Method, device, medium and program product for flue gas desulfurization and denitrification control

Abstract

The present disclosure provides a flue gas desulfurization and denitrification control method, device, medium and program product, wherein the flue gas desulfurization and denitrification control method comprises: obtaining flue gas parameters of a flue gas desulfurization and denitrification system, obtaining second flue gas emission data of the system based on the credibility score of CEMS data, CEMS data and output data of a soft measurement model; obtaining the lag time of the flue gas desulfurization and denitrification system based on the cross-correlation method, historical reagent dosage, first flue gas inlet data and second flue gas emission data; obtaining second flue gas inlet data based on the lag time, the second flue gas inlet data comprising the sulfur and nitrogen content in the system inlet flue gas in a future time window; and controlling the reagent dosage based on the lag time and the second flue gas inlet data. In this way, the reagent dosage can be precisely controlled, the sulfur and nitrogen content in the flue gas discharged from the system can reach the standard, and the reagent can be saved, thereby reducing the cost.

Description

Technical Field

[0001] This disclosure relates to the field of artificial intelligence technology, specifically to a control method, equipment, medium, and program product for flue gas desulfurization and denitrification. Background Technology

[0002] Flue gas desulfurization and denitrification are core components of industrial waste gas treatment, directly determining whether a plant can consistently achieve ultra-low emission standards and its operating costs. Existing flue gas desulfurization and denitrification control methods are mostly based on PID control (Proportional-Integral-Derivative). However, this method has the following problems:

[0003] Firstly, the flue gas desulfurization and denitrification process exhibits a triple lag, specifically: detection lag (CEMS, a continuous emission monitoring system, is an online monitoring system for flue gas, but it cannot output monitoring results in real time; it requires steps such as extraction, condensation, dust removal, pretreatment, and instrument analysis, with an analysis cycle of approximately 5-15 minutes); transmission lag (after the desulfurization and denitrification agents are injected into the reactor inlet, the flue gas carries the reaction products all the way to the outlet monitoring point, and the flow time is determined by physical distance and flow velocity; the flow velocity is fast when the flue gas flow rate is high and slow when the flow rate is low, with a transmission time of approximately 1-5 minutes); and reaction lag (the desulfurization and denitrification agents need to undergo steps such as dissolution, diffusion, and SCR catalytic reaction, with a reaction time of approximately 2-10 minutes; the reaction is fast at high temperatures and slow at low temperatures). Therefore, the flue gas desulfurization and denitrification process not only has a lag, but the lag time is also dynamically changing, and the PID control method, based on fixed parameters, is difficult to adapt to the dynamically changing lag characteristics.

[0004] Secondly, in actual operation, CEMS may experience problems such as zero drift, range drift, sampling tube blockage, and step jump after calibration. Existing control methods directly use the raw CEMS data as feedback signals. When the CEMS data is distorted, the control system will make incorrect decisions based on the erroneous data, resulting in excessive emissions or serious waste of auxiliary materials.

[0005] As can be seen from the above, the desulfurization and denitrification process of flue gas is complex. Existing technologies use fixed-parameter control methods to precisely control this process. Furthermore, although the parameters of multi-agent collaborative knowledge reasoning based on large language models are not fixed, the large models rely on text statistics to generate answers and do not understand the real physical reaction mechanisms, resulting in accuracy issues and making them difficult to apply to factual reasoning in industrial environments. Therefore, existing technologies cannot precisely control the desulfurization and denitrification process of flue gas. Summary of the Invention

[0006] The technical problem to be solved by this disclosure is to overcome the shortcomings of existing technologies in that they cannot accurately control the desulfurization and denitrification process of flue gas, and to provide a control method, equipment, medium and program product for flue gas desulfurization and denitrification.

[0007] This disclosure solves the above-mentioned technical problems through the following technical solution:

[0008] This disclosure provides a method for controlling flue gas desulfurization and denitrification, including:

[0009] Obtain flue gas parameters of the flue gas desulfurization and denitrification system. The flue gas parameters include first flue gas inlet data and first flue gas emission data. The first flue gas inlet data includes the sulfur and nitrate concentrations in the flue gas at the system inlet. The first flue gas emission data includes CEMS data and output data of the soft measurement model.

[0010] The second flue gas emission data of the system is obtained based on the credibility score of the CEMS data, the CEMS data, and the output data of the soft measurement model.

[0011] The lag time of the flue gas desulfurization and denitrification system is obtained based on the cross-correlation method, historical reagent dosage, first flue gas inlet data, and second flue gas emission data.

[0012] The second flue gas inlet data is obtained based on the lag time, and the second flue gas inlet data includes the sulfur and nitrate concentrations in the system inlet flue gas under a future time window;

[0013] The agent is added based on the lag time and the second flue gas inlet data, and the agent is used to desulfurize and denitrify the flue gas.

[0014] Optionally, the steps preceding the acquisition of the second flue gas emission data of the system based on the confidence score of the CEMS data, the CEMS data, and the output data of the soft sensor model include:

[0015] A credibility score is obtained based on at least one of the following: data continuity score, material balance consistency score, soft measurement cross-validation score, and calibration time decay score.

[0016] The material balance consistency score is obtained based on the first flue gas inlet data, reagent dosing data, and CEMS data, and / or the soft measurement cross-validation score is obtained based on the CEMS data and the output data of the soft measurement model;

[0017] Optionally, the step of obtaining the second flue gas emission data of the system based on the confidence score of the CEMS data, the CEMS data, and the output data of the soft sensor model includes:

[0018] The CEMS weight and soft measurement weight are obtained based on the credibility score, the first credibility threshold, and the second credibility threshold.

[0019] The second flue gas emission data is obtained based on the CEMS weights, the soft measurement weights, the CEMS data, and the output data of the soft measurement model.

[0020] Optionally, the step of obtaining the lag time of the flue gas desulfurization and denitrification system based on cross-correlation method, historical reagent dosage, first flue gas inlet data, and second flue gas emission data includes:

[0021] Obtain the confidence level of the cross-correlation method;

[0022] The lag time of the flue gas desulfurization and denitrification system is obtained based on the confidence level of the cross-correlation method.

