2 concentration and the predicted value of the in-furnace SO2 concentration output by the SO2 dynamic emission prediction model or the actual measured value of the in-furnace SO2 concentration. S5: The external generalized predictive controller calculates the control increment of the external limestone slurry addition amount based on the optimal set value of the external desulfurization SO2 concentration and the predicted value of the external SO2 concentration output by the SO2 dynamic emission prediction model or the actual measured external SO2 concentration value. S6: Based on the real-time operating conditions, dynamically adjust the weight coefficients of various performance indicators in the objective function of the in-furnace and out-of-furnace generalized predictive controllers through the fuzzy inference system; S7: Apply the control increment of limestone feed rate inside and outside the furnace to the controlled object or the SO2 dynamic emission prediction model, and return to step S2 in the next sampling cycle to achieve rolling optimization control.

[0005] In conjunction with the first aspect, the optimization algorithm mentioned in step S2 is either particle swarm optimization algorithm or genetic algorithm.

[0006] In conjunction with the first aspect, the optimal setpoints for SO2 concentration in the furnace and the external desulfurization process in step S2 are calculated using the following formulas: , , in, This is the optimal setpoint for SO2 concentration in the furnace desulfurization process. This is the optimal setpoint for SO2 concentration in the external desulfurization process. The amount of limestone fed into the furnace. This refers to the amount of limestone supplied outside the furnace. , These represent the desulfurization efficiencies inside and outside the furnace, respectively. , These are the correction factors for the calcium-sulfur ratio inside and outside the furnace, respectively. For flue gas flow rate, , These represent the sulfur content inside and outside the furnace, respectively.

[0007] In conjunction with the first aspect, the input variables of the fuzzy inference system in step S6 include one or more of the following: SO2 concentration change rate, load change rate, and limestone usage change rate. The output variables are the weighting coefficients of tracking error, economic indicators, and control quantity stability in the objective function.

[0008] In conjunction with the first aspect, the fuzzy inference system uses the centroid method for defuzzification, and its fuzzy rules dynamically adjust the weight allocation based on the operating conditions.

[0009] In conjunction with the first aspect, the generalized predictive controller uses the CARIMA model as the prediction model, and its objective function is as follows: , in, For the first The predicted output of the step, To predict the time domain, To control the time domain, To control the weighting coefficients, For the first The incremental control of limestone feed rate in each step.

[0010] A second aspect of this disclosure provides a control system for SO2 emissions from a circulating fluidized bed unit, comprising: The data communication module is used to acquire real-time operating data of the circulating fluidized bed unit; The SO2 dynamic emission prediction model module is used to predict the dynamic changes in SO2 concentration based on the operational data. The optimized setpoint calculation module is used to run optimization algorithms to calculate the optimal SO2 concentration setpoints for desulfurization inside and outside the furnace; The generalized predictive control module includes an in-furnace generalized predictive control submodule and an out-of-furnace generalized predictive control submodule, which are respectively used to receive the optimal setpoint and generate control commands; The fuzzy adaptive weight adjustment module is used to dynamically adjust the objective function weights of the generalized prediction control module according to the real-time operating status. Each of these modules is integrated into a simulation computing platform or an industrial control system.

[0011] In conjunction with the second aspect, the optimization setpoint calculation module is configured to execute a particle swarm optimization algorithm or a genetic algorithm, and the input of the fuzzy adaptive weight adjustment module includes one or more of the following: SO2 concentration deviation, load change rate, and limestone dosage change rate.

[0012] A third aspect of this disclosure provides an electronic device comprising: One or more processors; A storage unit is used to store one or more programs that, when executed by one or more processors, enable the one or more processors to implement the SO2 emission control method of the circulating fluidized bed unit.

[0013] A fourth aspect of this disclosure provides a computer-readable storage medium having a computer program stored thereon, characterized in that the computer program, when executed by a processor, enables a method for controlling SO2 emissions from the circulating fluidized bed unit.

