Multivariable self-adaptive optimization control method for deacidification of waste incineration flue gas

Abstract

The invention belongs to the technical field of waste incineration flue gas deacidification, and particularly relates to a waste incineration flue gas deacidification multivariable adaptive optimization control method, which comprises the following steps: collecting combustion side data and feedback control side data in a waste incineration process, determining system lag time, and performing translation alignment on the combustion side data according to the system lag time; establishing a deacidification reaction process model, calculating a forgetting factor, and updating parameters of the deacidification reaction process model by using the forgetting factor; and calculating a theoretical control quantity based on the updated deacidification reaction process model parameters, calculating an environmental impedance compensation gain coefficient, and correcting the theoretical control quantity by using the environmental impedance compensation gain coefficient to obtain a final control quantity. According to the method, the self-adaptive capacity of the system can be remarkably improved by utilizing the feedforward characteristics of the combustion side and environmental impedance correction, and flue gas emission is ensured to stably reach the standard.

Description

Technical Field

[0001] This invention relates to the field of flue gas desulfurization technology for waste incineration. More specifically, this invention relates to a multivariate adaptive optimization control method for flue gas desulfurization in waste incineration. Background Technology

[0002] As the main method of urban waste disposal, power generation is subject to increasingly stringent emission standards for flue gas generated during the process. In the purification process of flue gas generated during waste incineration power generation, semi-dry or dry deacidification is the key step in removing acidic gases. This process usually involves injecting lime slurry or slaked lime powder into the reaction tower of waste incineration, so that the lime slurry or slaked lime powder can chemically neutralize the flue gas generated during waste incineration, thereby achieving the purpose of purifying the flue gas generated during waste incineration.

[0003] Currently, the control of acid removal in waste incineration mainly relies on feedback data from the flue gas online monitoring system, which uses PID control or conventional model predictive control. This control method has significant limitations: First, there is a transmission and detection lag of several minutes as the flue gas flows from the combustion chamber through the reaction tower and bag filter to the CEMS sampling point. When the calorific value of the feed waste suddenly changes, causing a sudden surge in acidic gases, the feedback control often fails to respond in time, resulting in instantaneous emissions exceeding the standard. Second, the model parameters cannot adapt to drastic changes in operating conditions. Existing adaptive control algorithms usually use a fixed forgetting factor, which cannot take into account the need for long memory under stable operating conditions and the need for short memory under drastic fluctuations. This can easily lead to the system tracking slowly under sudden operating conditions or oscillations due to noise interference under steady-state conditions. Summary of the Invention

[0004] To address the technical problem of ineffective response to sudden changes in waste incineration conditions and resulting in excessive flue gas emissions due to feedback lag and fixed model parameters, this invention proposes a multivariate adaptive optimization control method for acid removal from waste incineration flue gas. This method can significantly improve the system's adaptive capability by utilizing combustion-side feedforward characteristics and environmental impedance correction, and ensure stable compliance of flue gas emissions.

[0005] In a first aspect, the present invention provides a multivariate adaptive optimization control method for desulfurization of waste incineration flue gas, comprising: collecting combustion-side data and feedback control-side data during the waste incineration process, calculating the temporal correlation between the combustion-side data and the feedback control-side data, determining the system lag time, aligning the combustion-side data according to the system lag time to form a training dataset; extracting load fluctuation rate, combustion intensity gain, and air volume mutation characteristics from the aligned combustion-side data to construct a combustion disturbance energy index intensity characterizing the degree of operational instability; establishing a desulfurization reaction process model, calculating a variable forgetting factor based on the combustion disturbance energy index intensity and the current hydrogen chloride concentration at the chimney inlet, iteratively updating the parameters of the desulfurization reaction process model using a recursive least squares method with a variable forgetting factor; calculating a theoretical control quantity based on the updated parameters of the desulfurization reaction process model, calculating an environmental impedance compensation gain coefficient based on the flue gas temperature and humidity at the desulfurization tower inlet, correcting the theoretical control quantity using the environmental impedance compensation gain coefficient to obtain the final control quantity, and completing the desulfurization control by adjusting the action amplitude of the actuator.