[0023] Optionally, the step of obtaining the second flue gas inlet data based on the lag time includes:

[0024] The second flue gas inlet data is obtained based on time-series data under historical time windows, lag duration, LSTM and self-attention mechanism. The time-series data includes flue gas parameters and at least one of the following: ambient temperature, gas consumption, and production load.

[0025] Optionally, the step of obtaining the second flue gas parameters based on time-series data under a historical time window, lag duration, LSTM, and self-attention mechanism includes:

[0026] Time series prediction data is obtained based on LSTM and self-attention mechanism;

[0027] The event disturbance curve is obtained based on the event signal from the production scheduling.

[0028] The second flue gas inlet data is obtained based on the event disturbance curve, the confidence level of the event disturbance curve, and the time-series prediction data.

[0029] Optionally, the step of controlling the agent dosing based on the lag time and the second flue gas inlet data includes:

[0030] The first dosage of the drug is obtained based on a forward feedback model;

[0031] The second dosage of the drug is obtained based on the PID method;

[0032] The target dosage of the agent is obtained based on the first dosage and / or the second dosage;

[0033] Add the target dosage of the agent;

[0034] The first dosage is the output data of the forward feedback model, and the input data of the forward feedback model includes: the lag time, the first flue gas inlet data, the second flue gas inlet data, the second flue gas emission data, and the corresponding historical dosage.

[0035] The formula for obtaining the second dosage of the drug based on the PID method is as follows:

[0036] u_pid(t) = Kp(t)·e(t) + Ki(t)·∫e(τ)dτ + Kd(t)·de(t) / dt;

[0037] The formula for calculating e(t) is:

[0038] e(t) = C_target(t) - C_measured(t-τ_detect);

[0039] Where τ is the lag time, t is the sampling time, C_target is the target flue gas emission data, and C_measured is the second flue gas emission data.

[0040] Optionally, the input data of the prefeedback model further includes: a seasonal compensation coefficient, which is obtained based on a CNN model and a self-attention mechanism, and is used to characterize the seasonal periodic variation of the dosage of the agent.

[0041] Optionally, the input data of the pre-feedback model further includes: catalyst efficiency, which is the output data of the Bayesian state-space model. The input data of the Bayesian state-space model includes: cumulative operating time of the catalyst within the system, cumulative treated flue gas volume, denitrification efficiency trend, temperature history, alkali metal content proxy index of raw material batches, soot blowing cycle and effect data, and flue gas concentration. Concentration data;

[0042] Optionally, the step of adding the target dosage of the agent to the system includes:

[0043] The target dosing mode is selected based on the caking risk of the system and the rate of change of flue gas inlet data. The dosing mode includes continuous dosing mode, high pulse dosing mode and variable speed dosing mode.

[0044] The target dosage of the agent is added to the system according to the target dosing pattern;

[0045] The system's caking risk is based on environmental humidity, the rate of change of silo pressure difference, the duration of continuous dosing of the agent, and the amount of agent added at one time.

[0046] If the caking risk is less than the first risk threshold, the target dosing mode is continuous dosing mode; if the caking risk is greater than the first risk threshold but less than the second risk threshold, the target dosing mode is high-frequency pulse dosing mode; if the flue gas inlet change rate is greater than the first change threshold, the target dosing mode is variable speed dosing mode; if the caking risk is greater than the second risk threshold, an operation and maintenance warning is generated.

[0047] Optionally, the step of obtaining the target dosage of the agent based on the first dosage and the second dosage includes:

[0048] Based on the first dosage, the second dosage, and the target dosage of the agent obtained by the model prediction control layer, the objective function of the model prediction control layer is:

[0049] ;

[0050] in, For the second flue gas emission data concentration, For the target flue gas emission data of flue gas concentration, For the second flue gas emission data concentration, For the target flue gas emission data of flue gas concentration, for Number of escapees The base dosage is obtained based on the first dosage and the second dosage. The target dosage is obtained based on the base dosage and the dosage bias value, which is the dosage bias value.

[0051] Optionally, the target flue gas emission data is obtained based on a dynamic edge-checking strategy, specifically using the following formula:

[0052] Margin(t) = Margin_base × α_confidence × α_stability × α_history;

[0053] Margin_base = Emission limit × 15% (base margin);

[0054] Where Margin(t) is the target flue gas emission data, α_confidence is the prediction confidence adjustment factor, which is obtained based on the confidence of the model prediction control layer, α_stability is the operating condition stability adjustment factor, which is obtained based on the first flue gas inlet data, and α_history is the historical exceedance frequency adjustment factor, which is obtained based on the second flue gas emission data.

[0055] Optionally, the agent includes ammonia, and the control method further includes:

[0056] If the temperature of the ammonia water is lower than the first temperature threshold, the ammonia water is preheated based on the first power.

[0057] If the ammonia water temperature is greater than the second temperature threshold and less than the first temperature threshold, the ammonia water is preheated based on the second power. The step of obtaining the target dosage of the agent based on the first dosage, the second dosage, and the model prediction control layer includes:

[0058] The target dosage of the agent is obtained based on the first dosage, the second dosage, the model prediction control layer, and the first ammonia water temperature compensation coefficient.

[0059] If the ammonia water temperature is greater than the third temperature threshold, the step of obtaining the target dosage of the agent based on the first dosage, the second dosage, and the model prediction control layer includes:

[0060] The target dosage of the reagent is obtained based on the first dosage, the second dosage, the model prediction control layer, and the second ammonia water temperature compensation coefficient.

