Intelligent circulating fluidized bed method desulfurization and carbon-pollution deep purification cooperating system

Abstract

The invention relates to the field of energy conservation and environmental protection, in particular to an intelligent circulating fluidized bed method desulfurization and carbon-pollution deep purification system, which constructs a high-reliability and explainable accurate prediction model for the concentration of various pollutants through a modeling method cooperatively driven by knowledge and data in a multi-pollutant generation-removal process, predicts the change of the concentration of the pollutants 90 seconds ahead of time, and improves the efficiency of desulfurization and carbon-pollution deep purification. The average relative percentage error is smaller than 2%, and the problems of control lag and difficulty in real-time accurate optimization caused by data measurement delay are solved; an intelligent regulation and control technology of slaked lime feeder frequency / bed pressure / water spraying amount multi-parameter collaborative optimization-model predictive control double-layer structure coupling is established, and while full-time stable ultra-low emission is achieved, outlet pollutant concentration fluctuation is reduced by more than 50%, outlet temperature fluctuation is within + / -2 DEG C, bed pressure drop fluctuation is within + / -75 Pa, slaked lime consumption is reduced by more than 10%, and full-time stable ultra-low emission is achieved. The water consumption is reduced by 5% or above, CO2 emission is reduced by 6% or above, and the problems of efficient and stable removal and synergistic carbon reduction of multiple pollutants in flue gas under complex working conditions are solved.

Description

Technical Field

[0001] The present invention belongs to the field of energy conservation and environmental protection, and specifically relates to an intelligent circulating fluidized bed desulfurization synergistic carbon-pollution deep purification system. Background Art

[0002] At present, circulating fluidized bed flue gas desulfurization technology has been applied on a large scale and industrialized in key industries such as coal-fired power plants, industrial boilers, steel sintering machines, and sludge incineration. Circulating fluidized bed flue gas desulfurization and carbon-pollution deep purification are complex industrial processes with multiple inputs and multiple outputs, strong coupling, strong nonlinearity, variable parameters, multiple operating conditions, and large-scale variable loads. How to achieve intelligent control of circulating fluidized bed flue gas desulfurization and carbon-pollution deep purification systems faces many challenges, which are mainly manifested as follows: Circulating fluidized bed desulfurization involves a complex process of flow-mass transfer-reaction, and the removal mechanisms of different pollutants are very different. There is a time delay in the measurement of key operating parameters, which leads to serious lags in optimization and control based on instrument data, making it difficult to support real-time and accurate optimization and control; therefore, how to construct a prediction model for key parameters of circulating fluidized bed flue gas desulfurization and carbon-pollution efficient purification process and accurately and quantitatively describe the operating characteristics of the circulating fluidized bed flue gas desulfurization system is a difficult problem that needs to be overcome. The operation control in the circulating fluidized bed desulfurization synergistic carbon-pollutant efficient purification system is often dynamic. For example, frequent changes in load (20% to 110%) and frequent fluctuations in coal quality / ore quality (fluctuations in sulfur content, ash content, etc., fluctuations in the ratio of coal / ore blending, etc.) will cause changes in the system's operating characteristics. The current operation mode based on manual experience and traditional PID control makes it difficult to accurately adjust and optimize the operation of the pollutant removal device in real time when the operating conditions change, resulting in large fluctuations in pollutant emission concentrations, poor control performance of the pollutant removal device, high material and energy consumption of the pollutant removal device, and bed collapse. Therefore, in view of the fluctuations in flue gas treatment volume and pollutant concentration caused by variable load, variable parameters, and variable fuel / raw materials, how to realize the control system from "single device, single pollutant" control to "multiple devices, multiple pollutants" operation optimization and intelligent control, from "standard control" to "high reliability card edge control", significantly reducing the fluctuation range of the system outlet pollutant emission concentration and reducing the operating energy and material consumption is a difficult problem that needs to be broken through.

[0003] The adiabatic saturation temperature difference is the difference between the flue gas temperature at the outlet of the desulfurization tower and the adiabatic saturation temperature of the flue gas. The adiabatic saturation temperature difference plays an important role in the desulfurization efficiency of the desulfurization device and the stable operation of the system. On the one hand, the lower the adiabatic saturation temperature difference, the higher the humidity in the desulfurization tower, which is more conducive to the desulfurization reaction and improves the desulfurization efficiency. On the other hand, if the adiabatic saturation temperature difference is too low, it will cause condensation in the flue gas, corrode the inner wall and separate dust removal equipment. Therefore, how to control the adiabatic saturation temperature difference during operation becomes the key to improving the performance of the flue gas desulfurization process.

[0004] The purpose of flue gas temperature control is to control the adiabatic saturation temperature difference by adjusting the amount of water sprayed at the bottom of the desulfurization tower to ensure the economic and safe operation of the desulfurization system. The adjustment of the amount of water sprayed can be achieved through simple PID control. The key to flue gas temperature control is the online real-time measurement of the adiabatic saturation temperature difference. However, in the actual desulfurization process, the adiabatic saturation temperature difference cannot be measured online, so it can only be calculated according to its definition. To calculate the adiabatic saturation temperature difference, it is necessary to measure the flue gas temperature at the outlet of the desulfurization tower and the adiabatic saturation temperature of the flue gas. The flue gas outlet temperature can be measured online. The key is that there is no equipment that can directly measure the adiabatic saturation temperature in the actual process, so some indirect methods are generally used to obtain the adiabatic saturation temperature.

[0005] In view of the above problems, the present invention proposes an intelligent circulating fluidized bed desulfurization synergistic carbon-pollution deep purification system. Through a modeling method driven by knowledge and data of multiple pollutant generation-removal processes, combined with multi-dimensional analysis of the dynamic correlation characteristics between key parameters and pollutant concentrations, a highly reliable and interpretable multi-section SO 2 / SO 3 / CO 2 / HCl and other pollutant concentrations, overcoming the control lag caused by data measurement delays and the difficulty in real-time accurate optimization; invented a multi-parameter collaborative optimization-model predictive control double-layer structure coupled pollutant removal device intelligent control technology, while ensuring stable ultra-low emissions of multiple pollutants at all times, reducing the fluctuation range of outlet pollutant concentrations and flue gas temperature, and achieving precise control of bed pressure drop and temperature, further reducing slaked lime consumption and process water consumption, and reducing CO 2 The emission reduction solves the problem of efficient and stable removal of multiple pollutants in flue gas and coordinated carbon reduction under complex operating conditions. Summary of the invention

[0006] The present invention proposes an intelligent circulating fluidized bed desulfurization coordinated carbon-pollution deep purification system, the system includes a slaked lime preparation and transportation system, an ash circulation system, a temperature and humidity control system, a flue gas recirculation system, a pre-dust collector, a bag filter, a desulfurization tower and an intelligent control system. It is characterized in that the temperature and humidity control system is provided with a multi-stage humidification activation spray gun, the slaked lime preparation and transportation system is provided with a variable frequency feeder, the ash circulation system is provided with an ash circulation variable frequency feeder, the flue gas recirculation system is provided with a clean flue gas recirculation variable frequency fan, and the intelligent control system includes an online monitoring module, a key parameter prediction module, and a key parameter control module.

[0007] The online monitoring module includes the total coal feed amount, total ore amount, coal quality parameters, mineral composition, sintering process outlet flue gas temperature, wind box temperature, combustion air flow, main exhaust fan flow, flue gas oxygen content parameters, desulfurization tower inlet and outlet SO 2Concentration, flue gas temperature and moisture content of inlet flue gas, feeder frequency, ash circulation feeder frequency, clean flue gas recirculation variable frequency fan frequency, and the number of multi-stage humidification activation spray guns put into use.