[0006] By adopting the above technical solution, combustion-side data and feedback control-side data during the waste incineration process are collected, and time-series translation and alignment are performed according to the system lag time, eliminating the negative impact of physical lag on modeling accuracy. At the same time, the adaptive update of the deacidification reaction process model parameters is driven by the intensity of combustion disturbance energy index, and combined with environmental physical impedance correction, the system's adaptability to sudden changes in operating conditions and environmental changes is improved, ensuring stable compliance of flue gas emissions.

[0007] Preferably, the collection of combustion-side data and feedback control-side data during the waste incineration process includes: using the main steam flow rate collected by the vortex flow meter, the average furnace temperature collected by the thermocouple array, and the air volume data collected by the air volume measurement device as combustion-side data; and using the flue gas temperature at the desulfurization tower inlet collected by the temperature sensor, the flue gas humidity at the desulfurization tower inlet collected by the humidity sensor, and the hydrogen chloride concentration at the chimney inlet collected by the flue gas online monitoring system as feedback control-side data.

[0008] By adopting the above technical solution, precision sensors such as vortex flow meters, thermocouple arrays, and air volume measurement devices are used to obtain feedforward combustion-side data, and temperature sensors, humidity sensors, and flue gas online monitoring systems are combined to obtain feedback control-side data. This ensures the accuracy and reliability of the control system input signals and lays a solid data foundation for the subsequent construction of an accurate deacidification reaction process model.

[0009] Preferably, determining the system lag time includes: establishing a historical data sliding window, using Spearman's rank correlation coefficient to calculate the correlation between the main steam flow rate and the hydrogen chloride concentration at the chimney inlet at different time offsets, and selecting the time offset corresponding to the largest absolute value of the correlation coefficient as the system lag time; forming a training dataset includes: shifting the sampling time point of the combustion-side data backward by the system lag time to align it with the sampling time point of the feedback control-side data.

[0010] By adopting the above technical solution, a sliding window of historical data is established, and the Spearman rank correlation coefficient is used to calculate the time-series correlation. This enables the dynamic and accurate determination of the system lag time under different operating conditions, thereby achieving a strict match between combustion-side data and feedback control-side data in terms of causal relationship. This effectively reduces the probability of model distortion caused by misalignment of data causal relationship.

[0011] Preferably, the intensity of the combustion disturbance energy index satisfies the following relationship:

[0012] in, Indicates the first The intensity of the combustion disturbance energy index at any given moment; and These represent the main steam flow rates at the current time and the previous time, respectively. This represents the moving average of the main steam flow rate over a preset time period. This indicates the current average temperature of the furnace. Indicates the rated reference furnace temperature under design operating conditions; and These represent the primary air volume at the current moment and the previous moment, respectively; and This represents the preset dimensionless weighting coefficient; This represents the temperature nonlinearity factor.

[0013] By adopting the above technical solution, a combustion disturbance energy index intensity that includes the relative change rate of main steam flow, combustion intensity gain, and air volume mutation characteristics is constructed. This transforms the changes in multi-dimensional physical parameters into a single scalar index that reflects the degree of instability of the operating conditions, enabling the system to detect potential disturbances caused by fluctuations in waste composition in advance, thereby significantly enhancing the system's feedforward regulation sensitivity.

[0014] Preferably, the construction of the combustion disturbance energy index intensity characterizing the degree of operational instability includes: calculating the square of the normalized difference of the main steam flow rate relative to the moving average as a term reflecting the boiler load fluctuation energy; calculating the exponent of the ratio of the current furnace average temperature to the rated reference furnace temperature as a combustion intensity gain term reflecting the amplification effect of reaction kinetics at high temperatures; calculating the natural logarithm of the absolute value of the primary air volume change as an air volume mutation term reflecting the degree of airflow disturbance; and taking the square root after weighted coupling of the term reflecting the boiler load fluctuation energy, the combustion intensity gain term, and the air volume mutation term to obtain the combustion disturbance energy index intensity.

[0015] Preferably, the variable forgetting factor satisfies the following relationship:

[0016] in, Indicates the first The forgetting factor changes over time; and These represent the upper and lower limits of the forgetting factor, respectively; This represents the sensitivity adjustment coefficient; Indicates the first The intensity of the combustion disturbance energy index at any given moment; Indicates the inflection point of the disturbance threshold; Represents the safety bias constant; This indicates the current actual monitored concentration of hydrogen chloride at the chimney inlet; This represents the safety baseline concentration constant.