[0061] Optionally, the control method further includes:

[0062] Obtain the security level of the system;

[0063] If the safety level of the system is level zero, the steps for controlling the addition of the agent include: obtaining the target addition amount of the agent based on the first addition amount, the second addition amount, and the model prediction control layer;

[0064] If the system's security level is Level 1, the steps for controlling the drug dosage include: obtaining the target dosage of the drug based on the first dosage and the second dosage;

[0065] If the safety level of the system is level two, the step of adding the control agent includes: increasing the safety margin of the second flue gas emission data;

[0066] If the system has a safety level of three, the steps for controlling the addition of the agent include: obtaining the target dosage of the agent based on a fixed ratio;

[0067] If the system's security level is level four, the steps for controlling the drug dosing include: prompting that the drug dosing be manually controlled.

[0068] This disclosure also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and used to run on the processor, wherein the processor executes the computer program to implement the aforementioned flue gas desulfurization and denitrification control method.

[0069] This disclosure also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the aforementioned control method for flue gas desulfurization and denitrification.

[0070] This disclosure also provides a computer program product, including a computer program that, when executed by a processor, implements the aforementioned control method for flue gas desulfurization and denitrification.

[0071] Based on common knowledge in the field, the above-mentioned preferred conditions can be combined arbitrarily to obtain various preferred embodiments of this disclosure.

[0072] The positive and progressive effects of this disclosure are as follows: Second flue gas emission data are obtained by using the credibility score of CEMS data. When the credibility score of CEMS data is low, the output data of the soft sensor model can be used to further extrapolate the flue gas emission data. Compared with obtaining flue gas emission data solely through CEMS, the obtained second flue gas emission data is more accurate and has stronger credibility.

[0073] By using the cross-correlation method, the first flue gas inlet data, and the second flue gas emission data to obtain the lag time, the correlation between the inlet data and the emission data can be obtained. This allows for the accurate acquisition of the total system lag time from the time the flue gas enters the system to the time it is discharged from the system and detected by the CEMS. Furthermore, since the accuracy of the second flue gas emission data is more precise than that of the flue gas emission data obtained by traditional methods, the calculation accuracy of the lag time is also further improved.

[0074] Based on the ability to accurately obtain the total system lag time, and based on accurate historical flue gas emission data and its corresponding flue gas inlet data, that is, based on the second flue gas emission data and the corresponding first flue gas inlet data, it is possible to accurately predict the flue gas inlet data under the future window, that is, the second flue gas inlet data.

[0075] By accurately predicting the second flue gas inlet data, i.e. the sulfur and nitrate concentrations in the system inlet flue gas under the future time window, the addition of reagents can be precisely controlled, so that the sulfur and nitrate concentrations in the flue gas discharged from the system meet the standards, and reagents are saved and costs are reduced. Attached Figure Description

[0076] Figure 1 A flowchart of a control method for flue gas desulfurization and denitrification provided as an exemplary embodiment of this disclosure;

[0077] Figure 2 This is a schematic diagram of an electronic device provided as an exemplary embodiment of the present disclosure. Detailed Implementation

[0078] The present disclosure is further illustrated below by way of embodiments, but the present disclosure is not limited to the scope of the embodiments described herein.

[0079] The prefixes such as "first" and "second" used in this disclosure are merely for distinguishing different descriptive objects and do not limit the position, order, priority, quantity, or content of the described objects. The use of ordinal numbers and other prefixes used to distinguish descriptive objects in this disclosure does not constitute a limitation on the described objects. The description of the described objects is given in the claims or the context of the embodiments, and should not be construed as an unnecessary limitation. Furthermore, in the description of this embodiment, unless otherwise stated, "multiple" means two or more.

[0080] Figure 1 This is a flowchart of a control method for flue gas desulfurization and denitrification provided as an exemplary embodiment of the present disclosure. This method can be executed by a flue gas desulfurization and denitrification system. This system can monitor or acquire core operating data such as flue gas parameters, operating condition parameters, auxiliary material parameters, auxiliary material status parameters, environmental parameters, and equipment parameters. In addition, the system can also be used to perform relevant calculations or data storage for this method, and based on the final calculation results, execute corresponding control steps, specifically including but not limited to agent preheating, controlling the type, amount, time, and mode of agent addition, etc.

[0081] The method includes the following steps:

[0082] S1. Obtain the flue gas parameters of the flue gas desulfurization and denitrification system.

[0083] The flue gas parameters include the first flue gas inlet data and the first flue gas emission data. The first flue gas inlet data includes the sulfur and nitrate concentrations in the flue gas at the system inlet, and the first flue gas emission data includes CEMS data and the output data of the soft measurement model.

[0084] S2. The second flue gas emission data of the system is obtained by using CEMS data, CEMS data and soft measurement model output data based on the credibility score of CEMS data.

[0085] S3. Based on the cross-correlation method, historical reagent dosage, first flue gas inlet data, and second flue gas emission data, obtain the lag time of the flue gas desulfurization and denitrification system.

[0086] S4. Obtain the second flue gas inlet data based on the lag time. The second flue gas inlet data includes the sulfur and nitrate concentrations in the system inlet flue gas under the future time window.

[0087] S5. The dosage of the reagent is controlled based on the lag time and the second flue gas inlet data. The reagent is used to desulfurize and denitrify the flue gas.

[0088] The first flue gas inlet data can be obtained through hardware monitoring or calculated using a mathematical model based on relevant factory data. The CEMS data is the sulfur and nitrate concentrations of the flue gas after chemical treatment and discharge from the system, as monitored by the continuous emission monitoring system. The soft measurement model is the sulfur and nitrate concentrations of the flue gas after chemical treatment and discharge from the system calculated by the mathematical model. When the reliability of the CEMS data is low, the output data of the soft measurement model can be combined to calculate the flue gas emission data, thereby obtaining more accurate second flue gas emission data.

[0089] This embodiment of the disclosure obtains second flue gas emission data through the credibility score of CEMS data. When the credibility score of CEMS data is low, the output data of the soft measurement model can be used to further calculate the flue gas emission data. Compared with obtaining flue gas emission data solely through CEMS, the obtained second flue gas emission data is more accurate and has stronger credibility.