[0014] Beneficial Effects: The SO2 emission control method and system provided in this disclosure for a circulating fluidized bed unit constructs a dynamic SO2 emission prediction model that includes economic objectives and adopts a hierarchical collaborative optimization control architecture. First, the optimal SO2 concentration setpoint is calculated in real time using an optimization algorithm. Then, a dual-loop generalized predictive controller (GPC) is used for precise tracking control. At the same time, a fuzzy inference system is introduced to dynamically and adaptively adjust the weight of the control objective. This achieves stable, economical, and environmentally friendly synergistic optimization control of SO2 emissions under complex operating conditions such as deep peak shaving. It effectively solves the problems of slow response, unstable control, and high cost of traditional methods, and achieves the beneficial effects of significantly reducing desulfurization operating costs, enhancing emission concentration stability, and improving the system's adaptive capability. Attached Figure Description

[0015] Figure 1 This is a schematic flowchart illustrating a method for controlling SO2 emissions from a circulating fluidized bed unit according to an embodiment of this disclosure. Figure 2 This is a schematic diagram of the structure of a control system for SO2 emissions from a circulating fluidized bed unit according to an embodiment of this disclosure; Figure 3 An electronic device according to an embodiment of this disclosure. Detailed Implementation

[0016] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with those disclosed herein.

[0017] The terminology used in this disclosure is for the purpose of describing particular embodiments only and is not intended to be limiting of the present disclosure. The singular forms “a,” “the,” and “the” as used in this disclosure and the appended claims are also intended to include the plural forms unless the context clearly indicates otherwise. It should also be understood that the term “and / or” as used herein refers to and includes any and all possible combinations of one or more of the associated listed items.

[0018] Figure 1 This is a schematic flowchart illustrating a method for controlling SO2 emissions from a circulating fluidized bed unit according to an embodiment of the present disclosure, including: S1. Construct a dynamic SO2 emission prediction model for circulating fluidized bed units. The model includes an SO2 generation sub-model, an in-furnace desulfurization sub-model, an out-of-furnace desulfurization sub-model, and a desulfurization economic cost sub-model.

[0019] First, it is necessary to extensively collect historical and real-time operating parameters related to SO2 generation and removal from real-time databases such as Distributed Control Systems (DCS) and System-on-Systems (SIS). These key data include at least: unit load (MW), main steam flow rate, total boiler air volume, coal feed rate and coal sulfur content analysis, primary / secondary air flow rate and temperature, bed temperature, furnace outlet oxygen content, flue gas flow rate, measured values ​​of raw flue gas SO2 concentration, measured values ​​of clean flue gas SO2 concentration, limestone feed rate in the furnace, and limestone slurry dosage outside the furnace. The collected data needs to undergo preprocessing such as filtering, outlier removal, and normalization to ensure data quality.

[0020] Subsequently, based on the above data and the understanding of the desulfurization process mechanism, four sub-models were constructed. The SO2 generation sub-model mainly establishes the dynamic relationship between the coal feed rate, the sulfur content in the coal, and the initial SO2 concentration generated after combustion in the furnace. It can usually be represented as a time lag or transfer function model with load and sulfur content as the main inputs. The in-furnace desulfurization sub-model describes the influence of factors such as the limestone (CaCO3) feed rate, bed temperature, and calcium-sulfur ratio on the initial desulfurization efficiency in the furnace. Its core is to calculate the amount of SO2 removed in the furnace, which is usually a dynamic process model with pure time lag. The out-of-furnace desulfurization sub-model (such as dry or semi-dry desulfurization) establishes the dynamic relationship between the amount of limestone slurry added outside the furnace and the final SO2 removal efficiency, considering the influence of factors such as slurry concentration, spray effect, and flue gas temperature.

[0021] The desulfurization economic cost sub-model is key to embodying the "economic" optimization objective of this invention. This model quantifies the operating cost of the desulfurization process, and its calculation mainly includes two parts: first, the cost of limestone consumption, calculated in real-time based on the unit price of limestone and the feed rates inside and outside the furnace; second, the plant power cost of the desulfurization system, including the power consumption costs of equipment such as the limestone crusher, slurry pump, and spray system. The output of this model is the total desulfurization cost per unit time.

[0022] Finally, the four sub-models mentioned above are integrated and coupled according to the actual process flow to form a unified SO2 dynamic emission prediction model. This model can receive current control commands (such as limestone feed rate setpoint) and operating conditions (such as load and coal quality), and predict the key outputs of the system over a period of time, including the changing trajectories of SO2 concentrations in raw and clean flue gas, as well as the corresponding desulfurization economic costs. The model is built in a simulation platform (such as MATLAB / Simulink) and uses historical field data for parameter identification and model verification to ensure that its prediction accuracy meets control requirements, providing a reliable predictive basis for subsequent optimization setpoint calculations and predictive control.

[0023] S2. Based on the SO2 dynamic emission prediction model and real-time operating data in the current control cycle, use an optimization algorithm to calculate the optimal setpoints for SO2 concentration in the furnace and SO2 concentration outside the furnace that minimize the total desulfurization cost in the future period.