[0017] By adopting the above technical solution, a variable forgetting factor is calculated, which includes the intensity of the combustion disturbance energy index and the current actual monitored hydrogen chloride concentration at the chimney inlet. The nonlinear characteristics of the Sigmoid function are used to establish a mapping relationship between the degree of disturbance and the model memory length, thereby realizing a smooth switch between the strong filtering characteristics of the algorithm under steady-state conditions and the fast tracking capability under sudden change conditions. This effectively solves the technical problem that traditional algorithms cannot balance robustness and sensitivity.

[0018] Preferably, the calculation process of the variable forgetting factor includes: determining whether the calculated variable forgetting factor value is greater than 1, and if it is greater than 1, forcing it to be 1; when the intensity of the combustion disturbance energy index exceeds the inflection point of the disturbance threshold, the variable forgetting factor is reduced to the lower limit of the forgetting factor through the Sigmoid function term.

[0019] Preferably, the calculation of the theoretical control quantity based on the updated model parameters includes: using the generalized minimum variance control law, with the set value of the hydrogen chloride concentration at the chimney inlet as the target, and based on the updated deacidification reaction process model parameters, calculating the theoretical lime slurry required to maintain the target concentration.

[0020] Preferably, the final control quantity satisfies the following relationship:

[0021] in, Indicates the first The final control quantity at any given moment; This represents the calculated theoretical control quantity; This indicates the actual flue gas temperature at the inlet of the deacidification tower; This indicates the center value of the optimal deacidification reaction temperature window; This indicates the actual humidity at the inlet of the deacidification tower; Indicates the basic humidity constant; This represents the impedance compensation gain coefficient.

[0022] By adopting the above technical solution, the actual flue gas temperature and humidity at the inlet of the deacidification tower are used to calculate the environmental impedance compensation gain coefficient, which transforms the objective laws of chemical reaction kinetics into compensation commands. In environments where the reaction efficiency decreases due to deviation from the optimal reaction temperature or low humidity, the system automatically increases the theoretical lime slurry demand, making up for the shortcomings of the pure data-driven model in extreme physical environments, thereby ensuring that the final control quantity can still meet the actual deacidification requirements in complex environments.

[0023] Preferably, after obtaining the final control quantity, the method further includes: converting the final control quantity into a frequency command for the rotary spray pump or an opening command for the regulating valve, and sending it to the field actuator to counteract the effects of combustion-side disturbances and environmental impedance on the deacidification efficiency by changing the flow rate of lime slurry injected into the deacidification tower.

[0024] The beneficial effects of this invention are as follows: This invention proposes a control strategy based on combustion-side feedforward feature extraction. By constructing the combustion disturbance energy index intensity, it can detect the operating condition fluctuations caused by sudden changes in waste composition in advance, effectively solving the problem of large lag in traditional feedback control.

[0025] Furthermore, this invention designs a variable forgetting factor update mechanism based on the intensity of combustion disturbance energy exponent and safety bias. Utilizing the nonlinear characteristics of the Sigmoid function, it achieves adaptive switching between steady-state and transient states, effectively resolving the contradiction between robustness and sensitivity. In addition, this invention establishes a control quantity correction model based on physical environmental impedance, converting the influence of flue gas temperature and humidity at the desulfurization tower inlet on reaction efficiency into impedance coefficients to compensate for the theoretical control quantity, significantly improving the system's control accuracy and stability under complex environments. Attached Figure Description

[0026] Figure 1 This is a schematic flowchart illustrating a multivariate adaptive optimization control method for desulfurization of waste incineration flue gas according to the present invention; Figure 2 This is a schematic diagram of the phase plane distribution of combustion disturbance energy in this invention; Figure 3 This is a schematic diagram of the distribution of the environmental impedance field and operating point in this invention. Detailed Implementation

[0027] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are some embodiments of the present invention, but not all embodiments.

[0028] This invention discloses a multivariable adaptive optimization control method for acid removal from waste incineration flue gas, referring to... Figure 1 This includes steps S1-S4: S1. Collect combustion-side data and feedback control-side data during the waste incineration process, calculate the temporal correlation between combustion-side data and feedback control-side data, determine the system lag time, and perform translation and alignment of combustion-side data according to the system lag time to form a training dataset.