[0008] The key parameter prediction module includes the desulfurization tower inlet and outlet CO 2 concentration, SO 2 concentration, SO 3 The generation and removal prediction model of concentration and HCl concentration, the prediction model of flue gas temperature and flue gas pressure at the outlet of the desulfurization tower; the key parameter control module includes a slaked lime variable frequency feeder control model, a bed pressure drop prediction control model, and a multi-stage humidification activation spray gun control model;

[0009] The desulfurization tower inlet and outlet CO 2 concentration, SO 2 concentration, SO 3 The construction of the generation and removal prediction model of concentration and HCl concentration mainly includes the following steps:

[0010] Step S1: Analysis of CO during combustion / sintering 2 、SO 2 、SO 3 The generation principle of HCl acid gas, combined with data analysis methods, was used to screen the factors affecting the inlet CO of the circulating fluidized bed desulfurization tower. 2 、SO 2 、SO 3 , the key parameter of HCl acid gas concentration;

[0011] Step S2: Compensate for the delay between key variables through data processing and construct CO based on LSTM neural network with autoregressive variables 2 、SO 2 、SO 3 The knowledge and data fusion drive model of the acid gas generation concentration of HCl and HF is used to achieve the CO 2 、SO 2 、SO 3 , accurate prediction of HCl acid gas concentration in advance;

[0012] Step S3: Analyze CO 2 、SO 2 、SO 3 , HC acid gas adsorption and removal mechanism with absorbent in flue gas circulating fluidized bed desulfurization tower, establish acid gas adsorption and removal mechanism model, and further 2 、SO 2 、SO 3 The correction parameters were introduced into the acid gas adsorption removal mechanism model of HCl and the PSO algorithm was used to solve it, and a CO2-based parameter identification method was constructed. 2 、SO2 、SO 3 , HCl acid gas adsorption removal model;

[0013] Step S4: LSTM neural network is further used to compensate for the parameter identification model error, and combined with actual operation data, a knowledge and data fusion driven acid gas concentration prediction model with double-layer data correction is established to achieve the CO2 concentration prediction of the desulfurization device outlet. 2 、SO 2 、SO 3 , accurate prediction of HCl acid gas concentration.

[0014] As a preference, the knowledge and data fusion driven desulfurization tower inlet SO 2 The concentration prediction model is expressed as follows:

[0015]

[0016] in: is the predicted value of sulfur dioxide concentration at the desulfurization tower inlet, K r SO caused by boiler / sintering process operating conditions 2 Emission factor, K r =f(T, B, L 1 , L 2 …); T is the flue gas temperature of the furnace or sintering process, B is the coal feed rate, L 1 is the primary air volume, L 2 is the secondary air volume, a fh is the fly ash coefficient, A ar To receive the base ash, S ar Received basic sulfur content, Q net，ar To receive the base low heat, ω CaO ,ω MgO , is the mass fraction of the corresponding substance in the ash, n is the coal consumption per hour, and Q is the flue gas volume at the desulfurization tower inlet.

[0017] As a preference, the knowledge and data fusion driven desulfurization tower inlet SO 3 The concentration prediction model is expressed as follows:

[0018]

[0019] in: is the predicted value of sulfur trioxide concentration at the desulfurization tower entrance, T is the flue gas temperature of the furnace or sintering process, B is the coal feed rate, L 1 is the primary air volume, L 2 is the secondary air volume, is the predicted value of sulfur dioxide concentration at the desulfurization tower inlet, and Q is the flue gas volume at the desulfurization tower inlet.

[0020] Preferably, the desulfurization tower inlet HCl concentration prediction model driven by knowledge and data fusion is expressed as follows:

[0021]

[0022] Where: C HCl,in，yc is the predicted value of hydrogen chloride concentration at the desulfurization tower inlet, K r is the HCl emission coefficient caused by the operating conditions of the boiler / sintering process; T is the flue gas temperature of the furnace or sintering process, B is the coal feed rate, L 1 is the primary air volume, L 2 is the secondary air volume, a fh is the fly ash coefficient, A ar To receive the base ash, Q net，ar To receive the base low heat, ω CaO ,ω MgO , is the mass fraction of the corresponding substance in the ash, n is the coal consumption per hour, and Q is the flue gas volume at the desulfurization tower inlet.

[0023] As a preference, the knowledge and data fusion driven desulfurization tower inlet CO 2 The concentration prediction model is expressed as follows:

[0024]

[0025] in: CO at the desulfurization tower inlet 2 volume, B is the amount of coal fed, Q F is the total air volume entering the combustion process, C ar is the carbon content of the fuel / ore as received, H ar is the received hydrogen content of the fuel / ore, O ar is the received oxygen content of the fuel / ore, V CO is the carbon monoxide volume at the desulfurization tower inlet, β is the fuel characteristic coefficient, is the predicted value of sulfur dioxide concentration at the desulfurization tower inlet, Q is the flue gas volume at the desulfurization tower inlet;

[0026] As a preference, the SO at the outlet of the desulfurization tower based on parameter identification 2 The prediction model of concentration is expressed as follows:

[0027]

[0028] Where: M Ca(OH)2 is the amount of slaked lime fed; Z h is the conversion rate of slaked lime, f(t l , t f …); t f is the time it takes for the desulfurization agent to complete its reaction in the absorption tower; t lThe residence time of the desulfurizer in the absorption tower; SO in the desulfurization tower 2 partial pressure; T is temperature, V is the volume of the desulfurization tower, a is the air leakage coefficient of the desulfurization tower, is the predicted value of sulfur dioxide concentration at the desulfurization tower inlet, Q is the flue gas volume at the desulfurization tower inlet; R is a constant.

[0029] Further based on the above model, t f is the reaction completion time of the desulfurization agent in the absorption tower and t l There are two parameters that need to be identified for the residence time of the desulfurizer in the absorption tower. The parameter identification process can be expressed as an optimization problem, namely:

[0030]

[0031] In the formula, RMSE(t l , t f ) is the model calculation export SO 2 The root mean square error between the predicted and measured concentrations.

[0032] The establishment of the prediction model of the flue gas temperature at the outlet of the desulfurization tower and the adiabatic saturated temperature difference of the desulfurized flue gas mainly includes the following steps:

[0033] Step W1: Analyze the adiabatic cooling and humidification mechanism of flue gas in the flue gas circulating fluidized bed desulfurization process, construct a flue gas humidity prediction mechanism model at the outlet of the desulfurization tower, collect the moisture content of the flue gas at the inlet of the desulfurization tower, the flue gas volume at the inlet of the desulfurization tower, the flue gas temperature at the inlet of the desulfurization tower, the flue gas pressure at the inlet of the desulfurization tower, the water spray volume at the inlet of the desulfurization tower, the flue gas pressure at the outlet of the desulfurization tower, and the flue gas temperature at the outlet of the desulfurization tower, and construct a desulfurization flue gas adiabatic saturated temperature difference prediction model database;

[0034] Step W2: Combine data analysis methods to screen key parameters that affect the humidity of flue gas at the outlet of the desulfurization tower, further analyze the pure delay time from the change of key variables to the humidity of flue gas at the outlet of the desulfurization tower, and compensate for the delay between key variables through data processing;

[0035] Step W3: By introducing correction parameters into the flue gas humidity mechanism model at the outlet of the desulfurization tower and solving it through the PSO algorithm, a parameter-identified desulfurization tower outlet flue gas humidity control mechanism model is constructed;

[0036] Step W4: Further construct a flue gas humidity prediction model at the outlet of the desulfurization tower based on the LSTM neural network with autoregressive variables, which integrates data and humidity control mechanism;

[0037] Step W5: Further analyze the adiabatic saturated temperature control mechanism of desulfurized flue gas, and construct a desulfurized flue gas adiabatic saturated temperature difference prediction model based on the fusion of empirical knowledge and mechanism;

[0038] Step W6: Further combine the flue gas temperature at the outlet of the desulfurization tower, the flue gas pressure data at the inlet of the desulfurization tower and the prediction model of the flue gas humidity at the outlet of the desulfurization tower to construct a desulfurization flue gas adiabatic saturated temperature difference prediction model that integrates the mechanism and operation data correction, thereby realizing the accurate prediction of the adiabatic saturated temperature difference of the desulfurization flue gas under variable load conditions.

[0039] Preferably, the parameter-identified flue gas humidity control mechanism model at the outlet of the desulfurization tower can be expressed as follows:

[0040] H=λ*(G*(1-δ) / (Q*ρ)+H i )

[0041] Among them, G is the amount of water sprayed into the desulfurization tower, Q is the volume flow rate of flue gas, ρ is the density of flue gas, H i is the humidity of flue gas at the inlet of desulfurization tower, δ is the coefficient of residual moisture in desulfurization ash in the total water spraying volume, and λ is the correction coefficient of flue gas humidity at the outlet of desulfurization tower.

[0042] Furthermore, the parameters of δ and λ in the empirical model of flue gas humidity at the outlet of the desulfurization tower are identified. The parameter identification process can be expressed as an optimization problem, namely:

[0043]

[0044] Among them, RMSE (δ, λ) is expressed as the root mean square error between the predicted value and the actual value of the flue gas humidity at the outlet of the desulfurization tower.