[0024] The optimization algorithm mentioned in step S2 is either particle swarm optimization algorithm or genetic algorithm.

[0025] The optimal setpoints for SO2 concentration in in-furnace desulfurization and SO2 concentration in external desulfurization mentioned in step S2 are calculated using the following formulas: , , in, This is the optimal setpoint for SO2 concentration in the furnace desulfurization process. This is the optimal setpoint for SO2 concentration in the external desulfurization process. The amount of limestone fed into the furnace. This refers to the amount of limestone supplied outside the furnace. , These represent the desulfurization efficiencies inside and outside the furnace, respectively. , These are the correction factors for the calcium-sulfur ratio inside and outside the furnace, respectively. For flue gas flow rate, , These represent the sulfur content inside and outside the furnace, respectively.

[0026] At the start of the current control cycle k, the system acquires the latest real-time operating data through the data communication module, including the current unit load, flue gas flow rate, raw flue gas SO2 concentration, and current limestone dosage. Subsequently, the optimization algorithm (taking the particle swarm optimization algorithm PSO as an example) is activated.

[0027] Optimization algorithms (such as Particle Swarm Optimization, PSO) use the setpoints for in-furnace desulfurization SO2 concentration and external desulfurization SO2 concentration in future time periods as decision variables for optimization. To accelerate convergence and ensure the feasibility of the solution, the algorithm uses a simplified computational model based on the process mechanism to initialize the population or define an efficient search space. This simplified model is characterized by the following formula: , , in, This is the optimal setpoint for SO2 concentration in the furnace desulfurization process. This is the optimal setpoint for SO2 concentration in the external desulfurization process. The amount of limestone fed into the furnace. This refers to the amount of limestone supplied outside the furnace. , These represent the desulfurization efficiencies inside and outside the furnace, respectively. , These are the correction factors for the calcium-sulfur ratio inside and outside the furnace, respectively. For flue gas flow rate, , These represent the sulfur content inside and outside the furnace, respectively.

[0028] The optimization algorithm uses minimizing the total desulfurization cost as its objective function, which comprehensively considers the cost of limestone consumption, the power consumption of the desulfurization system, and the potential environmental penalties for exceeding emission standards. For each pair of candidate setpoints generated by the PSO algorithm, the upper-level optimizer substitutes them into a complete SO2 dynamic emission prediction model, rolling out predictions over several sampling periods to show how the system's operating state and total cost will change if the setpoints are tracked. By comparing the predicted total costs corresponding to different candidate solutions, the PSO algorithm ultimately finds the optimal setpoints that minimize the cost and outputs them to the lower-level predictive controller. This mechanism of re-optimizing based on the latest operating conditions in each period ensures that the system can dynamically track the globally most economical operating point, thereby substantially reducing operating costs.

[0029] S3. The optimal setpoint for SO2 concentration in the furnace is input to the furnace generalized predictive controller, and the optimal setpoint for SO2 concentration outside the furnace is input to the outside generalized predictive controller.

[0030] The generalized predictive controller uses the CARIMA model as the prediction model, and its objective function is as follows: , in, For the first The predicted output of the step, To predict the time domain, To control the time domain, To control the weighting coefficients, For the first The incremental control of limestone feed rate in each step.

[0031] Specifically, the real-time optimal economic setpoint calculated by the optimization algorithm in step S2— and These serve as the tracking targets for two independent generalized predictive controllers (GPCs) located inside and outside the furnace, respectively. This means that the lower-level controller no longer tracks a fixed setpoint, but rather an optimal setpoint that dynamically fluctuates with changes in unit operating conditions, always pointing towards the lowest operating cost. The inputs to these two setpoints act as a bridge connecting economic objectives with specific control actions.

[0032] The core of the lower-level generalized predictive controller is to predict the future behavior of the system based on the controlled autoregressive integral moving average (CARIMA) model. This model can well describe the process dynamics under non-stationary disturbances. For the SO2 emission control process of this invention, the model output is the SO2 concentration, and the control input is the limestone feed rate.

[0033] The controller's goal is to calculate a series of future control increments that make the predicted output as close as possible to the optimal economic setpoint from the upper layer, while ensuring that the control action is not too drastic. This goal is mathematically formalized as an objective function that needs to be minimized online: , in, For the first The predicted output of the step, To predict the time domain, To control the time domain, To control the weighting coefficients, For the first The incremental control of limestone feed rate in each step.