[0029] In an optional embodiment, the present invention addresses the problem of unclear data sources by specifying the data acquisition method. Specifically, the present invention acquires feedforward combustion-side data and feedback control-side data at a sampling frequency of 1Hz through the power plant's distributed control system.

[0030] The feedforward combustion-side data collected by the distributed control system includes the main steam flow rate. Average furnace temperature and primary air volume The feedback control side data collected by the distributed control system includes the flue gas temperature at the inlet of the desulfurization tower. Humidity of flue gas at the inlet of the deacidification tower and the concentration of hydrogen chloride at the chimney inlet .

[0031] Specifically, main steam flow rate The unit is tons per hour, and the data comes from a vortex flow meter installed on the steam pipe at the outlet of the waste heat boiler. The main steam flow reflects the current overall load level of the boiler; the average furnace temperature... The unit is standard cubic meters per hour, and the data comes from a K-type thermocouple array evenly distributed at different heights around the incinerator furnace. The arithmetic mean of all valid thermocouple readings is taken. The average furnace temperature characterizes the current combustion intensity; primary air volume The unit is standard cubic meters per hour, and the data comes from the Venturi flow meter on the primary air fan outlet duct. The primary air volume characterizes the supply of combustion air; the flue gas temperature at the inlet of the deacidification tower... The unit is degrees Celsius, which is measured by a resistance temperature detector (RTD) sensor installed in the inlet flue of the deacidification reaction tower; the humidity of the flue gas at the inlet of the deacidification tower... The unit is a percentage, measured by a high-temperature humidity meter; hydrogen chloride concentration at the chimney inlet. The unit is milligrams per cubic meter, which is measured by an online flue gas monitoring system analyzer installed behind the bag filter and at the chimney inlet.

[0032] It is important to note that due to the time lag between physical transport and chemical reaction in the transmission of changes from the combustion end to the emission end, directly using data from the same moment in time for modeling would lead to a misalignment of causal relationships. Therefore, this invention establishes a length of... This problem can be solved using a historical data sliding window, and the specific operation method is as follows: Within this historical data sliding window, the main steam flow sequence is calculated using the Spearman rank correlation coefficient. Sequence of hydrogen chloride concentration at chimney inlet At different time offsets The correlation below, where, The search range was set to 60 to 600 seconds, selecting the period corresponding to the largest absolute value of the correlation coefficient. The value is used as the current system lag time. Finally, the sampling time of the combustion-side data... Data shift corresponds to the data sampling time on the feedback side. The data is used to form an input-output aligned training dataset.

[0033] As one implementation of this invention, when the calculation yields... If the time interval is 180 seconds, then during model training, the input term is taken from the time of 12:00:00. , , The corresponding output item is taken from the time 12:03:00. .

[0034] In this way, by clearly identifying the sensor source and performing strict timing alignment, the impact of large physical lag on modeling accuracy is eliminated, ensuring that the model input variables can truly reflect the root causes of output changes.

[0035] S2. Based on the aligned combustion-side data, extract the load fluctuation rate, combustion intensity gain and air volume change characteristics to construct the combustion disturbance energy index intensity, which characterizes the degree of instability of the operating conditions.

[0036] In an optional embodiment, in order to unify the collected multi-dimensional data and provide an accurate basis for subsequent control strategy switching, the present invention constructs a combustion disturbance energy index intensity. Combustion disturbance energy index intensity The calculation method is as follows: ; in, For the first The intensity of the combustion perturbation energy index at any given moment, a dimensionless scalar. This represents the current main steam flow rate. This represents the main steam flow rate at the previous moment; The moving average of the main steam flow rate over the past 10 minutes; it serves as a normalization benchmark to prevent the denominator from being zero, and in this invention, its lower limit is set at 10 t / h; The squared rate of change of the main steam flow rate is used to characterize the relative change rate of the main steam flow rate in order to obtain the energy concept and amplify the fluctuation. This represents the current average temperature of the furnace. The rated reference furnace temperature under design conditions is set to 850 degrees Celsius in this embodiment of the invention. This is a temperature nonlinearity factor, which is set to 2 in this embodiment of the invention; The gain in combustion intensity reflects the physical fact that the reaction rate is accelerated at high temperatures, and small disturbances are amplified. This represents the airflow at the current moment. This refers to the airflow volume at the previous moment; It is used to characterize sudden changes in wind volume. It is important to note that... Taking the absolute value of the numerical value is used in logarithmic functions to compress the data range and prevent excessive overflow caused by drastic airflow adjustments. and These are all preset dimensionless weighting coefficients used to balance the numerical ranges of terms with different dimensions. In this embodiment of the invention, Set to 100. Set to 1.