[0045] Preferably, the adiabatic saturation temperature difference of the desulfurized flue gas can be expressed by the following formula:

[0046] ΔT s =TT s

[0047] Where, ΔT s is the adiabatic saturation temperature difference of the desulfurized flue gas, T is the flue gas temperature at the outlet of the desulfurization tower, and T s is the adiabatic saturation temperature;

[0048] Furthermore, the adiabatic saturated temperature difference prediction model of desulfurized flue gas based on the fusion of empirical knowledge and mechanism can be expressed by the following formula:

[0050]

[0051] Among them, C is the specific heat of wet flue gas, M is the molecular weight of flue gas, T is the flue gas temperature at the outlet of desulfurization tower, T s is the adiabatic saturation temperature, H is the humidity of the flue gas at T, P f is the flue gas pressure.

[0052] The multi-stage humidification module is not only the module with the highest water consumption in the desulfurization device, but also the most important subsystem to ensure the flue gas temperature at the desulfurization outlet and the adiabatic saturation temperature difference of the desulfurized flue gas. Therefore, when constructing the intelligent control problem of the multi-stage humidification module of the flue gas desulfurization device, the operating water consumption, the flue gas temperature at the desulfurization outlet and the adiabatic saturation temperature difference of the desulfurized flue gas should be considered at the same time. Therefore, the intelligent control process can also be described by the following optimization problem:

[0053] minWater(L 1 ,L 2 ,L 3 ,L 4 =L 1 p 1 +L 2 p 2 +L 3 p 3 +L 4 p 4

[0054]

[0055] Among them, L 1 , L 2 , L 3 , L 4 It is the operating state of the 1st to 4th level humidification nozzle. When the humidification nozzle is turned on, the operating state is 1; when the humidification nozzle is turned off, the operating state is 0. 1 、p 2 、p 3 、p 4 Corresponding water flow rate for the humidifying nozzle. is the predicted value of outlet flue gas temperature, It is the predicted value of adiabatic saturated temperature difference of desulfurized flue gas.

[0056] The slaked lime optimization control module calculates the predicted value of sulfur dioxide, the calculated value of sulfur trioxide, the calculated value of HCl, and the calculated value of carbon dioxide through the established sulfur dioxide concentration prediction model, sulfur trioxide concentration prediction model, HCl concentration prediction model, and carbon dioxide concentration prediction model, and further combines the real-time operating data of the desulfurization tower furnace temperature, the desulfurization tower furnace pressure, and the slaked lime feed amount, and after adjustment by the multi-model prediction optimization control strategy, sends the slaked lime feed amount instruction to the DCS controller through the data communication module, so that the DCS system sends the instruction to the field equipment.

[0057] Preferably, the predicted sulfur dioxide value, calculated sulfur trioxide value, calculated HCl value, calculated carbon dioxide value and the desulfurization tower inlet sulfur dioxide value are used as given instructions, and the desulfurization tower furnace temperature and desulfurization tower furnace pressure are used as feedforward instructions.

[0058] Furthermore, f(X1) is the CO at the desulfurization tower inlet.2 The broken line function of volume conversion into lime mass; f(X2) is the SO at the desulfurization tower inlet 3 The broken line function of the concentration converted into the mass of slaked lime; f(X3) is the SO 2 f(X4) is the broken line function of converting the HCl concentration at the desulfurization tower inlet into the mass of slaked lime; f(X5) is the broken line function of converting the desulfurization tower temperature into the mass of slaked lime; f(X6) is the broken line function of converting the desulfurization tower furnace pressure into the mass of slaked lime.

[0059] As a preferred option, the multi-model prediction optimization control strategy is to set the export SO 2 Under the concentration conditions, the method of minimizing the feeding amount of the slaked lime feeder is selected. At the same time, in the actual operation process, since the frequency of the slaked lime feeder is difficult to achieve fine control, the slaked lime flow can only be adjusted by a specific frequency modulation amplitude. Frequent slaked lime feeder frequency will greatly affect the service life of the equipment; in addition, in order to avoid the outlet SO2 caused by untimely adjustment of the slaked lime feeder frequency, the slaked lime flow rate can be adjusted by adjusting the frequency of the slaked lime feeder. 2 Concentration exceeds the limit, set outlet SO 2 The concentration emission limit is to ensure SO 2 Meet emission standards.

[0060] Furthermore, the frequency switching condition of the slaked lime feeder can be expressed by the following mathematical expression:

[0061]

[0062] S time >S time,set

[0063] In the formula, Based on SO 2 The removal process model calculates the outlet SO when maintaining the existing operation strategy 2 concentration, To set the export SO 2 The upper limit of concentration emission is set to 35mg / m 3 ; S time,set The frequency switching time of the slaked lime feeder can be determined according to the corresponding operating procedures.

[0064] Furthermore, the specific calculation logic of the multi-model prediction optimization control strategy mainly includes the following steps: first, the best operation strategy under the current working conditions is selected according to the optimized operating condition library, and whether the current operation strategy of the slaked lime feeder is consistent with the best strategy is determined. If they are consistent, no adjustment is required. If they are inconsistent, first determine whether the outlet SO 2Whether the concentration exceeds the set emission upper limit. If so, ignore other conditions and switch the feeder frequency immediately to prioritize ensuring that the flue gas meets the emission standards. If the emission standards are met, judge the switching time. If the switching time is greater than the specified time, switch. If the specified switching time has not been reached since the last switch, maintain the original operation strategy.

[0065] The bed pressure drop control module optimization strategy mainly includes the following steps:

[0066] Step 1: Construct a bed pressure drop prediction model based on parameter identification;

[0067] Step 2: Given a target value for bed pressure drop, the key input variables of inlet pressure, flue gas flow, fresh slaked lime particle flow, clean flue gas circulation flow, and ash material circulation flow are discretized;

[0068] Step 3: Obtain the optimal configuration results of the clean flue gas circulation fan regulating valve opening and the ash material circulation regulating valve opening under different working conditions through discretization, build a control rule table, and use it as the basis for judging the implementation of control actions to achieve intelligent control of the clean flue gas circulation fan and ash material circulation;

[0069] Step 4: Further introduce feedback variables into the control rules. By using the idea of ​​feedback correction in model predictive control, the model error is obtained by comparing the predicted bed pressure drop value output by the bed pressure drop prediction model based on parameter identification with the actual measured bed pressure drop value. The error is then fed back to correct the given reference trajectory and input it into the inference module, thereby ensuring the effectiveness of the controller.

[0070] Preferably, the actual absorption tower bed pressure drop is used as the feedback value, and the bed pressure drop control strategy output instruction is used as the material circulation control valve tracking value calculation parameter, and this value is multiplied by the corresponding control valve balance coefficient as the tracking value of the control valve. In order to prevent the return trough valve from frequently and slightly moving, a valve position holding program is set when the fluctuation amplitude is less than 1.

[0071] Furthermore, the continuous material levels of different ash hoppers are averaged, and the actual material level of each ash hopper is divided by the average value as the balancing coefficient of the ash return valve to achieve balanced control of the material level of each ash hopper.

[0072] The beneficial effects of the present invention are mainly as follows:

[0073] (1) This paper proposes an intelligent circulating fluidized bed desulfurization synergistic carbon-pollutant deep purification system. Through a modeling method driven by knowledge and data of multiple pollutant generation-removal processes, combined with a multi-dimensional analysis of the dynamic correlation characteristics between key parameters and pollutant concentrations, a highly reliable and interpretable multi-section SO 2 / SO 3 / CO 2The accurate prediction model of pollutant concentrations such as / HCl can predict pollutant concentration changes 90 seconds in advance, with an average relative percentage error of less than 2%, overcoming the control lag caused by data measurement delay and the difficulty in real-time accurate optimization;

[0074] (2) Invented a multi-parameter collaborative optimization-model predictive control double-layer structure coupled pollutant removal device intelligent control technology, which ensures stable ultra-low emissions of multiple pollutants throughout the entire period, reduces the fluctuation range of outlet pollutant concentration by more than 60%, fluctuates within ±2°C at the outlet, fluctuates within ±75Pa at the bed pressure drop, reduces slaked lime consumption by more than 10%, reduces process water consumption by more than 5%, and reduces CO by more than 5% to a certain extent. 2 The emission reduction solves the problem of efficient and stable removal of multiple pollutants in flue gas and coordinated carbon reduction under complex operating conditions.