[0034] The first term of this function represents the sum of squares of the deviations between the predicted concentration and the optimal setpoint for SO2 concentration in the furnace desulfurization process over the entire prediction time domain. Its purpose is to achieve excellent tracking performance and ensure that the emission concentration remains stable near the optimal economic point. The second term is the penalty for the magnitude of the control increment in the future control time domain, and is a control weighting coefficient. Used to regulate the intensity of control actions, it can effectively smooth control signals and prevent frequent large-amplitude movements of limestone feed valves, thereby improving equipment life and system operational stability.

[0035] In each control cycle, the generalized predictive controller uses the latest real-time measurement data to predict the future SO2 concentration trajectory through the CARIMA model and solves for the minimum of the objective function, ultimately obtaining a control increment sequence. Only the first element of this sequence is actually applied to the controlled object. The entire process is repeated with the arrival of new measurement data in the next sampling cycle; this mechanism is called "rolling optimization."

[0036] S4. The in-furnace generalized predictive controller calculates the control increment of the limestone feed rate in the furnace based on the optimal set value of the in-furnace desulfurization SO2 concentration and the predicted value of the in-furnace SO2 concentration output by the SO2 dynamic emission prediction model or the actual measured value of the in-furnace SO2 concentration.

[0037] Step S4 is the core computational step in generating specific control commands by the lower-level in-furnace generalized predictive controller (GPC). This step receives the target, namely the optimal setpoint for the in-furnace SO2 concentration from the upper-level optimizer, and uses it as the benchmark value that the controller needs to strive to track within the current control cycle. Simultaneously, the controller obtains the current system state for feedback correction in two ways: first, by directly using the actual measured in-furnace SO2 concentration value; and second, by using the predicted in-furnace SO2 concentration value output by the SO2 dynamic emission prediction model. This predicted value, based on historical data and control variables, provides a model-perspective estimate of the current state. In practical applications, the measured value is usually compared with the model prediction value; the deviation can be used to update the prediction model online, thereby enhancing its accuracy.

[0038] Based on the aforementioned setpoints and feedback values, the in-furnace GPC controller initiates its rolling optimization calculation. Its core task is to solve an optimization problem with a sequence of future control increments as the decision variable. Specifically, the controller utilizes its built-in CARIMA prediction model to predict the trajectory of SO2 concentration changes in the furnace over the next N sampling points (prediction time domain) based on current and past data. Subsequently, it compares this predicted trajectory with the economic setpoint sequence and calculates a series of future control increments (changes in limestone feed rate) by minimizing a predefined objective function, ensuring that the predicted concentration output approximates the economic setpoint as smoothly and accurately as possible.

[0039] objective function The solution process can be mathematically transformed into a standard quadratic programming (QP) problem. Using mathematical tools such as the Diophantine equation, the predicted output can be expressed as a function of future control increments and known historical data, thus transforming the objective function into a quadratic form of the control increments. Solving this quadratic programming problem yields the optimal sequence of control increments for the next Nu control time domains, starting from the current time k.

[0040] Following the basic principles of predictive control, the controller does not execute the entire calculated control sequence, but only adopts the first element of the sequence, i.e., the control increment at the current moment. This "rolling optimization, first-item-only" strategy is a key characteristic of GPC. It means that in each control cycle, the optimization calculation is re-performed based on the latest system state, thereby continuously compensating for the effects of model errors and unknown disturbances with new measurements, ensuring the robustness of the control system. Finally, this calculated control increment is added to the control quantity from the previous moment to obtain the actual control command to be applied now, driving the limestone feeding device (such as feeder speed or valve opening) to operate, thus achieving precise, stable, and economical control of the in-furnace desulfurization process.

[0041] S5. The external generalized predictive controller calculates the control increment of the external limestone slurry addition amount based on the optimal set value of the external desulfurization SO2 concentration and the predicted value of the external SO2 concentration output by the SO2 dynamic emission prediction model or the actual measured external SO2 concentration value.

[0042] Steps S5 and S4 are logically parallel and together constitute the lower-level dual-loop control architecture of the economic generalized predictive control method, but their control object is the external desulfurization system. This step is executed by the external generalized predictive controller (GPC), whose core function is to receive another economic target, the optimal setpoint for the external desulfurization SO2 concentration, from the upper-level optimizer, and generate precise control commands for the external limestone slurry dosing equipment accordingly.

