Wet desulphurization oxidation wind demand prediction model establishment method

Abstract

The invention discloses a method for establishing a wet desulphurization oxidation wind demand prediction model. The method comprises five stages of model variable setting, theoretical model prediction control, model data and dissolved oxygen deviation data collection, model data filtering and prediction model generation. And according to the set model characteristic variable and the label variable, collecting data which is obtained by theoretical model prediction control and meets a model building requirement. A machine learning or data fitting method is adopted for the screened data, and an oxidation wind demand prediction model with the feature variables as independent variables and the label variables as prediction variables is generated. The method can be used as a precise feedforward for controlling the supply quantity of the wet desulphurization oxidation air, and the control precision of the supply quantity of the wet desulphurization oxidation air is remarkably improved.

Description

Technical Field

[0001] The invention belongs to the technical field of environmental protection and relates to a method for establishing a wet desulfurization oxidation air demand prediction model. Background Art

[0002] In recent years, the proportion of renewable energy power generation in my country has been rising year by year, and the positioning of coal-fired power generation has gradually changed from basic load to flexible peak load, so the load adjustment of thermal power units has changed frequently. As an important environmental protection device for coal-fired thermal power units, the energy-saving operation of wet desulfurization devices during the main engine load adjustment process has become a technical research direction that the industry is currently focusing on.

[0003] In the wet flue gas desulfurization system, the oxidizing air is used to oxidize the calcium sulfite in the circulating slurry into calcium sulfate dihydrate, which continuously promotes the wet flue gas desulfurization reaction. The oxidizing air supply directly affects the content of calcium sulfite in the circulating slurry, but the calcium sulfite content in the slurry cannot be measured and monitored online in real time; it indirectly affects the SO2 concentration at the outlet of the desulfurization system, but it is not the only influencing factor. Therefore, there is a lack of basis for online real-time regulation and control of the oxidizing air volume, which is not conducive to the energy-saving operation of the oxidizing fan.

[0004] Some new technologies in the industry propose to control the oxidation air supply by measuring the dissolved oxygen or redox potential in the slurry. For example, patent CN117244387A uses the inlet SO2 flow rate to estimate the oxidation air demand according to the fixed oxidation air utilization rate as the feedforward for the oxidation air supply control, and uses the slurry dissolved oxygen or redox potential as the feedback for the oxidation air supply control. However, the above technology does not fully consider other major factors affecting the oxidation air demand, and does not use feedback optimization to improve the reliability of the feedforward. Therefore, the calculation error of the oxidation air demand by this method is large, and the feedback control lag is significant, and the oxidation air supply control accuracy is low.

[0005] Although the oxidation wind theoretical calculation model can be used to predict the oxidation wind demand online in real time, the oxidation wind system is designed according to full-load conditions. When operating at low load, the oxidation wind distribution system deviates from the design conditions, and the oxidation wind utilization rate changes, resulting in a deviation between the theoretical calculation results of the oxidation wind demand under the deviated design load conditions and the actual demand. In addition, factors such as the deviation of the device measurement instrument will also cause deviations between the theoretical calculation of the oxidation wind demand and the actual supply measurement.

[0006] Based on the above situation, the present invention proposes a method for establishing a wet desulfurization oxidation air demand prediction model, which can be used as an accurate feedforward for the control of the oxidation air supply, and significantly improves the control accuracy of the wet desulfurization oxidation air supply. Summary of the invention

[0007] The technical problem to be solved by the present invention is the problem of delayed control of the oxidation air supply amount and low control accuracy.

[0008] In order to achieve the above-mentioned purpose, the present invention proposes a method for establishing a wet flue gas desulfurization oxidation air demand prediction model. The method for establishing a wet flue gas desulfurization oxidation air demand prediction model includes five stages: model variable setting, theoretical model prediction control, model data and dissolved oxygen deviation data collection, model data filtering, and prediction model generation.

[0009] The model variable setting includes setting 9 characteristic variables and 1 label variable. The characteristic variables include: 1) oxygen concentration of flue gas at the desulfurization inlet; 2) SO2 concentration at the desulfurization inlet; 3) SO2 concentration at the desulfurization outlet; 4) flue gas flow at the desulfurization outlet; 5) moisture content of flue gas at the desulfurization outlet; 6) liquid level of the desulfurization tower slurry pool; 7) total flow of circulating slurry; 8) chloride ion concentration of desulfurization circulating slurry; 9) magnesium ion concentration in desulfurization circulating slurry. The label variable is the demand for oxidation air.

[0010] The theoretical model predictive control adopts the oxidation air theoretical calculation model to calculate the oxidation air demand online in real time, and uses it as the feedforward of the oxidation air supply control; the circulating slurry dissolved oxygen is measured online in real time, and used as the feedback of the oxidation air supply control; the circulating slurry dissolved oxygen set value is used as the oxidation air supply control target; the above-mentioned oxidation air supply control scheme is adopted to adjust and control the oxidation fan air supply in real time.