[0075] (3) The present invention overcomes the hysteresis of the slaked lime control system and solves the SO caused by the delay of the slaked lime control system when lifting and lowering the load. 2 It can also avoid the problem of excessive emissions, increase in heat loss and cost caused by excessive slaked lime input; at the same time, it can also utilize the flue gas inlet CO 2 The amount of slaked lime can be used to control the feed amount, which reduces CO to a certain extent. 2 emissions. BRIEF DESCRIPTION OF THE DRAWINGS

[0076] Figure 1 It is a schematic diagram of the intelligent circulating fluidized bed desulfurization coordinated carbon-pollution deep purification system;

[0077] Figure 2 Import and export CO for desulfurization tower 2 concentration, SO 2 concentration, SO 3 Concentration, HCl concentration

[0078] Flow chart of building a removal prediction model;

[0079] Figure 3 Build a process for predicting the adiabatic saturated temperature difference of desulfurized flue gas;

[0080] Figure 4 Control strategy diagram for slaked lime optimization control module;

[0081] Figure 5 Optimize strategy for bed pressure drop control module:

[0082] Figure 6 The application effect of the knowledge and data fusion modeling method based on double-layer data correction;

[0083] Figure 7This is the intelligent control effect diagram of the multi-stage humidification module;

[0084] Figure 8 Bed pressure drop control effect diagram. DETAILED DESCRIPTION

[0085] The present invention proposes an intelligent circulating fluidized bed desulfurization coordinated carbon-pollution deep purification system, which is described in detail below with reference to the accompanying drawings and specific examples.

[0086] Figure 1 This is a schematic diagram of an intelligent circulating fluidized bed desulfurization coordinated carbon-pollution deep purification system, which includes a slaked lime preparation and transportation system, an ash circulation system, a temperature and humidity control system, a flue gas recirculation system, a pre-dust collector, a bag filter, a desulfurization tower, and an intelligent control system. The temperature and humidity control system is provided with a multi-stage humidification activation spray gun, the slaked lime preparation and transportation system is provided with a variable frequency feeder, the ash circulation system is provided with an ash circulation variable frequency feeder, the flue gas recirculation system is provided with a clean flue gas recirculation variable frequency fan, and the intelligent control system includes an online monitoring module, a key parameter prediction module, and a key parameter control module.

[0087] The DCS system is connected to the circulating fluidized bed desulfurization synergistic carbon-pollution deep purification system and the intelligent control system. The intelligent control system includes: data communication module, SO 2 / SO 3 / CO 2 / HCl multiple pollutant concentration accurate prediction module, SO 2 / SO 3 / CO 2 / HCl multiple pollutant quantity calculation module, desulfurization tower outlet flue gas temperature and flue gas pressure prediction module, slaked lime variable frequency feeder control module, bed pressure drop prediction control module, multi-stage humidification activation spray gun control module.

[0088] Among them, the data communication module exchanges data with the DCS system, the precise prediction module of multiple pollutant concentrations connects the data communication module and the slaked lime variable frequency feeder control module, and the slaked lime variable frequency feeder control module connects the communication module and SO 2 / SO 3 / CO 2 / HCl multiple pollutant quantity calculation module, desulfurization tower outlet flue gas temperature and flue gas pressure prediction module connected to the data communication module and the bed pressure drop prediction control module, multi-stage humidification activation spray gun investment quantity prediction module connected to the data communication module and the multi-stage humidification activation spray gun control module.

[0089] The online monitoring module includes the total coal feed amount, total ore amount, coal quality parameters, mineral composition, sintering process outlet flue gas temperature, wind box temperature, combustion air flow, main exhaust fan flow, flue gas oxygen content parameters, desulfurization tower inlet and outlet SO 2 Concentration, flue gas temperature and moisture content of inlet flue gas, feeder frequency, ash circulation feeder frequency, clean flue gas recirculation variable frequency fan frequency, and the number of multi-stage humidification activation spray guns put into use.

[0090] The data communication module exchanges data with the DCS system through the ModBus communication protocol, and reads real-time data including: total coal feed amount, total ore amount, coal quality parameters, mineral composition, flue gas temperature at the outlet of the sintering process, wind box temperature, combustion air flow, main exhaust fan flow, flue gas oxygen content parameters, desulfurization tower inlet and outlet SO 2 Concentration, flue gas temperature and moisture content of inlet flue gas, feeder frequency, ash circulation feeder frequency, clean flue gas recirculation variable frequency fan frequency, and the number of multi-stage humidification activation spray guns put into use.

[0091] The real-time data read by the communication module may be distorted and missing due to instrument problems / data transmission problems / purging reasons. At the same time, to ensure that the model can fully extract the influence of each feature, the data is standardized / normalized, purged / data missing is processed before building the model. The specific steps are as follows:

[0092] Step 1.1: Identify and process data anomalies, filter out zero, negative, and null abnormal data, and clean them up.

[0093] Step 1.2: CEMS purge and maintenance data processing. The data during the purge period can be judged by the following formula.

[0094]

[0095] In the formula, is the SO2 at the desulfurization device inlet measured at time i 2 concentration; is the SO at the desulfurization device inlet measured at the i+jth moment 2 concentration, where j = 1, 2, 3, ..., 20; To determine the SO2 of the desulfurization device inlet during purging 2 Minimum change in concentration.

[0096] Step 1.3: In order to eliminate the impact of different feature magnitudes on the model, the data needs to be normalized, as shown in the following formula:

[0097]

[0098] In the formula, is the data of the feature after normalization at time i, x i is the original data of the feature at time i, x min is the minimum value of the feature, x max is the maximum value of the feature.

[0099] Figure 2 The desulfurization tower imports and exports CO 2 concentration, SO 2 concentration, SO 3 Flow chart for constructing a prediction model for the generation and removal of concentration and HCl concentration.

[0100] The desulfurization tower inlet and outlet CO 2 concentration, SO 2 concentration, SO 3 The construction of the generation and removal prediction model of concentration and HCl concentration mainly includes the following steps:

[0101] Step S1: Analysis of CO during combustion / sintering 2 / SO 2 / SO 3 The generation principle of various pollutants such as / HCl / HF, combined with data analysis methods, was used to screen the influence of CO inlet of circulating fluidized bed desulfurization tower. 2 / SO 2 / SO 3 / HCl / HF multiple pollutant concentration key parameters;

[0102] Step S2: Compensate for the delay between key variables through data processing, and build a knowledge and data fusion driven CO based on the LSTM neural network with autoregressive variables 2 / SO 2 / SO 3 / HCl / HF multiple pollutant generation concentration model to achieve desulfurization device inlet CO 2 / SO 2 / SO 3 / HCl / HF multiple pollutant concentrations can be accurately predicted in advance;

[0103] Step S3: Analyze CO 2 / SO 2 / SO 3 The adsorption and removal mechanism of various pollutants such as / HCl / HF and absorbent in the flue gas circulating fluidized bed desulfurization tower was established, and the acid gas adsorption and removal mechanism model was further established in CO 2 / SO 2 / SO 3 The correction parameters were introduced into the adsorption and removal mechanism model of various pollutants such as / HCl / HF, and the PSO algorithm was used to solve it. 2 / SO2 / SO 3 / HCl / HF multiple pollutant adsorption removal model;

[0104] Step S4: Further use the LSTM neural network to compensate for the parameter identification model error, combine the actual operation data, and establish a double-layer data corrected CO 2 / SO 2 / SO 3 / HCl / HF multiple pollutant removal prediction model realizes the CO removal at the outlet of the desulfurization device 2 / SO 2 / SO 3 Accurate prediction of the concentrations of multiple pollutants such as / HCl / HF.

[0105] As a preference, the knowledge and data fusion driven desulfurization tower inlet SO 2 The concentration prediction model is expressed as follows:

[0106]

[0107] in: is the predicted value of sulfur dioxide concentration at the desulfurization tower inlet, K r SO caused by boiler / sintering process operating conditions 2 Emission factor, K r =f(T,B,L 1 ,L 2 …); T is the flue gas temperature of the furnace or sintering process, B is the coal feed rate, L 1 is the primary air volume, L 2 is the secondary air volume, a fh is the fly ash coefficient, A ar To receive the base ash, S ar Received basic sulfur content, Q net,ar To receive the base low heat, ω CaO ,ω MgO , is the mass fraction of the corresponding substance in the ash, n is the coal consumption per hour, and Q is the flue gas volume at the desulfurization tower inlet.