[0090] By using the cross-correlation method, the first flue gas inlet data, and the second flue gas emission data to obtain the lag time, the correlation between the inlet data and the emission data can be obtained. This allows for the accurate acquisition of the total system lag time from the time the flue gas enters the system to the time it is discharged from the system and detected by the CEMS. Furthermore, since the accuracy of the second flue gas emission data is more precise than that of the flue gas emission data obtained by traditional methods, the calculation accuracy of the lag time is also further improved.

[0091] Based on the ability to accurately obtain the total system lag time, and based on accurate historical flue gas emission data and its corresponding flue gas inlet data, that is, based on the second flue gas emission data and the corresponding first flue gas inlet data, it is possible to accurately predict the flue gas inlet data under the future window, that is, the second flue gas inlet data.

[0092] By accurately predicting the second flue gas inlet data, i.e. the sulfur and nitrate concentrations in the system inlet flue gas under the future time window, the addition of reagents can be precisely controlled, so that the sulfur and nitrate concentrations in the flue gas discharged from the system meet the standards, and reagents are saved and costs are reduced.

[0093] In an alternative implementation, the credibility score of CEMS data can be obtained based on at least one of the following scores: data continuity score, material balance consistency score, soft measurement cross-validation score, and calibration time decay score.

[0094] Among them, the data continuity score is based on the continuous acquisition of CEMS data over a past time window; the material balance consistency score is based on the acquisition of first flue gas inlet data, reagent dosing data, and CEMS data, the principle of which is the inlet flue gas... Total amount minus the exhaust gas The total amount should be approximately equal to the desulfurizing agent consumption multiplied by the theoretical desulfurization efficiency. If the deviation exceeds the threshold, the CEMS data is deemed unreliable, and a lower material balance consistency score can be set. The calibration time decay score can be set according to practical conditions. Since the measurement accuracy of the CEMS may gradually deteriorate over time after calibration, the calibration time decay score can be obtained based on the usage time of the CEMS after calibration.

[0095] In an optional implementation, step S2 specifically includes:

[0096] The CEMS weight and soft measurement weight are obtained based on the credibility score, the first credibility threshold, and the second credibility threshold; the second flue gas emission data are obtained based on the CEMS weight, the soft measurement weight, the CEMS data, and the output data of the soft measurement model.

[0097] For example, the first confidence threshold can be set to 0.9, and the second confidence threshold can be set to 0.7. When the confidence score α(t) > 0.9, CEMS data is used normally, and the CEMS weight can be 1, while the soft measurement weight can be 0. When 0.7 < α(t) ≤ 0.9, the CEMS data and the soft measurement model output are fused together according to confidence weight, and the CEMS weight can be 0.5, while the soft measurement weight can be 0.5. When α(t) ≤ 0.7, the output data of the soft measurement model is switched to be used as the second flue gas emission data, and the CEMS weight can be 0, while the soft measurement weight can be 1.

[0098] In an optional implementation, step S3 may include:

[0099] Obtain the confidence level of the cross-correlation method; obtain the hysteresis of the flue gas desulfurization and denitrification system based on the confidence level of the cross-correlation method.

[0100] Specifically, cross-correlation analysis can be used as the main channel, with the reagent dosage change sequence Δu as the excitation signal and the outlet concentration change sequence (i.e., the second flue gas emission data) Δy as the response signal. The total system lag time can be estimated in real time through sliding window cross-correlation analysis, as shown in the following formula:

[0101] τ_total(t) = argmax_τ { Corr[Δu(t-τ), Δy(t)]};

[0102] Where τ represents the lag time and t represents the sampling time. Since the reagent addition point is at the flue gas inlet, by matching the correlation between the reagent addition amount and the second flue gas emission data, the reagent addition amount with the highest correlation coefficient is identified as matching the second flue gas emission data. Therefore, the total lag time of the current system, i.e., τ_total(t), can be obtained by using the time point of reagent addition and the time point of second flue gas emission data collection. The search range of the cross-correlation analysis method can be 5-35 minutes, with a sliding calculation update every 5 minutes.

[0103] Based on the main channel, a physical model channel and an empirical model channel can be added. The two auxiliary channels are selected appropriately according to the confidence level of the cross-correlation method. Since the matching relationship between the agent dosage and the second flue gas emission data is obtained based on the level of the cross-correlation coefficient, when the cross-correlation coefficient is low, the confidence level of the main channel is low. At this time, the calculation results of the physical model channel and the empirical model channel can be appropriately referenced to more accurately calculate the total lag time of the system.

[0104] The physical model channel calculates the transmission lag of flue gas from the system inlet to the system outlet based on the flue gas flow rate. The specific formula is as follows:

[0105] τ_transport = L / v (pipe length / flue gas velocity);

[0106] Where τ_transport is the system's transmission lag, L is the system's pipe length, and v is the flue gas velocity.

[0107] The empirical model channel can calculate the system's reaction hysteresis based on catalyst temperature and first flue gas inlet data, using the following formula:

[0108] τ_reaction = f(T_catalyst, C_inlet);

[0109] Where τ_reaction is the system's reaction hysteresis, T_catalyst is the catalyst temperature, and C_inlet is the first flue gas inlet data.

[0110] In an optional implementation, step S4 may include:

[0111] The second flue gas inlet data is obtained based on time-series data under historical time windows, lag duration, LSTM (Long Short-Term Memory) network and self-attention mechanism. The time-series data includes flue gas parameters and at least one of the following: ambient temperature, gas consumption, and production load.

[0112] For example, if the historical time window is the past two hours, the flue gas parameters are the first flue gas inlet data, the second flue gas emission data, and the flue gas flow rate of the system in the past two hours. By constructing an LSTM + Attention (self-attention mechanism) time series prediction channel, the trend of sulfur and nitrate concentration changes and confidence intervals in the flue gas at the system inlet under a future time window can be predicted based on the aforementioned time series data and lag duration. In other words, the second flue gas inlet data can be obtained directly from the time series prediction data, or it can be obtained in combination with other methods.

[0113] In an optional implementation, step S4 may specifically include:

[0114] S41. Obtain time series prediction data based on LSTM and self-attention mechanism.