[0043] The controller's input also includes a feedback loop. It receives either the actual measured SO2 concentration outside the furnace (usually the purified net flue gas concentration) or the predicted SO2 concentration outside the furnace output from the SO2 dynamic emission prediction model. This feedback signal is crucial, as it reflects the actual effect of the desulfurization tower in real time, as well as any unforeseen disturbances (such as sudden changes in flue gas volume, slurry quality changes, nozzle blockage, etc.), providing the controller with a basis for correcting its behavior.

[0044] The external GPC controller employs the same CARIMA model predictive control principle as the internal controller, but its internal model parameters and dynamic characteristics are specifically identified and established for external desulfurization processes (such as semi-dry desulfurization towers). This model can describe the dynamic relationship between the slurry dosage and the final net flue gas SO2 concentration, typically including more significant pure time delay and time constants. The controller's core operation is to minimize its objective function, which also balances tracking performance (ensuring the predicted net flue gas concentration closely follows the economic setpoint) with control stability (limiting drastic changes in the slurry dosage).

[0045] By solving a quadratic programming problem, the external GPC controller outputs its optimization result: the control increment of the external limestone slurry dosage for the current control cycle k. This increment signal is translated into specific actuator actions, such as adjusting the frequency of the slurry pump or regulating the valve opening, thereby precisely controlling the amount of slurry injected into the desulfurization tower. Using an incremental rather than absolute value output effectively avoids actuator saturation and impact, ensuring the stability of system operation.

[0046] Thus, steps S4 and S5 together complete a full control cycle. The upper-level optimizer provides the economic target, while the two lower-level GPC controllers act as two tacit "executors," precisely controlling the two different processes of dry limestone feeding inside the furnace and wet slurry addition outside the furnace. They operate in parallel and cooperate to ensure that the SO2 emission concentration of the entire CFB unit not only stably meets environmental protection requirements but also always operates near the optimal economic setpoint, dynamically found by the upper-level optimizer, which minimizes the overall cost of energy and material consumption.

[0047] S6. Based on real-time operating conditions, dynamically adjust the weight coefficients of various performance indicators in the objective function of the in-furnace and out-of-furnace generalized predictive controllers through a fuzzy inference system.

[0048] Step S6 introduces a key adaptive mechanism of the present invention, aiming to solve the core problem that fixed weight parameters in traditional generalized predictive control (GPC) are unable to cope with complex and variable operating conditions. This step uses a fuzzy inference system to dynamically adjust the weight coefficients in the objective function of the lower-level in-furnace and external generalized predictive controllers in real time, thereby enabling the controller's behavior to intelligently adapt to different operating states and achieve an optimal balance among multiple objectives (tracking performance, economy, and control stability).

[0049] The design of this fuzzy inference system first requires determining its input and output variables. Input variables are key indicators that accurately reflect the current system operating status and performance, typically selected from a real-time database. These inputs include at least: the real-time deviation of SO2 concentration (the difference between the setpoint and the measured value), the rate of change of SO2 concentration (reflecting the magnitude and trend of disturbances), the rate of change of unit load (characterizing the severity of external disturbances), and the rate of change of limestone dosage (reflecting the working status of the actuators). All these input variables undergo normalization before entering the fuzzy inference engine, mapping them to a specific domain (e.g., [0, 1]). Output variables are the dynamically adjusted weight coefficients, namely the weight w1 for penalizing tracking errors, the weight w2 for penalizing economic costs, and the weight w3 for penalizing control increments in the objective function.

[0050] For each input and output variable, its fuzzy set and membership function need to be defined. For example, for the input "SO2 concentration deviation", multiple fuzzy sets such as "negative large (NB)", "negative medium (NM)", "zero (ZO)", "positive medium (PM)", and "positive large (PB)" can be defined, and a triangular or Gaussian membership function can be used to describe the degree to which a precise deviation value belongs to a certain fuzzy set. The weight coefficient, as the output, may have its fuzzy set defined as "small (S)", "medium (M)", and "large (L)".

[0051] The core of the system is a fuzzy rule base built upon expert knowledge and extensive historical data analysis and simulation verification. Each rule adopts the "IF-THEN" form, describing how the weights should be adjusted under specific operating conditions.

[0052] For example, a typical rule might be: "IF large load change rate AND large SO2 concentration deviation THEN tracking weight w1 is 'large', economic weight w2 is 'small', and control weight w3 is 'medium'." The logic of this rule is: when the load changes rapidly and emissions deviate significantly from the setpoint, the control system should prioritize ensuring environmental compliance. Therefore, it is necessary to increase the tracking weight to quickly eliminate the deviation, while temporarily reducing the economic weight, allowing for a short-term increase in costs in exchange for system stability.