[0108] As a preference, the knowledge and data fusion driven desulfurization tower inlet SO 3 The concentration prediction model is expressed as follows:

[0109]

[0110] in: is the predicted value of sulfur trioxide concentration at the desulfurization tower entrance, T is the flue gas temperature of the furnace or sintering process, B is the coal feed rate, L 1 is the primary air volume, L 2 is the secondary air volume, is the predicted value of sulfur dioxide concentration at the desulfurization tower inlet, and Q is the flue gas volume at the desulfurization tower inlet.

[0111] Preferably, the desulfurization tower inlet HCl concentration prediction model driven by knowledge and data fusion is expressed as follows:

[0112]

[0113] Where: C HCl,in,yc is the predicted value of hydrogen chloride concentration at the desulfurization tower inlet, K r is the HCl emission coefficient caused by the operating conditions of the boiler / sintering process; T is the flue gas temperature of the furnace or sintering process, B is the coal feed rate, L 1 is the primary air volume, L 2 is the secondary air volume, a fh is the fly ash coefficient, A ar To receive the base ash, Q net,ar To receive the base low heat, ω CaO ,ω MgO , is the mass fraction of the corresponding substance in the ash, n is the coal consumption per hour, and Q is the flue gas volume at the desulfurization tower inlet.

[0114] As a preference, the knowledge and data fusion driven desulfurization tower inlet CO 2 The concentration prediction model is expressed as follows:

[0115]

[0116] in: CO at the desulfurization tower inlet 2 volume, B is the amount of coal fed, O F is the total air volume entering the combustion process, C ar is the carbon content of the fuel / ore as received, H ar is the received hydrogen content of the fuel / ore, O ar is the received oxygen content of the fuel / ore, V CO is the carbon monoxide volume at the desulfurization tower inlet, β is the fuel characteristic coefficient, is the predicted value of sulfur dioxide concentration at the desulfurization tower inlet, Q is the flue gas volume at the desulfurization tower inlet;

[0117] As a preference, the SO at the outlet of the desulfurization tower based on parameter identification 2 The concentration prediction model is expressed as follows:

[0118]

[0119] Where: M Ca(OH)2 is the amount of slaked lime fed; Z h is the conversion rate of slaked lime, f(t l , tf …); t f is the time it takes for the desulfurization agent to complete its reaction in the absorption tower; t l The residence time of the desulfurizer in the absorption tower; SO in the desulfurization tower 2 partial pressure; T is temperature, V is the volume of the desulfurization tower, a is the air leakage coefficient of the desulfurization tower, is the predicted value of sulfur dioxide concentration at the desulfurization tower inlet, Q is the flue gas volume at the desulfurization tower inlet; R is a constant.

[0120] Furthermore, f is the reaction completion time of the desulfurization agent in the absorption tower and t l There are two parameters that need to be identified for the residence time of the desulfurizer in the absorption tower. The parameter identification process can be expressed as an optimization problem, namely:

[0121]

[0122] In the formula, RMSE(t l , t f ) is the model calculation export SO 2 The root mean square error between the predicted and measured concentrations.

[0123] Figure 3 The process of constructing a desulfurized flue gas adiabatic saturated temperature difference prediction model is as follows:

[0124] Step W1: Analyze the adiabatic cooling and humidification mechanism of flue gas in the flue gas circulating fluidized bed desulfurization process, construct a flue gas humidity prediction mechanism model at the outlet of the desulfurization tower, collect the moisture content of the flue gas at the inlet of the desulfurization tower, the flue gas volume at the inlet of the desulfurization tower, the flue gas temperature at the inlet of the desulfurization tower, the flue gas pressure at the inlet of the desulfurization tower, the water spray volume at the inlet of the desulfurization tower, the flue gas pressure at the outlet of the desulfurization tower, and the flue gas temperature at the outlet of the desulfurization tower, and construct a desulfurization flue gas adiabatic saturated temperature difference prediction model database;

[0125] Step W2: Combine data analysis methods to screen key parameters that affect the humidity of flue gas at the outlet of the desulfurization tower, further analyze the pure delay time from the change of key variables to the humidity of flue gas at the outlet of the desulfurization tower, and compensate for the delay between key variables through data processing;

[0126] Step W3: By introducing correction parameters into the flue gas humidity mechanism model at the outlet of the desulfurization tower and solving it through the PSO algorithm, a parameter-identified desulfurization tower outlet flue gas humidity control mechanism model is constructed;

[0127] Step W4: Further based on the LSTM neural network with autoregressive variables, a flue gas humidity prediction model at the outlet of the desulfurization tower is constructed by integrating data and humidity control mechanism;

[0128] Step W5: Further analyze the adiabatic saturated temperature control mechanism of desulfurized flue gas, and construct a desulfurized flue gas adiabatic saturated temperature difference prediction model based on the fusion of empirical knowledge and mechanism;

[0129] Step W6: Further combine the flue gas temperature at the outlet of the desulfurization tower, the flue gas pressure data at the inlet of the desulfurization tower and the prediction model of the flue gas humidity at the outlet of the desulfurization tower to construct a desulfurization flue gas adiabatic saturated temperature difference prediction model that integrates the mechanism and operation data correction, thereby realizing the accurate prediction of the adiabatic saturated temperature difference of the desulfurization flue gas under variable load conditions.

[0130] Preferably, the parameter-identified flue gas humidity control mechanism model at the outlet of the desulfurization tower can be expressed as follows:

[0131] H=λ*(G*(1-δ) / (Q*ρ)+H i )

[0132] Among them, G is the amount of water sprayed into the desulfurization tower, Q is the volume flow rate of flue gas, ρ is the density of flue gas, H i is the humidity of flue gas at the inlet of desulfurization tower, δ is the coefficient of residual moisture in desulfurization ash in the total water spraying volume, and λ is the correction coefficient of flue gas humidity at the outlet of desulfurization tower.

[0133] Furthermore, the parameters of δ and λ in the empirical model of flue gas humidity at the outlet of the desulfurization tower are identified. The parameter identification process can be expressed as an optimization problem, namely:

[0134]

[0135] Among them, RMSE (δ, λ) is expressed as the root mean square error between the predicted value and the actual value of the flue gas humidity at the outlet of the desulfurization tower.

[0136] Preferably, the adiabatic saturation temperature difference of the desulfurized flue gas can be expressed by the following formula:

[0137] ΔT s =TT s

[0138] Where, ΔT s is the adiabatic saturation temperature difference of the desulfurized flue gas, T is the flue gas temperature at the outlet of the desulfurization tower, and T s is the adiabatic saturation temperature;

[0139] Furthermore, the adiabatic saturated temperature difference prediction model of desulfurized flue gas based on the fusion of empirical knowledge and mechanism can be expressed by the following formula:

[0140]

[0141] Among them, C is the specific heat of wet flue gas, M is the molecular weight of flue gas, T is the flue gas temperature at the outlet of desulfurization tower, T sis the adiabatic saturation temperature, H is the humidity of the flue gas at T, P f is the flue gas pressure.

[0142] The multi-stage humidification module is not only the module with the highest water consumption in the desulfurization device, but also the most important subsystem to ensure the flue gas temperature at the desulfurization outlet and the adiabatic saturation temperature difference of the desulfurized flue gas. Therefore, when constructing the intelligent control problem of the multi-stage humidification module of the flue gas desulfurization device, the operating water consumption, the flue gas temperature at the desulfurization outlet and the adiabatic saturation temperature difference of the desulfurized flue gas should be considered at the same time. Therefore, the intelligent control process can also be described by the following optimization problem:

[0143] minWater(L 1 ,L 2 ,L 3 ,L 4 )=L 1 p 1 +L 2 p 2 +L 3 p 3 +L 4 p 4

[0144]

[0145] Among them, L 1 , L 2 , L 3 , L 4 It is the operating state of the 1st to 4th level humidification nozzle. When the humidification nozzle is turned on, the operating state is 1; when the humidification nozzle is turned off, the operating state is 0. 1 、p 2 、p 3 、p 4 Corresponding water flow rate for the humidifying nozzle. is the predicted value of outlet flue gas temperature, It is the predicted value of adiabatic saturated temperature difference of desulfurized flue gas.