[0115] The time-series forecast data refers to the trend and confidence interval of sulfur and nitrate concentration changes in the flue gas at the system inlet within a future time window.

[0116] S42. Obtain the corresponding event disturbance curve based on the event signal of production scheduling.

[0117] Event signals can specifically include: blast furnace maintenance plans, coke oven shift change times, raw material batch switching notifications, etc. Then, based on the type of event, typical disturbance curves of similar historical events in the event response plan database can be matched, i.e., the aforementioned event disturbance curves. Event signals can be received or acquired from relevant production equipment.

[0118] S43. Obtain the second flue gas inlet data based on the event disturbance curve, the confidence level of the event disturbance curve, and time-series prediction data. The specific formula is as follows:

[0119] ;

[0120] in, This is the data for the second flue gas inlet. The flue gas concentration is predicted based on time-series prediction data. The event confidence level is the confidence level of the event-induced disturbance to the flue gas concentration, predicted based on the event disturbance curve. When no event is triggered, the prediction can be reduced to LSTM + Attention. The second flue gas inlet data can include inlet data for the next 1-5 control cycles. and Concentration predictions and confidence intervals.

[0121] Production events such as blast furnace maintenance, coke oven shift changes, and raw material batch switching can cause a sudden change in the composition of the inlet flue gas. (The concentration can fluctuate drastically within the range of 500-2000 mg / m³). Traditional control systems can only passively adjust after the outlet concentration has deviated significantly, resulting in a severely delayed response. This embodiment, however, combines... and Predicting flue gas concentration allows for accurate forecasting of the disturbances caused by events to flue gas concentration, enabling timely adjustment of reagent dosage.

[0122] In an optional implementation, step S5 may include:

[0123] S51. Obtain the first dosage of the agent based on the forward feedback model (FFNN).

[0124] S52. Obtain the second dosage of the agent based on the PID method.

[0125] S53. Obtain the target dosage of the agent based on the first dosage and the second dosage.

[0126] S54. Add the target dosage of the agent.

[0127] The first dosage is the output data of the forward feedback model. The input data of the forward feedback model includes: lag time, first flue gas inlet data, second flue gas inlet data, second flue gas emission data, and the corresponding historical dosage.

[0128] The formula for obtaining the second dosage of the drug based on the PID method is:

[0129] u_pid(t) = Kp(t)·e(t) + Ki(t)·∫e(τ)dτ + Kd(t)·de(t) / dt;

[0130] The formula for calculating e(t) is:

[0131] e(t) = C_target(t) - C_measured(t-τ_detect);

[0132] Where τ is the lag time, t is the sampling time, C_target is the target flue gas emission data, and C_measured is the second flue gas emission data.

[0133] In summary, the second flue gas inlet data u(t) = u_ffnn(t) + u_pid(t), where u_ffnn(t) is the first dosage calculated based on FFNN, and u_pid(t) is the second dosage calculated based on PID.

[0134] In an optional implementation, the PID parameters can be dynamically adjusted according to the current operating conditions (such as flue gas flow rate, sulfur and nitrate concentration, and whether it is a sudden transition period): when the operating conditions change abruptly, Kp / Ki is reduced and Kd is increased to reduce overshoot and enhance the ability to resist disturbances; when the sulfur and nitrate concentration of the flue gas is high, Kp is increased to speed up the response; when the flue gas flow rate is low, Kp is reduced to avoid oscillation.

[0135] In an alternative implementation, the input data for the forward feedback model may also include seasonal compensation coefficients.

[0136] The seasonal compensation coefficient is obtained based on a CNN model and Attention (self-attention mechanism). The seasonal compensation coefficient is used to characterize the periodic changes in the dosage of the drug with the season.

[0137] Because the extreme low temperatures in winter and the temperature difference between summer and winter in regions like Northeast China can reach 50°C, this directly affects key parameters such as the evaporation efficiency of ammonia in the reagent, the activity window of the catalyst, and the consumption of coal gas. These key parameters drift slowly with the seasons, and fixed-parameter control methods cannot adapt to them. However, the temperature in these regions changes cyclically with the seasons. Therefore, by matching the current seasonal conditions with the approximate temperature of the current region, the dosage can be adjusted according to the temperature changes. This allows for the acquisition of the corresponding seasonal compensation coefficient based on the event situation, thus meeting the requirements of the cyclical seasonal changes.

[0138] In an alternative implementation, the input data for the forward feedback model may also include catalyst efficiency.

[0139] Catalyst efficiency can be calculated based on a Bayesian state-space model. The catalyst efficiency is the output data of the Bayesian state-space model, and the input data includes: cumulative catalyst operating time within the system, cumulative flue gas volume treated, denitrification efficiency trend, historical temperature, alkali metal content proxy index of raw material batches, soot blowing cycle and effect data, and the concentration of alkali metals in the flue gas. Concentration data.

[0140] The system can establish a logic for distinguishing three deactivation modes of the catalyst: slow and monotonous decrease in efficiency (stable slope), which is determined to be thermal aging (irreversible), and the catalyst replacement time is planned according to the predicted curve; sudden and accelerated decrease in efficiency (abrupt slope), which is determined to be chemical poisoning (partially reversible), triggering the inspection of alkali metal content in the raw materials and the evaluation of acid washing and regeneration; and efficiency can be recovered by soot blowing after a sudden drop, which is determined to be physical blockage (reversible), automatically shortening the soot blowing cycle and checking whether the ammonia dosage is too high.

[0141] In an optional implementation, the calculation of the target dosage, based on the first and second dosages, can be further adjusted based on the Model Predictive Control (MPC) layer, whose objective function is:

[0142] ;

[0143] in, For the second flue gas emission data concentration, For the target flue gas emission data of flue gas concentration, For the second flue gas emission data concentration, For the target flue gas emission data of flue gas concentration, for Number of escapees The base dosage is obtained based on the first dosage and the second dosage. The target dosage is obtained based on the base dosage and the dosage bias value, which is the dosage bias value. The weight can be set or changed according to practical needs.