[0037] To more clearly illustrate the intensity of the combustion disturbance energy index The function and calculation process of this invention will be explained by example in the following embodiments: First, assume that the system is at the moment of a sudden change in operating conditions. Moment It is 60. for Current moment of It is 63. for , for benchmark value It is 60. It is 850; but .

[0038] When the system is in steady state, the difference in flow rate and air volume is close to 0, so the calculated intensity of the combustion disturbance energy index is also close to 0. However, when the system is in a sudden change in operating conditions, the calculated intensity of the combustion disturbance energy index is much greater than the value in steady state, which clearly indicates that the system is in a high disturbance state at this time.

[0039] In this way, multi-dimensional changes in physical parameters can be characterized as a single disturbance index, providing an accurate basis for switching subsequent control strategies.

[0040] S3. Establish a deacidification reaction process model, calculate the variable forgetting factor based on the intensity of combustion disturbance energy index and the current hydrogen chloride concentration at the chimney inlet, and iteratively update the parameters of the deacidification reaction process model using the recursive least squares method with the variable forgetting factor.

[0041] In an optional embodiment, to address the problem that traditional fixed forgetting factors cannot simultaneously achieve steady-state accuracy and dynamic tracking speed, this invention constructs a variable forgetting factor. Furthermore, the deacidification reaction model was identified online using the recursive least squares method, and the forgetting factor was varied. The construction method is as follows: ; in, In this embodiment of the invention, it is set to 0.995; In this embodiment of the invention, it is set to 0.95; This is the sensitivity coefficient, which is set to 0.5 in this embodiment of the invention; The perturbation threshold is set to 2 in this embodiment of the invention. This is a safety bias constant, which is set to 1 in this embodiment of the invention; The safety baseline concentration is set to 50 in this embodiment of the invention, and its unit is mg / m³. The concentration of HCl is measured, and its unit is mg / m³. It is a Sigmoid term; This is a safety bias term.

[0042] To more clearly explain the forgetting factor The function and calculation process of this invention will be illustrated by the following examples in this embodiment: First, using the data from the previous description of the system at the instant of a sudden change in operating conditions, the intensity of the combustion disturbance energy index at this time is... It is 3.15; assuming at this time If it is 70, then it can be calculated that = .

[0043] The calculations above show that due to the large perturbation and high emission concentration, the safety bias term and the sigmoid term are both relatively small, leading to a lower calculated variable forgetting factor. The value is close to its lower limit of 0.95.

[0044] It's important to note that a smaller forgetting factor allows the algorithm to quickly forget old data, primarily updating the model based on the latest data, thus rapidly tracking sudden changes in operating conditions. Conversely, if... for , If the value is 10, then the exponent of the Sigmoid term is a large negative value, the denominator approaches 1, and the fraction value approaches 1. The value of the safety bias term is The final calculated variable forgetting factor The value is greater than 1 and is forced to be 1. At this time, the model has a long memory and strong noise resistance.

[0045] Thus, through the variable forgetting factor mechanism, the system automatically switches between fast tracking and steady-state filtering, ensuring that the model parameters are always optimal.

[0046] S4. Calculate the theoretical control quantity based on the updated deacidification reaction process model parameters, and calculate the environmental impedance compensation gain coefficient according to the flue gas temperature and humidity at the deacidification tower inlet. Use the environmental impedance compensation gain coefficient to correct the theoretical control quantity to obtain the final control quantity. Deacidification control is completed by adjusting the action amplitude of the actuator.