[0146] Figure 4 This is the control strategy diagram for the lime optimization control module. Through the established sulfur dioxide concentration prediction model, sulfur trioxide concentration prediction model, HCl concentration prediction model and carbon dioxide concentration prediction model, the sulfur dioxide prediction value, sulfur trioxide calculation value, HCl calculation value and carbon dioxide calculation value are calculated. Further combined with the real-time operating data of the desulfurization tower furnace temperature, desulfurization tower furnace pressure and slaked lime feed rate, after multi-model prediction optimization control strategy adjustment, the slaked lime feed rate instruction is sent to the DCS controller through the data communication module, and then the DCS system sends the instruction to the field equipment.

[0147] Preferably, the predicted sulfur dioxide value, calculated sulfur trioxide value, calculated HCl value, calculated carbon dioxide value and the desulfurization tower inlet sulfur dioxide value are used as given instructions, and the desulfurization tower furnace temperature and desulfurization tower furnace pressure are used as feedforward instructions.

[0148] Furthermore, f(X1) is the CO at the desulfurization tower inlet. 2 The broken line function of volume conversion into lime mass; f(X2) is the SO at the desulfurization tower inlet 3 The broken line function of the concentration converted into the mass of slaked lime; f(X3) is the SO 2 f(X4) is the broken line function of converting the HCl concentration at the desulfurization tower inlet into the mass of slaked lime; f(X5) is the broken line function of converting the desulfurization tower temperature into the mass of slaked lime; f(X6) is the broken line function of converting the desulfurization tower furnace pressure into the mass of slaked lime.

[0149] As a preferred option, the multi-model prediction optimization control strategy is to set the export SO 2 Under the concentration conditions, the method of minimizing the feeding amount of the slaked lime feeder is selected. At the same time, in the actual operation process, since the frequency of the slaked lime feeder is difficult to achieve fine control, the slaked lime flow can only be adjusted by a specific frequency modulation amplitude. Frequent slaked lime feeder frequency will greatly affect the service life of the equipment; in addition, in order to avoid the outlet SO2 caused by untimely adjustment of the slaked lime feeder frequency, the slaked lime flow rate can be adjusted by adjusting the frequency of the slaked lime feeder. 2 Concentration exceeds the limit, set outlet SO 2 The concentration emission limit is to ensure SO 2 Meet emission standards.

[0150] Furthermore, the frequency switching condition of the slaked lime feeder can be expressed by the following mathematical expression:

[0151]

[0152] S time >S time,set

[0153] In the formula, Based on SO 2 The removal process model calculates the outlet SO when maintaining the existing operation strategy 2 concentration, To set the export SO 2 The upper limit of concentration emission is set to 35mg / m 3 ; S time,set The frequency switching time of the slaked lime feeder can be determined according to the corresponding operating procedures.

[0154] Furthermore, the specific calculation logic of the multi-model prediction optimization control strategy mainly includes the following steps: first, the best operation strategy under the current working conditions is selected according to the optimized operating condition library, and whether the current operation strategy of the slaked lime feeder is consistent with the best strategy is determined. If they are consistent, no adjustment is required. If they are inconsistent, first determine whether the outlet SO 2 Whether the concentration exceeds the set emission upper limit. If so, ignore other conditions and switch the feeder frequency immediately to prioritize ensuring that the flue gas meets the emission standards. If the emission standards are met, judge the switching time. If the switching time is greater than the specified time, switch. If the specified switching time has not been reached since the last switch, maintain the original operation strategy.

[0155] Figure 5 The optimization strategy for the bed pressure drop control module mainly includes the following steps:

[0156] Step X1: construct a bed pressure drop prediction model based on parameter identification;

[0157] Step X2: Given a target value for bed pressure drop, the key input variables of inlet pressure, flue gas flow, fresh slaked lime particle flow, clean flue gas circulation flow, and ash material circulation flow are discretized;

[0158] Step X3: Obtain the optimal configuration results of the clean flue gas circulation fan regulating valve opening and the ash material circulation regulating valve opening under different working conditions through discretization, construct a control rule table, and use it as the basis for judging the implementation of control actions to achieve intelligent control of the clean flue gas circulation fan and ash material circulation;

[0159] Step X4: Further introduce feedback variables into the control rules. By using the idea of ​​feedback correction in model predictive control, the model error is obtained by comparing the predicted bed pressure drop value output by the bed pressure drop prediction model based on parameter identification with the actual measured bed pressure drop value. The error is then fed back to correct the given reference trajectory and input it into the inference module, thereby ensuring the effectiveness of the controller.

[0160] Preferably, the actual absorption tower bed pressure drop is used as the feedback value, and the bed pressure drop control strategy output instruction is used as the material circulation control valve tracking value calculation parameter, and this value is multiplied by the corresponding control valve balance coefficient as the tracking value of the control valve. In order to prevent the return trough valve from frequently and slightly moving, a valve position holding program is set when the fluctuation amplitude is less than 1.

[0161] Furthermore, the continuous material levels of different ash hoppers are averaged, and the actual material level of each ash hopper is divided by the average value as the balancing coefficient of the ash return valve to achieve balanced control of the material level of each ash hopper.

[0162] In order to verify the accuracy and effectiveness of the knowledge and data fusion modeling method based on double-layer data correction, the present invention has conducted engineering application research in a domestic flue gas circulating fluidized bed desulfurization project. The selected operating conditions include stable, sudden load, low load, medium load, and high load. A set of operating condition data is collected every 5 seconds, and a total of 72 hours of continuous operating condition data is collected for verification. The results are as follows: Figure 6 As shown, SO 2 The model prediction value of emission concentration is consistent with the actual measured value. Based on parameter identification, a recurrent neural network is used to construct a data correction model to further compensate for the model error. The RMSE can reach 0.51 mg / m 3 , explaining the SO based on the knowledge and data fusion modeling method of double-layer data correction 2 The emission concentration model has high prediction accuracy.

[0163] Through the established sulfur dioxide concentration prediction model, sulfur trioxide concentration prediction model, HCl concentration prediction model and carbon dioxide concentration prediction model, the sulfur dioxide prediction value, sulfur trioxide calculation value, HCl calculation value and carbon dioxide calculation value are calculated, and further combined with the real-time operating data of the desulfurization tower furnace temperature, desulfurization tower furnace pressure and slaked lime feed amount, after multi-model prediction optimization control strategy adjustment, the slaked lime feed amount instruction is sent to the DCS controller through the data communication module, and then the DCS system sends the instruction to the field equipment.

[0164] Based on the accurate prediction model of sulfur dioxide emission concentration based on knowledge and data fusion of double-layer data correction, the optimization control of sulfur dioxide and slaked lime was carried out, and the results are as follows: Figure 7 As shown in the figure, the average emission concentration of sulfur dioxide before control in the present invention is 20 mg / m 3 The average emission concentration after optimization is 30mg / m 3 , export SO 2 Concentration fluctuations were reduced by more than 40%, achieving SO 2 High-precision "edge-cutting" control of emission concentration reduces the amount of slaked lime feed by 10% and carbon dioxide emissions by 9%, and the desulfurization efficiency in the desulfurization tower is as high as 97.5%.

[0165] Based on the accurate prediction model of adiabatic saturated temperature difference of desulfurized flue gas, intelligent control of multi-stage humidification module was carried out. The results are as follows: Figure 7As shown, before the application of this patent, the flue gas temperature at the outlet of the desulfurization tower was between 95 and 110°C, and the fluctuating temperature of the flue gas at the outlet of the desulfurization tower was about ±7.5°C, with a large fluctuation range; after the application of the results of this patent, the flue gas temperature at the outlet of the desulfurization tower was stably controlled at 105°C, and the fluctuating temperature of the flue gas at the outlet of the desulfurization tower was about ±2°C. Compared with before the application of the patent results, the flue gas temperature fluctuation at the outlet of the desulfurization tower was reduced by more than 70%, realizing high-precision "edge-cutting" control of the flue gas temperature at the outlet of the desulfurization tower, and thereby reducing the water consumption of the multi-stage humidification module by more than 5%, further realizing the control of the adiabatic saturated temperature difference at an appropriate value, improving the desulfurization efficiency while avoiding the adiabatic saturated temperature difference being too low, which will cause flue gas condensation and corrosion of the inner wall and separation and dust removal equipment.

[0166] When the opening of the material circulation regulating valve increases, the bed pressure drop will increase accordingly. Conversely, when the opening of the material circulation regulating valve decreases, the bed pressure drop will decrease accordingly. The function of the bed pressure drop control loop is to control the bed pressure drop by controlling the opening of the material circulation regulating valve. After the application of this patent result, such as Figure 8 As shown, the bed pressure drop under variable loads can be operated continuously and stably, controlled between 1500Pa and 1650Pa, and the fluctuation range is reduced by more than 30% compared with that before the application of the results.