[0144] Using the above objective function, we can... Calculations are performed to further refine the baseline dosage obtained from the first and second dosage amounts, making the target dosage more accurate. , The three emission indicators of ammonia slip are interrelated. Adjusting one indicator will inevitably affect the others. Traditional PID control uses a single loop and cannot handle multi-variable coupling. However, this embodiment combines FFNN, PID and MPC methods to control the dosage, which can meet the multi-variable coupling requirements.

[0145] In an optional implementation, the target flue gas emission data is obtained based on a dynamic edge-checking strategy, specifically using the following formula:

[0146] Margin(t) = Margin_base × α_confidence × α_stability × α_history;

[0147] Margin_base = Emission limit × 15% (base margin);

[0148] Wherein, Margin(t) represents the target flue gas emission data, and α_confidence is the prediction confidence adjustment factor, which is obtained based on the confidence level of the model predicting the control layer. For example, when the confidence level is >0.9, α_confidence = 0.5 (margin reduction, aggressive saving); when the confidence level is 0.7-0.9, α_confidence = 1.0 (standard margin); and when the confidence level is <0.7, α_confidence = 2.0 (margin expansion, conservative control).

[0149] α_stability is the operating condition stability adjustment factor, which is obtained based on the first flue gas inlet data. For example, when the coefficient of variation (CV) of the inlet concentration in the past hour is less than 10%, α_stability = 0.8; when CV is 10%-30%, α_stability = 1.0; and when CV is greater than 30%, α = 1.5.

[0150] α_history is a historical exceedance frequency adjustment factor, which is obtained based on the second flue gas emission data; for example, if there is no exceedance in the past 24 hours, α_history=0.9; if there is exceedance in the past 24 hours, α_history=1.5.

[0151] In one optional implementation, the agent includes ammonia, and the control method further includes:

[0152] If the temperature of the ammonia water is lower than the first temperature threshold, the ammonia water is preheated based on the first power.

[0153] If the ammonia water temperature is greater than the second temperature threshold and less than the first temperature threshold, the ammonia water is preheated based on the second power. The aforementioned steps for obtaining the target dosage of the agent based on the first dosage, the second dosage, and the model prediction control layer include:

[0154] The target dosage of the agent is obtained based on the first dosage, the second dosage, the model prediction control layer, and the first ammonia water temperature compensation coefficient.

[0155] If the ammonia water temperature is greater than the third temperature threshold, the aforementioned steps for obtaining the target dosage of the agent based on the first dosage, the second dosage, and the model prediction control layer include:

[0156] The target dosage of the reagent is obtained based on the first dosage, the second dosage, the model prediction control layer, and the second ammonia water temperature compensation coefficient.

[0157] For example, the first temperature threshold is 15℃, the second temperature threshold is 5℃, and the third temperature threshold is 30℃. When the ammonia water temperature T_nh3 ≥ 15℃, preheating is turned off and normal injection is performed; when 5℃ ≤ T_nh3 < 15℃, preheating is performed at low power, with the first power P = P_max × (15 - T_nh3) / 15; when T_nh3 < 5℃, the second power is full power, and the first ammonia water compensation coefficient is 1 + 0.03×(5 - T_nh3); when T_nh3 > 30℃ (in summer), the ammonia water dosage is reduced, and the second ammonia water temperature compensation coefficient is 1 - 0.02×(T_nh3 - 30).

[0158] In an optional implementation, step S54 includes:

[0159] The target dosing mode is selected based on the system's caking risk and the flue gas inlet data change rate. The dosing modes include continuous dosing mode, high pulse dosing mode, and variable speed dosing mode.

[0160] The drug dosing system is configured to deliver the target dosage amount according to the target dosing pattern.

[0161] The system's caking risk is determined based on ambient humidity, the rate of change of pressure differential in the silo, the duration of continuous dosing of the reagent, and the amount of reagent added at one time.

[0162] If the risk of caking is less than the first risk threshold, the target dosing mode is continuous dosing; if the risk of caking is greater than the first risk threshold but less than the second risk threshold, the target dosing mode is high-frequency pulse dosing; if the flue gas inlet data change rate is greater than the first change threshold, the target dosing mode is variable speed dosing; if the risk of caking is greater than the second risk threshold, an operation and maintenance warning is generated.

[0163] Since the system also includes a silo, when the risk of raw material caking inside the silo reaches a certain level, the dosing mode needs to be changed. For example: continuous dosing mode, which is used by default when the operating conditions are stable, with the dosing pump running at a constant speed; high-frequency pulse dosing mode, which automatically switches when R_caking > 0.7, with short-cycle (30-60 seconds) pulse dosing, thin material layer and continuous agitation, so that liquid bridges are broken up before they can form; variable speed dosing mode, which is used when the inlet concentration fluctuates greatly, with the dosing rate adjusted in real time according to the inlet concentration; when R_caking > 0.9, an operation and maintenance alarm is triggered, and a soft measurement model of material layer thickness H_layer≈ ∫(dosing amount - consumption amount)dt / (silo cross-sectional area × bulk density) is established, supplemented by silo pressure difference verification.

[0164] In an optional implementation, the method further includes:

[0165] Obtain the system's security level;

[0166] If the system's safety level is level zero (normal), step S5 includes: obtaining the target dosage of the agent based on the first dosage, the second dosage, and the model prediction control layer;

[0167] If the system's security level is Level 1 (minor anomaly), step S5 includes: obtaining the target dosage of the agent based on the first dosage and the second dosage;

[0168] If the system's safety level is Level 2 (sensor failure), step S5 includes: increasing the safety margin of the second flue gas emission data;

[0169] If the system's security level is Level 3 (critical fault), step S5 includes: obtaining the target dosage of the agent based on a fixed ratio;

[0170] If the system's security level is Level 4 (emergency), step S5 includes: prompting for manual control of the drug dosing.

[0171] The system can automatically switch between different safety levels based on the system status, ensuring that flue gas emissions meet standards under any fault conditions.