[0047] In an optional embodiment, the theoretical amount of lime slurry required to maintain emission compliance is first calculated based on the updated model. Since physical environmental factors such as low temperature and dryness can also hinder chemical reactions, this invention dynamically corrects the final control quantity. The specific correction method is as follows: ; in, This refers to the final amount of lime slurry injected; The inlet smoke temperature is measured in degrees Celsius. The optimal reaction temperature is expressed in degrees Celsius, and is set to 145 degrees Celsius in this embodiment of the invention. The measured inlet humidity is expressed in %; The basic humidity constant is set to 5% in this embodiment of the invention. The impedance compensation gain is set to 0.5 in this embodiment of the invention.

[0048] To more clearly illustrate the process of correcting the final amount of lime slurry injected in the embodiments of the present invention, the following example will be used: First, theoretical calculations The corresponding pump frequency is 40Hz, and the current operating conditions are quite harsh. Low ; Since the dryness level is 4%, the final amount of lime slurry sprayed is... Hz. It should be noted that if the actual device has a maximum frequency limit, such as 50Hz, then the output will operate at full load at 50Hz.

[0049] The calculations above show that the reaction efficiency is low in harsh environments with low temperature and low humidity. In such cases, the theoretically calculated 40Hz frequency is insufficient to meet the standard, and the injection rate must be significantly increased to force compliance. Finally, the final injection volume of the lime slurry is sent to the rotary spray pump frequency converter for adjustment.

[0050] In this way, by correcting for environmental impedance, the objective laws of chemical reaction kinetics are integrated into the control algorithm, ensuring that the control quantities can meet the actual deacidification requirements under complex climate and operating conditions, thus avoiding the failure of purely theoretical calculations.

[0051] Reference Figure 2 The steady-state control region is located in Figure 2 The central region, its data points The numerical value is relatively small, the system adopts a high forgetting factor, and the high-disturbance adaptive region is located on the periphery of the steady-state control region. When the value is large, the system automatically switches to a low forgetting factor. This distribution intuitively reflects that the present invention can effectively distinguish between steady-state and sudden operating conditions.

[0052] Reference Figure 3 The darker the background color, the greater the environmental impedance compensation gain coefficient. When the operating point deviates from the optimal response temperature window or falls into a low humidity area, the background color becomes darker, which intuitively demonstrates the mechanism of automatic compensation control based on the physical environment of this invention.

[0053] In the description of this specification, "multiple" or "several" means at least two, such as two, three or more, unless otherwise expressly and specifically defined.

Claims

1. A multivariable adaptive optimization control method for acid removal from waste incineration flue gas, characterized in that, include: Data from the combustion side and feedback control side during the waste incineration process are collected, and the temporal correlation between the combustion side data and the feedback control side data is calculated to determine the system lag time. The combustion side data is then shifted and aligned based on the system lag time to form a training dataset. Based on the aligned combustion-side data, load fluctuation rate, combustion intensity gain and air volume change characteristics are extracted to construct the combustion disturbance energy index intensity, which characterizes the degree of instability of the operating conditions. A deacidification reaction process model is established. The variable forgetting factor is calculated based on the intensity of the combustion disturbance energy index and the current hydrogen chloride concentration at the chimney inlet. The parameters of the deacidification reaction process model are iteratively updated using the recursive least squares method with the variable forgetting factor. The theoretical control quantity is calculated based on the updated deacidification reaction process model parameters, and the environmental impedance compensation gain coefficient is calculated based on the flue gas temperature and humidity at the deacidification tower inlet. The theoretical control quantity is then corrected using the environmental impedance compensation gain coefficient to obtain the final control quantity. The deacidification control is completed by adjusting the action amplitude of the actuator.

2. The multivariable adaptive optimization control method for desulfurization of waste incineration flue gas according to claim 1, characterized in that, The collection of combustion-side data and feedback control-side data during the waste incineration process includes: using the main steam flow rate collected by the vortex flow meter, the average furnace temperature collected by the thermocouple array, and the air volume data collected by the air volume measurement device as combustion-side data; and using the flue gas temperature at the desulfurization tower inlet collected by the temperature sensor, the flue gas humidity at the desulfurization tower inlet collected by the humidity sensor, and the hydrogen chloride concentration at the chimney inlet collected by the flue gas online monitoring system as feedback control-side data.