[0053] In each control cycle, the fuzzy inference engine performs the following operations: First, it performs fuzzification, transforming the precise input value into the membership of different fuzzy sets according to the membership function; then, it performs fuzzy inference, activating the corresponding consequent (THEN) rules based on the degree to which the antecedents (IF parts) of all fuzzy rules are satisfied; finally, it performs defuzzification, typically using the centroid method, merging all activated output fuzzy sets into a precise, continuous weight coefficient output value.

[0054] Ultimately, this fuzzy inference system calculates and outputs a completely new set of weighting coefficients best suited to the rapidly changing operating conditions for the two GPC controllers in real time. For example, when the unit is running stably, it can automatically increase the economic weight w2, making the controller more inclined to save limestone; while when a sudden increase or decrease in load is detected, it immediately increases the tracking weight w1 and decreases the economic weight w2, ensuring that the control system responds quickly to suppress fluctuations in SO2 concentration and prevent environmental violations from exceeding standards. This intelligent parameter self-tuning mechanism greatly enhances the robustness and adaptability of the entire control system, and is the key to its ability to handle complex operating conditions such as deep peak shaving.

[0055] S7. Apply the control increment of limestone feed rate inside and outside the furnace to the controlled object or the SO2 dynamic emission prediction model, and return to step S2 in the next sampling cycle to achieve rolling optimization control.

[0056] Step S7 lays the foundation for continuous optimization in the next cycle. The specific implementation of this step involves two core actions: First, the execution of control commands, which involves converting the limestone feed rate control increments calculated by the in-furnace and external generalized predictive controllers in steps S4 and S5 into actual operating signals. These control increments are accumulated with the control values ​​from the previous moment to generate the absolute control value to be applied at the current moment. This value is then used by the output module of an industrial control system (such as a DCS) to drive precise actions in the field actuators (such as frequency converters adjusting the limestone feeder or regulating valves controlling the slurry flow), thereby actually influencing the desulfurization process.

[0057] Secondly, while the control increments are applied to the controlled object (the actual CFB unit), these control commands are also simultaneously applied to the SO2 dynamic emission prediction model constructed in step S1. The main purpose of this is to use the latest control variables to update the internal state of the model, enabling the model to more accurately predict the dynamic response of the system at the next moment, and providing a more reliable predictive basis for subsequent optimization calculations.

[0058] After completing all operations in the current cycle k, the system does not stop but enters a waiting state until the next preset sampling time (k+1). At that time, the system will re-collect the latest real-time operating data (such as new SO2 concentration measurements, load commands, etc.) and immediately return to step S2 to start a new control cycle. In this new cycle, the upper-level optimizer will recalculate the economically optimal setpoint for future periods based on the updated model and the latest measurement data; while the lower-level GPC controller will also re-solve for the optimal control increment based on the new setpoint and new feedback.

[0059] This closed-loop process, running cyclically, constitutes what is known as rolling feedback control. Its core advantage lies in the fact that every control decision is made based on the latest system status information. This allows the control system to continuously perceive process changes, evaluate control effectiveness, and adjust strategies in a timely manner, much like an experienced operator. This enables it to respond quickly and accurately to internal and external disturbances such as unit load fluctuations and coal quality changes, ultimately ensuring that SO2 emission control remains on a stable, economical, and compliant optimal operating track.

[0060] The control system for SO2 emissions from the circulating fluidized bed unit described in the method is characterized by comprising: Data communication module 210 is used to acquire real-time operating data of the circulating fluidized bed unit; SO2 dynamic emission prediction model module 220 is used to predict the dynamic changes of SO2 concentration based on the operating data; The optimization setpoint calculation module 230 is used to run the optimization algorithm to calculate the optimal SO2 concentration setpoint for desulfurization inside and outside the furnace; The generalized predictive control module 240 includes an in-furnace generalized predictive control submodule and an out-of-furnace generalized predictive control submodule, which are respectively used to receive the optimal setpoint and generate control commands; The fuzzy adaptive weight adjustment module 250 is used to dynamically adjust the objective function weights of the generalized prediction control module according to the real-time operating status. Each of these modules is integrated into a simulation computing platform or an industrial control system.

[0061] The optimization setpoint calculation module is configured to execute a particle swarm optimization algorithm or a genetic algorithm, and the input of the fuzzy adaptive weight adjustment module includes one or more of the following: SO2 concentration deviation, load change rate, and limestone dosage change rate.