[0172] Figure 2 This is a schematic diagram of the structure of an electronic device according to an example embodiment of the present disclosure. The electronic device includes a memory, a processor, and a computer program stored in the memory and used to run on the processor. When the processor executes the computer program, it implements the flue gas desulfurization and denitrification control method described in any of the above embodiments. Figure 2 The electronic device 20 shown is merely an example and should not impose any limitation on the functionality and scope of use of the embodiments disclosed herein.

[0173] like Figure 2 As shown, the electronic device 20 can be manifested as a general-purpose computing device, such as a server device. The components of the electronic device 20 may include, but are not limited to: at least one processor 21, at least one memory 22, and a bus 23 connecting different system components (including memory 22 and processor 21).

[0174] Bus 23 includes a data bus, an address bus, and a control bus.

[0175] The memory 22 may include volatile memory, such as random access memory (RAM) 221 and / or cache memory 222, and may further include read-only memory (ROM) 223.

[0176] The memory 22 may also include a program tool 225 (or utility) having a set (at least one) program module 224, such program module 224 including but not limited to: an operating system, one or more application programs, other program modules and program data, each or some combination of these examples may include an implementation of a network environment.

[0177] The processor 21 executes various functional applications and data processing by running computer programs stored in the memory 22, such as the flue gas desulfurization and denitrification control method provided in any of the above embodiments.

[0178] Electronic device 20 can also communicate with one or more external devices 24 (e.g., keyboard, pointing device, etc.). This communication can be performed via input / output (I / O) interface 25. Furthermore, electronic device 20 can also communicate with one or more networks (e.g., local area network (LAN), wide area network (WAN), and / or public network, such as the Internet) via network adapter 26. As shown, network adapter 26 communicates with other modules of electronic device 20 via bus 23. It should be understood that, although... Figure 2 Not shown, other hardware and / or software modules may be used in conjunction with electronic device 20, including but not limited to: microcode, device drivers, redundant processors, external disk drive arrays, RAID (disk array) systems, tape drives, and data backup storage systems.

[0179] It should be noted that although several units / modules or sub-units / modules of the electronic device have been mentioned in the detailed description above, this division is merely exemplary and not mandatory. In fact, according to embodiments of this disclosure, the features and functions of two or more units / modules described above can be embodied in one unit / module. Conversely, the features and functions of one unit / module described above can be further divided and embodied by multiple units / modules.

[0180] This disclosure also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the flue gas desulfurization and denitrification control method provided in any of the above embodiments.

[0181] The readable storage medium may be more specifically adopted, including but not limited to: portable disk, hard disk, random access memory, read-only memory, erasable programmable read-only memory, optical storage device, magnetic storage device, or any suitable combination thereof.

[0182] This disclosure also provides a computer program product, including a computer program that, when executed by a processor, implements the flue gas desulfurization and denitrification control method described in any of the above embodiments.

[0183] The program code for executing the computer program product of this disclosure can be written in any combination of one or more programming languages, and the program code can be executed entirely on a user device, partially on a user device, as a stand-alone software package, partially on a user device and partially on a remote device, or entirely on a remote device.

[0184] While specific embodiments of this disclosure have been described above, those skilled in the art should understand that these are merely illustrative examples, and the scope of protection of this disclosure is defined by the appended claims. Those skilled in the art can make various changes or modifications to these embodiments without departing from the principles and essence of this disclosure, but all such changes and modifications fall within the scope of protection of this disclosure.

Claims

1. A method for controlling flue gas desulfurization and denitrification, characterized in that, include: Obtain flue gas parameters of the flue gas desulfurization and denitrification system. The flue gas parameters include first flue gas inlet data and first flue gas emission data. The first flue gas inlet data includes the sulfur and nitrate concentrations in the flue gas at the system inlet. The first flue gas emission data includes CEMS data and output data of the soft measurement model. The second flue gas emission data of the system is obtained based on the credibility score of the CEMS data, the CEMS data, and the output data of the soft measurement model. The lag time of the flue gas desulfurization and denitrification system is obtained based on the cross-correlation method, historical reagent dosage, first flue gas inlet data, and second flue gas emission data. The second flue gas inlet data is obtained based on the lag time, and the second flue gas inlet data includes the sulfur and nitrate concentrations in the system inlet flue gas under a future time window; The agent is added based on the lag time and the second flue gas inlet data, and the agent is used to desulfurize and denitrify the flue gas.

2. The control method as described in claim 1, characterized in that, The steps preceding the acquisition of the second flue gas emission data of the system based on the credibility score of the CEMS data, the CEMS data, and the output data of the soft sensor model include: A credibility score is obtained based on at least one of the following: data continuity score, material balance consistency score, soft measurement cross-validation score, and calibration time decay score. The material balance consistency score is obtained based on the first flue gas inlet data, reagent dosing data, and CEMS data, and / or the soft measurement cross-validation score is obtained based on the CEMS data and the output data of the soft measurement model; And / or, The step of obtaining the second flue gas emission data of the system based on the credibility score of the CEMS data, the CEMS data, and the output data of the soft sensor model includes: The CEMS weight and soft measurement weight are obtained based on the credibility score, the first credibility threshold, and the second credibility threshold. The second flue gas emission data is obtained based on the CEMS weights, the soft measurement weights, the CEMS data, and the output data of the soft measurement model.

3. The control method as described in claim 1, characterized in that, The step of obtaining the lag time of the flue gas desulfurization and denitrification system based on cross-correlation method, historical reagent dosage, first flue gas inlet data, and second flue gas emission data includes: Obtain the confidence level of the cross-correlation method; The lag time of the flue gas desulfurization and denitrification system is obtained based on the confidence level of the cross-correlation method. And / or, The step of obtaining the second flue gas inlet data based on the lag time includes: The second flue gas inlet data is obtained based on time-series data under historical time windows, lag duration, LSTM and self-attention mechanism. The time-series data includes flue gas parameters and at least one of the following: ambient temperature, gas consumption, and production load.