[0167] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. Intelligent circulating fluidized bed desulfurization synergistic carbon-pollution deep purification system, the system includes slaked lime preparation and transportation system, ash circulation system, temperature and humidity control system, flue gas recirculation system, pre-dust collector, bag dust collector, desulfurization tower and intelligent control system, characterized in that: The temperature and humidity control system is provided with a multi-stage humidification activation spray gun, the slaked lime preparation and transportation system is provided with a variable frequency feeder, the ash circulation system is provided with an ash circulation variable frequency feeder, the flue gas recirculation system is provided with a clean flue gas recirculation variable frequency fan, and the intelligent control system includes an online monitoring module, a key parameter prediction module, and a key parameter control module; The online monitoring module includes the total coal feeding amount, total ore amount, coal quality parameters, mineral composition, flue gas temperature at the outlet of the sintering process, bellows temperature, combustion air flow, main exhaust fan flow, flue gas oxygen content parameters, SO2 concentration at the inlet and outlet of the desulfurization tower, flue gas temperature and inlet flue gas moisture content, feeder frequency, ash circulation feeder frequency, clean flue gas recirculation variable frequency fan frequency, and the number of multi-stage humidification activation spray guns put into use; The key parameter prediction module includes the generation and removal prediction models of CO2 concentration, SO2 concentration, SO3 concentration and HCl concentration at the desulfurization tower inlet and outlet, and the flue gas temperature and flue gas pressure prediction models at the desulfurization tower outlet; the key parameter control module includes the slaked lime variable frequency feeder control model, the bed pressure drop prediction control model, and the multi-stage humidification activation spray gun control model.

2. The intelligent circulating fluidized bed desulfurization coordinated carbon-pollution deep purification system according to claim 1 is characterized by: The construction of the generation and removal prediction model of the CO2 concentration, SO2 concentration, SO3 concentration, and HCl concentration at the desulfurization tower inlet and outlet mainly includes the following steps: Step S1: Analyze the generation principle of CO2, SO2, SO3, and HCl acid gases during the combustion / sintering process, and screen the key parameters that affect the concentration of CO2, SO2, SO3, and HCl acid gases at the inlet of the circulating fluidized bed desulfurization tower by combining data analysis methods; Step S2: Compensate for the delay between key variables through data processing, and build a knowledge and data fusion driven model of the concentration of CO2, SO2, SO3, HCl and HF acid gases based on the LSTM neural network with autoregressive variables, so as to achieve the early and accurate prediction of the concentration of CO2, SO2, SO3 and HCl acid gases at the inlet of the desulfurization device; Step S3: Analyze the adsorption and removal mechanism of CO2, SO2, SO3, and HC acid gases with absorbent in the flue gas circulating fluidized bed desulfurization tower, establish an acid gas adsorption and removal mechanism model, further introduce correction parameters into the CO2, SO2, SO3, and HCl acid gas adsorption and removal mechanism model, solve it through the PSO algorithm, and construct a CO2, SO2, SO3, and HCl acid gas adsorption and removal model based on parameter identification; Step S4: Further adopting LSTM neural network to compensate for the error of parameter identification model, combining with actual operation data, establishing knowledge and data fusion driven acid gas concentration prediction model of double-layer data correction, realizing accurate prediction of CO2, SO2, SO3 and HCl acid gas concentration at the outlet of desulfurization device; The prediction model of SO2 concentration at the desulfurization tower inlet driven by knowledge and data fusion is described as follows: in: is the predicted value of sulfur dioxide concentration at the desulfurization tower inlet, K r is the SO2 emission factor caused by the operating conditions of the boiler / sintering process, K r =f(T,B,L1,L2…); T is the flue gas temperature of the furnace or sintering process, B is the coal feed rate, L1 is the primary air volume, L2 is the secondary air volume, a fh is the fly ash coefficient, A ar To receive the base ash, S ar Received basic sulfur content, Q net,ar To receive the base low heat, ω CaO ,ω MgO , is the mass fraction of the corresponding substance in the ash, n is the coal consumption per hour, and Q is the flue gas volume at the desulfurization tower inlet.

3. The intelligent circulating fluidized bed desulfurization coordinated carbon-pollution deep purification system according to claim 2 is characterized by: The prediction model of SO3 concentration at the desulfurization tower inlet driven by knowledge and data fusion is described as follows: in: is the predicted value of sulfur trioxide concentration at the desulfurization tower entrance, T is the flue gas temperature of the furnace or sintering process, B is the coal feed rate, L1 is the primary air volume, L2 is the secondary air volume, is the predicted value of sulfur dioxide concentration at the desulfurization tower inlet, Q is the flue gas volume at the desulfurization tower inlet; The prediction model of HCl concentration at the desulfurization tower inlet driven by knowledge and data fusion is described as follows: Where: C HCl,in,yc is the predicted value of hydrogen chloride concentration at the desulfurization tower inlet, K r is the HCl emission coefficient caused by the operating conditions of the boiler / sintering process; T is the flue gas temperature of the furnace or sintering process, B is the coal feed rate, L1 is the primary air volume, L2 is the secondary air volume, and a fh is the fly ash coefficient, A ar To receive the base ash, Q net,ar To receive the base low heat, ω CaO ,ω MgO , is the mass fraction of the corresponding substance in the ash, n is the coal consumption per hour, and Q is the flue gas volume at the desulfurization tower inlet; The knowledge and data fusion-driven desulfurization tower inlet CO2 concentration prediction model is described as follows: in: is the volume of CO2 at the desulfurization tower entrance, B is the amount of coal fed, Q F is the total air volume entering the combustion process, C ar is the carbon content of the fuel / ore as received, H ar is the received hydrogen content of the fuel / ore, O ar is the received oxygen content of the fuel / ore, V CO is the carbon monoxide volume at the desulfurization tower inlet, β is the fuel characteristic coefficient, is the predicted value of sulfur dioxide concentration at the desulfurization tower inlet, Q is the flue gas volume at the desulfurization tower inlet; The prediction model of SO2 concentration at the outlet of the desulfurization tower based on parameter identification is as follows: Where: M Ca(OH)2 is the amount of slaked lime fed; Z h is the conversion rate of slaked lime, f(t l , t f …); t f is the time it takes for the desulfurization agent to complete its reaction in the absorption tower; t l The residence time of the desulfurizer in the absorption tower; is the partial pressure of SO2 in the desulfurization tower; T is the temperature, V is the volume of the desulfurization tower, a is the air leakage coefficient of the desulfurization tower, is the predicted value of sulfur dioxide concentration at the desulfurization tower inlet, Q is the flue gas volume at the desulfurization tower inlet; R is a constant; Further based on the above model, t f is the reaction completion time of the desulfurization agent in the absorption tower and t l There are two parameters that need to be identified for the residence time of the desulfurizer in the absorption tower. The parameter identification process can be expressed as an optimization problem, namely: In the formula, RMSE(t l , t f ) is the root mean square error between the predicted value of outlet SO2 concentration calculated by the model and the actual measured value.

4. The intelligent circulating fluidized bed desulfurization coordinated carbon-pollution deep purification system according to claim 1 is characterized by: The establishment of the prediction model of the flue gas temperature at the outlet of the desulfurization tower and the adiabatic saturated temperature difference of the desulfurized flue gas mainly includes the following steps: Step W1: Analyze the adiabatic cooling and humidification mechanism of flue gas in the flue gas circulating fluidized bed desulfurization process, construct a flue gas humidity prediction mechanism model at the outlet of the desulfurization tower, collect the moisture content of the flue gas at the inlet of the desulfurization tower, the flue gas volume at the inlet of the desulfurization tower, the flue gas temperature at the inlet of the desulfurization tower, the flue gas pressure at the inlet of the desulfurization tower, the water spray volume at the inlet of the desulfurization tower, the flue gas pressure at the outlet of the desulfurization tower, and the flue gas temperature at the outlet of the desulfurization tower, and construct a desulfurization flue gas adiabatic saturated temperature difference prediction model database; Step W2: Combine data analysis methods to screen key parameters that affect the humidity of flue gas at the outlet of the desulfurization tower, further analyze the pure delay time from the change of key variables to the humidity of flue gas at the outlet of the desulfurization tower, and compensate for the delay between key variables through data processing; Step W3: By introducing correction parameters into the flue gas humidity mechanism model at the outlet of the desulfurization tower and solving it through the PSO algorithm, a parameter-identified desulfurization tower outlet flue gas humidity control mechanism model is constructed; Step W4: Further construct a flue gas humidity prediction model at the outlet of the desulfurization tower based on the LSTM neural network with autoregressive variables, which integrates data and humidity control mechanism; Step W5: Further analyze the adiabatic saturated temperature control mechanism of desulfurized flue gas, and construct a desulfurized flue gas adiabatic saturated temperature difference prediction model based on the fusion of empirical knowledge and mechanism; Step W6: Further combine the flue gas temperature at the outlet of the desulfurization tower, the flue gas pressure data at the inlet of the desulfurization tower and the prediction model of the flue gas humidity at the outlet of the desulfurization tower to construct a desulfurization flue gas adiabatic saturated temperature difference prediction model that integrates the mechanism and operation data correction, thereby realizing the accurate prediction of the adiabatic saturated temperature difference of the desulfurization flue gas under variable load conditions.