[0062] Specifically, the implementation of the control system relies on the collaborative work and data interaction between various modules to form a closed-loop intelligent control system. The data communication module, as the system's foundational sensing layer, uses standard protocols such as Industrial Ethernet and OPC UA to exchange data at high speed and reliably with the distributed control system (DCS) and plant-level monitoring information system (SIS) of the circulating fluidized bed unit. It continuously acquires real-time operational data throughout the entire process, including unit load, flue gas parameters, measured SO2 concentration, and limestone dosage, providing accurate and timely input for all subsequent advanced algorithm modules.

[0063] The SO2 dynamic emission prediction model module is the core of the system's "digital twin." After receiving real-time data from the data communication module, it runs a hybrid model that combines dynamic mechanisms with data-driven approaches. This module not only calculates the current SO2 concentration state estimate in real time, but more importantly, it can predict the changing trajectory of SO2 concentrations inside and outside the furnace and the corresponding desulfurization costs over several sampling periods based on assumed future control strategies. This provides crucial future scenario extrapolation capabilities for upper-level optimization decisions.

[0064] The optimized setpoint calculation module acts as the system's "economic brain," with its built-in intelligent optimization algorithms (such as particle swarm optimization) aiming to minimize the total desulfurization cost. In each control cycle, this module utilizes forward-looking data provided by predictive models to perform multi-objective trade-off calculations while meeting environmental emission constraints. Ultimately, it outputs a set of real-time updated, economically optimal SO2 concentration setpoints for both inside and outside the furnace, thus translating macroeconomic objectives into microeconomic control commands.

[0065] The generalized predictive control module acts as the system's "intelligent actuator," employing a dual-loop design comprising two independent predictive control sub-modules: one inside the furnace and one outside. Each sub-module receives setpoint instructions from the optimization module and, combined with real-time feedback data, calculates precise limestone feed control increments by solving a rolling time-domain optimization problem. Its core advantage lies in its ability to proactively predict future system dynamics and make smooth and precise control actions in advance, effectively overcoming the control challenges caused by large time lags.

[0066] The fuzzy adaptive weight adjustment module endows the control system with "contextual awareness" and "intelligent decision-making" capabilities. This module continuously monitors the changing trends of key performance indicators such as SO2 concentration deviation and load change rate. Through a built-in fuzzy inference rule base, it dynamically adjusts the weight coefficients of various performance indicators (such as tracking error, economy, and control stability) in the objective function of the generalized predictive control module in real time. This mechanism ensures that the control system can automatically switch the dominant strategy under different operating conditions (such as stable operation and deep peak shaving), always maintaining optimal overall performance.

[0067] Ultimately, all these functional modules are highly integrated into a unified simulation computing platform (such as the MATLAB / Simulink environment for algorithm verification) or directly embedded into industrial control systems (such as high-end IPC-based controllers). They exchange data and transmit commands through standardized software interfaces, forming a complete automated loop from sensing and decision-making to execution, achieving real-time, economical, stable, and adaptive advanced control of SO2 emissions from circulating fluidized bed units.

[0068] Electronic device 300 can be a desktop computer, laptop, handheld computer, cloud server, or other electronic device. Electronic device 300 may include, but is not limited to, processor 301 and memory 302. Those skilled in the art will understand that... Figure 3 This is merely an example of electronic device 300 and does not constitute a limitation on electronic device 300. It may include more or fewer components than shown, or combine certain components, or different components. For example, electronic device may also include input / output devices, network access devices, buses, etc.

[0069] Processor 301 can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor can be a microprocessor or any conventional processor.

[0070] The memory 302 can be an internal storage unit of the electronic device 300, such as a hard disk or RAM of the electronic device 300. The memory 302 can also be an external storage device of the electronic device 300, such as a plug-in hard disk, Smart Media Card (SMC), Secure Digital (SD) card, or Flash Card equipped on the electronic device 300. Furthermore, the memory 302 can include both internal and external storage units of the electronic device 300. The memory 302 is used to store the computer program 303 and other programs and data required by the electronic device. The memory 302 can also be used to temporarily store data that has been output or will be output.

[0071] The above embodiments are only used to illustrate the technical solutions of this disclosure, and are not intended to limit it. Although this disclosure has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this disclosure, and should all be included within the protection scope of this disclosure.