4. The control method as described in claim 3, characterized in that, The steps for obtaining the second flue gas parameters based on time-series data under historical time windows, lag duration, LSTM, and self-attention mechanism include: Time series prediction data is obtained based on LSTM and self-attention mechanism; The event disturbance curve is obtained based on the event signal from the production scheduling. The second flue gas inlet data is obtained based on the event disturbance curve, the confidence level of the event disturbance curve, and the time-series prediction data.

5. The control method as described in claim 1, characterized in that, The step of controlling the agent dosing based on the lag time and the second flue gas inlet data includes: The first dosage of the drug is obtained based on a forward feedback model; The second dosage of the drug is obtained based on the PID method; The target dosage of the agent is obtained based on the first dosage and / or the second dosage; Add the target dosage of the agent; The first dosage is the output data of the forward feedback model, and the input data of the forward feedback model includes: the lag time, the first flue gas inlet data, the second flue gas inlet data, the second flue gas emission data, and the corresponding historical dosage. The formula for obtaining the second dosage of the drug based on the PID method is as follows: u_pid(t) = Kp(t)·e(t) + Ki(t)·∫e(τ)dτ + Kd(t)·de(t) / dt; The formula for calculating e(t) is: e(t) = C_target(t) - C_measured(t-τ_detect); Where τ is the lag time, t is the sampling time, C_target is the target flue gas emission data, and C_measured is the second flue gas emission data.

6. The control method as described in claim 5, characterized in that, The input data of the prefeedback model also includes: a seasonal compensation coefficient, which is obtained based on a CNN model and a self-attention mechanism, and is used to characterize the seasonal periodic variation of the dosage of the drug. And / or, The input data of the prefeedback model also includes: catalyst efficiency, which is the output data of the Bayesian state-space model. The input data of the Bayesian state-space model includes: the cumulative operating time of the catalyst in the system, the cumulative amount of flue gas treated, the trend of denitrification efficiency, the temperature history, the alkali metal content proxy index of raw material batches, the soot blowing cycle and effect data, and the SO3 concentration data in the flue gas. And / or, The step of adding the target dosage of the agent includes: The target dosing mode is selected based on the caking risk of the system and the rate of change of flue gas inlet data. The dosing mode includes continuous dosing mode, high pulse dosing mode and variable speed dosing mode. The target dosage of the agent is added to the system according to the target dosing pattern; The system's caking risk is based on environmental humidity, the rate of change of silo pressure difference, the duration of continuous dosing of the agent, and the amount of agent added at one time. If the caking risk is less than the first risk threshold, the target dosing mode is continuous dosing mode; if the caking risk is greater than the first risk threshold but less than the second risk threshold, the target dosing mode is high-frequency pulse dosing mode; if the flue gas inlet data change rate is greater than the first change threshold, the target dosing mode is variable speed dosing mode; if the caking risk is greater than the second risk threshold, an operation and maintenance warning is generated.

7. The control method as described in claim 6, characterized in that, The step of obtaining the target dosage of the agent based on the first dosage and the second dosage includes: Based on the first dosage, the second dosage, and the target dosage of the agent obtained by the model prediction control layer, the objective function of the model prediction control layer is: ； in, For the second flue gas emission data concentration, For the target flue gas emission data of flue gas concentration, For the second flue gas emission data concentration, For the target flue gas emission data of flue gas concentration, for Number of escapees The base dosage is obtained based on the first dosage and the second dosage. The target dosage is obtained based on the base dosage and the dosage bias value, which is the dosage bias value.

8. The control method as described in claim 7, characterized in that, The target flue gas emission data is obtained based on a dynamic edge-checking strategy, specifically using the following formula: Margin(t) = Margin_base × α_confidence × α_stability × α_history; Margin_base = Emission limit × 15% (base margin); Where Margin(t) is the target flue gas emission data, α_confidence is the prediction confidence adjustment factor, which is obtained based on the confidence of the model prediction control layer, α_stability is the operating condition stability adjustment factor, which is obtained based on the first flue gas inlet data, and α_history is the historical exceedance frequency adjustment factor, which is obtained based on the second flue gas emission data. And / or, The reagent includes ammonia water, and the control method further includes: If the temperature of the ammonia water is lower than the first temperature threshold, the ammonia water is preheated based on the first power. If the ammonia water temperature is greater than the second temperature threshold and less than the first temperature threshold, the ammonia water is preheated based on the second power. The step of obtaining the target dosage of the agent based on the first dosage, the second dosage, and the model prediction control layer includes: The target dosage of the agent is obtained based on the first dosage, the second dosage, the model prediction control layer, and the first ammonia water temperature compensation coefficient. If the ammonia water temperature is greater than the third temperature threshold, the step of obtaining the target dosage of the agent based on the first dosage, the second dosage, and the model prediction control layer includes: The target dosage of the reagent is obtained based on the first dosage, the second dosage, the model prediction control layer, and the second ammonia water temperature compensation coefficient.

9. The control method as described in claim 8, characterized in that, The control method further includes: Obtain the security level of the system; If the safety level of the system is level zero, the steps for controlling the addition of the agent include: obtaining the target addition amount of the agent based on the first addition amount, the second addition amount, and the model prediction control layer; If the system's security level is Level 1, the steps for controlling the drug dosage include: obtaining the target dosage of the drug based on the first dosage and the second dosage; If the safety level of the system is level two, the step of adding the control agent includes: increasing the safety margin of the second flue gas emission data; If the system has a safety level of three, the steps for controlling the addition of the agent include: obtaining the target dosage of the agent based on a fixed ratio; If the system's security level is level four, the steps for controlling the drug dosing include: prompting that the drug dosing be manually controlled.

10. An electronic device comprising a memory, a processor, and a computer program stored in the memory and for running on the processor, characterized in that, When the processor executes the computer program, it implements the flue gas desulfurization and denitrification control method according to any one of claims 1 to 9.

11. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the flue gas desulfurization and denitrification control method according to any one of claims 1 to 9.

12. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by the processor, it implements the flue gas desulfurization and denitrification control method as described in any one of claims 1-9.

Images