5. The intelligent circulating fluidized bed desulfurization coordinated carbon-pollution deep purification system according to claim 4 is characterized in that: The parameter-identified flue gas humidity control mechanism model at the outlet of the desulfurization tower can be expressed as follows: H=λ*(G*(1-δ) / (Q*ρ)+H i ) Among them, G is the amount of water sprayed into the desulfurization tower, Q is the volume flow rate of flue gas, ρ is the density of flue gas, H i is the humidity of flue gas at the inlet of desulfurization tower, δ is the coefficient of residual water in desulfurization ash in the total water spraying amount, and λ is the correction coefficient of flue gas humidity at the outlet of desulfurization tower; The parameters of δ and λ in the empirical model of flue gas humidity at the outlet of the desulfurization tower are identified. The process of parameter identification can be expressed as an optimization problem, namely: Among them, RMSE (δ, λ) represents the root mean square error between the predicted value and the actual value of the flue gas humidity at the outlet of the desulfurization tower; The adiabatic saturation temperature difference of the desulfurized flue gas can be expressed by the following formula: ΔT s =T-T s Where, ΔT s is the adiabatic saturation temperature difference of the desulfurized flue gas, T is the flue gas temperature at the outlet of the desulfurization tower, and T s is the adiabatic saturation temperature; The desulfurized flue gas adiabatic saturated temperature difference prediction model based on empirical knowledge and mechanism fusion can be expressed as follows: Among them, C is the specific heat of wet flue gas, M is the molecular weight of flue gas, T is the flue gas temperature at the outlet of desulfurization tower, T s is the adiabatic saturation temperature, H is the humidity of the flue gas at T, P f is the flue gas pressure.

6. The intelligent circulating fluidized bed desulfurization coordinated carbon-pollution deep purification system according to claim 1 is characterized by: The intelligent control and optimization of the multi-stage humidification module of the flue gas desulfurization device can be described as follows: min Water(L1,L2,L3,L4)=L1p1+L2p2+L3p3+L4p4 Among them, L1, L2, L3, and L4 are the operating states of the 1st to 4th level humidification nozzles. When the humidification nozzle is turned on, the operating state is 1; when the humidification nozzle is turned off, the operating state is 0; p1, p2, p3, and p4 are the corresponding water flow rates of the humidification nozzles; is the predicted value of outlet flue gas temperature, It is the predicted value of adiabatic saturated temperature difference of desulfurized flue gas.

7. The intelligent circulating fluidized bed desulfurization coordinated carbon-pollution deep purification system according to claim 1 is characterized by: The slaked lime optimization control module calculates the predicted value of sulfur dioxide, the calculated value of sulfur trioxide, the calculated value of HCl, and the calculated value of carbon dioxide through the established sulfur dioxide concentration prediction model, sulfur trioxide concentration prediction model, HCl concentration prediction model, and carbon dioxide concentration prediction model, and further combines the real-time operation data of the desulfurization tower furnace temperature, the desulfurization tower furnace pressure, and the slaked lime feeding amount, and after adjustment by the multi-model prediction optimization control strategy, sends the slaked lime feeding amount instruction to the DCS controller through the data communication module, so that the DCS system sends the instruction to the field equipment; The predicted sulfur dioxide value, calculated sulfur trioxide value, calculated HCl value, calculated carbon dioxide value and the sulfur dioxide value at the desulfurization tower inlet are used as given instructions, and the desulfurization tower furnace temperature and desulfurization tower furnace pressure are used as feedforward instructions; f(X1) is the broken line function of the conversion of the volume of CO2 at the desulfurization tower inlet into the mass of slaked lime; f(X2) is the broken line function of the conversion of the concentration of SO3 at the desulfurization tower inlet into the mass of slaked lime; f(X3) is the broken line function of the conversion of the concentration of SO2 at the desulfurization tower inlet into the mass of slaked lime; f(X4) is the broken line function of the conversion of the concentration of HCl at the desulfurization tower inlet into the mass of slaked lime; f(X5) is the broken line function of the conversion of the temperature of the desulfurization tower into the mass of slaked lime; f(X6) is the broken line function of the conversion of the furnace pressure of the desulfurization tower into the mass of slaked lime; The multi-model prediction optimization control strategy is to select the method with the minimum feeding amount of the slaked lime feeder under the condition of setting the outlet SO2 concentration; at the same time, in the actual operation process, since the frequency of the slaked lime feeder is difficult to achieve fine control, the slaked lime flow can only be adjusted by a specific frequency modulation amplitude. Frequent slaked lime feeder frequency will greatly affect the service life of the equipment; in addition, in order to avoid the problem of excessive outlet SO2 concentration due to untimely adjustment of the slaked lime feeder frequency, the outlet SO2 concentration emission upper limit is set to ensure that SO2 emissions meet the standards; The frequency switching condition of the slaked lime feeder can be expressed as follows: S time >S time,set In the formula, To calculate the outlet SO2 concentration based on the SO2 removal process model while maintaining the existing operation strategy, The upper limit of the outlet SO2 concentration is set to 35 mg / m 3 ; S time,set The switching time of the frequency of the slaked lime feeder can be determined according to the corresponding operating procedures; The specific calculation logic of the multi-model prediction optimization control strategy mainly includes the following steps: first, the best operating strategy under the current operating conditions is selected according to the optimized operating condition library, and it is determined whether the current operating strategy of the slaked lime feeder is consistent with the best strategy. If they are consistent, no adjustment is required. If they are inconsistent, first determine whether the outlet SO2 concentration exceeds the set emission upper limit under the original operating strategy. If it exceeds, ignore other conditions and immediately switch the feeder frequency to give priority to ensuring that the flue gas meets the emission standards; if the emission standards are met, the switching time is judged. If the switching time is greater than the specified time, switch; if the specified switching time has not been reached since the last switch, maintain the original operating strategy.

8. The intelligent circulating fluidized bed desulfurization coordinated carbon-pollution deep purification system according to claim 1 is characterized by: The bed pressure drop control module optimization strategy mainly includes the following steps: Step 1: Construct a bed pressure drop prediction model based on parameter identification; Step 2: Given a target value for bed pressure drop, the key input variables of inlet pressure, flue gas flow, fresh slaked lime particle flow, clean flue gas circulation flow, and ash material circulation flow are discretized; Step 3: Obtain the optimal configuration results of the clean flue gas circulation fan regulating valve opening and the ash material circulation regulating valve opening under different working conditions through discretization, build a control rule table, and use it as the basis for judging the implementation of control actions to achieve intelligent control of the clean flue gas circulation fan and ash material circulation; Step 4: Further introduce feedback variables into the control rules. By using the idea of ​​feedback correction in model predictive control, the model error is obtained by comparing the predicted bed pressure drop value output by the bed pressure drop prediction model based on parameter identification with the actual measured bed pressure drop value. The error is then fed back to correct the given reference trajectory and input it into the inference module, thereby ensuring the effect of the controller. The actual absorption tower bed pressure drop is used as the feedback value, the bed pressure drop control strategy output instruction is used as the material circulation regulating valve tracking value calculation parameter, and the bed pressure drop control strategy output instruction value is multiplied by the corresponding regulating valve balance coefficient as the tracking value of the material circulation regulating valve. In order to prevent the return trough valve from frequent and small movements, a valve position holding program is set when the fluctuation amplitude is less than 1; the continuous material levels of different ash hoppers are averaged, and the actual material level of each ash hopper is divided by the average value as the return ash valve balance coefficient to achieve balanced control of the material level of each ash hopper.

Images