Claims

1. A method for controlling SO2 emissions from a circulating fluidized bed unit, characterized in that, Includes the following steps: S1: Construct a dynamic SO2 emission prediction model for circulating fluidized bed units. The model includes an SO2 generation sub-model, an in-furnace desulfurization sub-model, an out-of-furnace desulfurization sub-model, and a desulfurization economic cost sub-model. S2: Based on the SO2 dynamic emission prediction model and real-time operating data in the current control cycle, the optimal setpoints for SO2 concentration in the furnace desulfurization and SO2 concentration outside the furnace desulfurization are calculated using an optimization algorithm to minimize the total desulfurization cost in the future period. S3: Input the optimal setpoint of SO2 concentration in the furnace to the furnace generalized predictive controller, and input the optimal setpoint of SO2 concentration outside the furnace to the outside generalized predictive controller; S4: The in-furnace generalized predictive controller calculates the control increment of the limestone feed rate in the furnace based on the optimal setpoint of the in-furnace desulfurization SO2 concentration and the predicted value of the in-furnace SO2 concentration output by the SO2 dynamic emission prediction model or the actual measured value of the in-furnace SO2 concentration. S5: The external generalized predictive controller calculates the control increment of the external limestone slurry addition amount based on the optimal set value of the external desulfurization SO2 concentration and the predicted value of the external SO2 concentration output by the SO2 dynamic emission prediction model or the actual measured external SO2 concentration value. S6: Based on the real-time operating conditions, dynamically adjust the weight coefficients of various performance indicators in the objective function of the in-furnace and out-of-furnace generalized predictive controllers through the fuzzy inference system; S7: Apply the control increment of limestone feed rate inside and outside the furnace to the controlled object or the SO2 dynamic emission prediction model, and return to step S2 in the next sampling cycle to achieve rolling optimization control.

2. The method according to claim 1, characterized in that, The optimization algorithm mentioned in step S2 is either particle swarm optimization algorithm or genetic algorithm.

3. The method according to claim 1, characterized in that, The optimal setpoints for SO2 concentration in in-furnace desulfurization and SO2 concentration in external desulfurization mentioned in step S2 are calculated using the following formulas: ， ， in, This is the optimal setpoint for SO2 concentration in the furnace desulfurization process. This is the optimal setpoint for SO2 concentration in the external desulfurization process. The amount of limestone fed into the furnace. This refers to the amount of limestone supplied outside the furnace. , These represent the desulfurization efficiencies inside and outside the furnace, respectively. , These are the correction factors for the calcium-sulfur ratio inside and outside the furnace, respectively. For flue gas flow rate, , These represent the sulfur content inside and outside the furnace, respectively.

4. The method according to claim 1, characterized in that, The generalized predictive controller uses the CARIMA model as the prediction model, and its objective function is as follows: , in, For the first The predicted output of the step, To predict the time domain, To control the time domain, To control the weighting coefficients, For the first The incremental control of limestone feed rate in each step.

5. The method according to claim 1, characterized in that, The input variables of the fuzzy inference system in step S6 include one or more of the following: SO2 concentration change rate, load change rate, and limestone usage change rate. The output variables are the weighting coefficients of tracking error, economic indicators, and control quantity stability in the objective function.

6. The method according to claim 5, characterized in that, The fuzzy inference system uses the centroid method for defuzzification, and its fuzzy rules dynamically adjust the weight allocation based on the operating conditions.

7. A control system for SO2 emissions from a circulating fluidized bed unit implementing the method of any one of claims 1-6, characterized in that, include: The data communication module is used to acquire real-time operating data of the circulating fluidized bed unit; The SO2 dynamic emission prediction model module is used to predict the dynamic changes in SO2 concentration based on the operational data. The optimized setpoint calculation module is used to run optimization algorithms to calculate the optimal SO2 concentration setpoints for desulfurization inside and outside the furnace; The generalized predictive control module includes an in-furnace generalized predictive control submodule and an out-of-furnace generalized predictive control submodule, which are respectively used to receive the optimal setpoint and generate control commands; The fuzzy adaptive weight adjustment module is used to dynamically adjust the objective function weights of the generalized prediction control module according to the real-time operating status. Each of these modules is integrated into a simulation computing platform or an industrial control system.

8. The system according to claim 7, characterized in that, The optimization setpoint calculation module is configured to execute a particle swarm optimization algorithm or a genetic algorithm, and the input of the fuzzy adaptive weight adjustment module includes one or more of the following: SO2 concentration deviation, load change rate, and limestone dosage change rate.

9. An electronic device, characterized in that, include: One or more processors; A storage unit is used to store one or more programs that, when executed by one or more processors, enable the one or more processors to implement the SO2 emission control method of the circulating fluidized bed unit according to claims 1-7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it can implement the SO2 emission control method of the circulating fluidized bed unit according to claims 1-8.

Images