3. The multivariable adaptive optimization control method for deacidification of waste incineration flue gas according to claim 1, characterized in that, The determination of the system lag time includes: establishing a historical data sliding window, using Spearman's rank correlation coefficient to calculate the correlation between the main steam flow rate and the hydrogen chloride concentration at the chimney inlet at different time offsets, and selecting the time offset corresponding to the largest absolute value of the correlation coefficient as the system lag time; the formation of the training dataset includes: shifting the sampling time point of the combustion-side data backward by the system lag time to align it with the sampling time point of the feedback control-side data.

4. The multivariable adaptive optimization control method for deacidification of waste incineration flue gas according to claim 1, characterized in that, The intensity of the combustion disturbance energy index satisfies the following relationship: in, Indicates the first The intensity of the combustion disturbance energy index at any given moment; and These represent the main steam flow rates at the current time and the previous time, respectively. This represents the moving average of the main steam flow rate over a preset time period. This indicates the current average temperature of the furnace. Indicates the rated reference furnace temperature under design operating conditions; and These represent the primary air volume at the current moment and the previous moment, respectively; and This represents the preset dimensionless weighting coefficient; This represents the temperature nonlinearity factor.

5. The multivariable adaptive optimization control method for deacidification of waste incineration flue gas according to claim 4, characterized in that, The construction of the combustion disturbance energy index intensity characterizing the degree of operational instability includes: calculating the square of the normalized difference of the main steam flow rate relative to the moving average as a term reflecting the boiler load fluctuation energy; calculating the exponent of the ratio of the current furnace average temperature to the rated reference furnace temperature as a combustion intensity gain term reflecting the amplification effect of reaction kinetics at high temperatures; calculating the natural logarithm of the absolute value of the primary air volume change as an air volume mutation term reflecting the degree of airflow disturbance; and taking the square root after weighted coupling of the term reflecting the boiler load fluctuation energy, the combustion intensity gain term, and the air volume mutation term to obtain the combustion disturbance energy index intensity.

6. The multivariable adaptive optimization control method for deacidification of waste incineration flue gas according to claim 1, characterized in that, The variable forgetting factor satisfies the following relationship: in, Indicates the first The forgetting factor changes over time; and These represent the upper and lower limits of the forgetting factor, respectively; This represents the sensitivity adjustment coefficient; Indicates the first The intensity of the combustion disturbance energy index at any given moment; Indicates the inflection point of the disturbance threshold; Represents the safety bias constant; This indicates the current actual monitored concentration of hydrogen chloride at the chimney inlet; This represents the safety baseline concentration constant.

7. The multivariable adaptive optimization control method for desulfurization of waste incineration flue gas according to claim 6, characterized in that, The calculation process of the variable forgetting factor includes: determining whether the calculated variable forgetting factor value is greater than 1, and if it is greater than 1, forcing it to be 1; when the intensity of the combustion disturbance energy index exceeds the inflection point of the disturbance threshold, the variable forgetting factor is reduced to the lower limit of the forgetting factor through the Sigmoid function term.

8. The multivariable adaptive optimization control method for deacidification of waste incineration flue gas according to claim 1, characterized in that, The calculation of the theoretical control quantity based on the updated model parameters includes: using the generalized minimum variance control law, with the set value of the hydrogen chloride concentration at the chimney inlet as the target, and based on the updated deacidification reaction process model parameters, calculating the theoretical lime slurry required to maintain the target concentration.

9. The multivariable adaptive optimization control method for desulfurization of waste incineration flue gas according to claim 1, characterized in that, The final control quantity satisfies the following relationship: in, Indicates the first The final control quantity at any given moment; This represents the calculated theoretical control quantity; This indicates the actual flue gas temperature at the inlet of the deacidification tower; This indicates the center value of the optimal deacidification reaction temperature window; This indicates the actual humidity at the inlet of the deacidification tower; Indicates the basic humidity constant; This represents the impedance compensation gain coefficient.

10. The multivariable adaptive optimization control method for desulfurization of waste incineration flue gas according to claim 9, characterized in that, After obtaining the final control quantity, the method further includes: converting the final control quantity into a frequency command for the rotary spray pump or an opening command for the regulating valve, and sending it to the field actuator to counteract the effects of combustion-side disturbances and environmental impedance on the deacidification efficiency by changing the flow rate of lime slurry injected into the deacidification tower.

Images