[0011] The model data and dissolved oxygen deviation data collection, in the process of the theoretical model predictive control, collects the 9 characteristic variables, 1 tag variable and dissolved oxygen deviation data in real time, and aligns the collected data by time. The dissolved oxygen deviation is the deviation between the circulating slurry dissolved oxygen measured online in real time and the set value of the circulating slurry dissolved oxygen.

[0012] The model data is filtered by setting a dissolved oxygen deviation limit, filtering the collected model variable data, and filtering out all data that is higher than the dissolved oxygen deviation limit. The dissolved oxygen deviation limit is the maximum allowable value of the dissolved oxygen deviation.

[0013] The prediction model is generated by performing machine learning or data fitting on the filtered model variable data, taking the feature variables as independent variables and the label variables as predictive variables, to generate a calculation formula that describes the predictive variables with mathematical expressions of the independent variables, i.e., the prediction model. DETAILED DESCRIPTION

[0014] The present invention relates to a method for establishing a wet desulfurization oxidation air demand prediction model, the method comprising the following steps:

[0015] 1. Set model variables, including 9 characteristic variables and 1 label variable. The characteristic variables include: 1) oxygen concentration of flue gas at desulfurization inlet; 2) SO2 concentration at desulfurization inlet; 3) SO2 concentration at desulfurization outlet; 4) flue gas flow at desulfurization outlet; 5) moisture content of flue gas at desulfurization outlet; 6) liquid level of desulfurization tower slurry pool; 7) total flow of circulating slurry; 8) chloride ion concentration of desulfurization circulating slurry; 9) magnesium ion concentration in desulfurization circulating slurry. The label variable is the demand for oxidation air.

[0016] 2. Use theoretical models to predict and control the oxidation air supply. Use the oxidation air demand calculated online and in real time by the oxidation air theoretical calculation model as the feedforward for oxidation air supply control, use the circulating slurry dissolved oxygen measured online and in real time as the feedback for oxidation air supply control, use the circulating slurry dissolved oxygen set value as the control target, and adjust and control the oxidation air fan supply in real time. The circulating slurry dissolved oxygen set value ranges from 0.5 to 5 mg / L, and the circulating slurry dissolved oxygen measurement sampling location is the bottom of the desulfurization tower slurry pool.

[0017] 3. Collect model variable data and dissolved oxygen deviation data. In the process of predicting and controlling the oxidation air supply using the theoretical model in step 2, the 9 characteristic variables, 1 label variable and dissolved oxygen deviation data described in step 1 are collected in real time. The model variable data has unified time information, that is, each time point corresponds to 9 characteristic variable data, 1 label variable data and 1 dissolved oxygen real-time deviation data. The dissolved oxygen deviation is the deviation between the circulating slurry dissolved oxygen measured online in real time and the set value of the circulating slurry dissolved oxygen. The frequency of collecting model variable data is 50ms~10min / time.

[0018] 4. Filter the model variable data. Set the dissolved oxygen deviation limit, filter the model variable data collected in step 3, filter out all time points that are higher than the dissolved oxygen deviation limit, and remove all model variable data corresponding to the filtered time points. The dissolved oxygen deviation limit is the maximum allowable value of the dissolved oxygen deviation.

[0019] 5. Generate a prediction model. For the model variable data that has been filtered in step 4, use the feature variable as the independent variable and the label variable as the prediction variable to perform machine learning or data fitting to generate a prediction variable described by a mathematical expression of the independent variable, i.e., a prediction model.

Claims

1. A method for establishing a prediction model for wet flue gas desulfurization oxidation air demand, characterized in that: The method is divided into five stages: model variable setting, theoretical model predictive control, model data and dissolved oxygen deviation data collection, model data filtering, and predictive model generation.

2. The model variable setting according to claim 1, characterized in that The model variables include 9 characteristic variables and 1 label variable. The characteristic variables include: 1) oxygen concentration of flue gas at the desulfurization inlet; 2) SO2 concentration at the desulfurization inlet; 3) SO2 concentration at the desulfurization outlet; 4) flue gas flow at the desulfurization outlet; 5) moisture content of flue gas at the desulfurization outlet; 6) liquid level of the desulfurization tower slurry pool; 7) total flow of circulating slurry; 8) chloride ion concentration of desulfurization circulating slurry; 9) magnesium ion concentration in desulfurization circulating slurry. The label variable is the demand for oxidation air.

3. The theoretical model predictive control according to claim 1, characterized in that The oxidation air theoretical calculation model is used to calculate the oxidation air demand online in real time, and it is used as the feedforward for the oxidation air supply control; the dissolved oxygen in the circulating slurry is measured online in real time, and it is used as the feedback for the oxidation air supply control; the set value of the dissolved oxygen in the circulating slurry is used as the oxidation air supply control target; the above-mentioned oxidation air supply control scheme is used to adjust and control the oxidation fan air supply in real time.

4. The model data filtering according to claim 1, characterized in that A dissolved oxygen deviation limit is set, the collected model data and dissolved oxygen deviation data are screened, all time points above the dissolved oxygen deviation limit are screened out, and all model variable data corresponding to the screened out time points are eliminated.