Method for classifying data features of flue gas emission continuous monitoring system in cement industry

Abstract

The invention discloses a cement industry flue gas emission continuous monitoring system data feature classification method, which does not assume that normal cement enterprise kiln tail exhaust port flue gas emission continuous monitoring system data obeys any known distribution family, and only solves the distribution from the data. A kernel density estimation method is used for modeling data of a continuous monitoring system for flue gas emission of a kiln tail exhaust port of a normal cement enterprise, the method is used for constructing normal non-parametric probability distribution, probability density distribution of the normal non-parametric probability distribution is determined, and the probability of areas where normal examples densely occur is high. The method has the advantages that 1, a supervision index is provided, so that an ecological environment supervision department can supervise whether the state of the kiln tail discharge port of the cement enterprise is normal or not; 2, a cement enterprise is helped to quickly monitor the process state, so that the cement enterprise operates in a normal operation range; the state of a large number of cement discharge ports can be rapidly checked in a short time, cost is low, and speed is high.

Description

Technical Field

[0001] The present invention relates to the technical field of environmental monitoring and protection, and particularly relates to a method for classifying data characteristics of a continuous flue gas emission monitoring system in the cement industry. Background Art

[0002] The continuous flue gas emission monitoring system, also known as the continuous automatic waste gas monitoring system, continuously and automatically monitors the concentration of particulate matter and gaseous pollutants in fixed pollution sources as well as the total amount of pollutant emissions, and transmits the monitoring data to the competent ecological environment department through the network to ensure that the pollutant concentration and the total amount of pollutant emissions of polluting enterprises meet the standards. The monitored gaseous pollutant indicators of the continuous flue gas emission monitoring system are particulate matter, SO2, NO X , flue gas, and the parameters are O2, flow rate, pressure, flue gas temperature, and flue gas humidity. The continuous flue gas emission monitoring system is mainly applied to fixed pollution source flue gas emission devices such as thermal power plant boilers, other industrial and civil boilers, industrial furnaces, municipal waste incinerators, and hazardous waste incinerators. With the in-depth development of the supervision and law enforcement of pollution discharge and the reduction of the total amount of main pollutant emissions, a large number of pollutant treatment facilities including dust removal, desulfurization, and denitrification have been put into operation, resulting in a gradual expansion of the installation and use scale of the continuous flue gas emission monitoring system. The continuous flue gas emission monitoring system can continuously and real-time track and measure the concentration and emission rate of particulate matter and gaseous pollutants in the flue gas emitted by the above-mentioned fixed pollution sources. The monitoring of fixed pollution source waste gas emissions is the basic work and data source of ecological environmental protection, and an important part of ecological environmental protection supervision and law enforcement and total emission reduction accounting. The continuous flue gas emission monitoring system provides a large amount of basic data and reference basis for the supervision and law enforcement of flue gas emissions from gas-related fixed pollution sources, the collection of pollution discharge fees, and the verification of total emission reduction. The data of the continuous flue gas emission monitoring system is the data generated after the joint action of the production process, working conditions, pollutant treatment facilities, and the continuous flue gas emission monitoring system. Conversely, the data generated by the continuous flue gas emission monitoring system implies the information of the production process, working conditions, pollutant treatment facilities, and the continuous flue gas emission monitoring system. With the in-depth application of the continuous flue gas emission monitoring system, the continuous flue gas emission monitoring system also continuously generates new data and accumulates a large amount of flue gas monitoring data. The total amount of these data is huge, the structure is complex, the format is rich, and the value density is low, but the value is large. The mining of these data is an important task. Through data mining technology, the inherent data value can be emerged from it to assist enterprises and administrative management departments in the management of the continuous flue gas emission monitoring system and the production process. The quality guarantee and mining of the monitoring data of the continuous flue gas emission monitoring system are the technical basis for key tasks such as environmental supervision, environmental monitoring, and total emission reduction accounting. At the same time, this is also a new method of remote monitoring, which can be used in aspects such as ecological environment supervision, condition monitoring, and emergency control.

[0003] To meet the regulatory requirements of the ecological environment supervision department, a method is needed to quickly and at low cost obtain the deviation characteristics of the data of the flue gas emission continuous monitoring system of each cement enterprise for remote high-frequency and near-real-time supervision. At the same time, this method can also meet the needs of enterprises to monitor the status of the data of their own flue gas emission continuous monitoring systems. Summary of the Invention

[0004] The object of the present invention is to provide a method for classifying the data characteristics of the flue gas emission continuous monitoring system in the cement industry.

[0005] The present invention does not assume that the data of the flue gas emission continuous monitoring system at the kiln tail discharge port of normal cement enterprises follows any known distribution family, but only obtains its distribution from the data. Using the kernel density estimation method, the data of the flue gas emission continuous monitoring system at the kiln tail discharge port of normal cement enterprises is modeled, and a non-parametric probability distribution of the normal class is constructed using this method, and its probability density distribution is determined. Areas where normal instances occur densely have a higher probability.

[0006] The specific steps are as follows:

[0007] Step 1. Data extraction, obtain the monitoring data of the flue gas emission continuous monitoring system at the discharge ports of all cement industry enterprises (set as set c), and the data includes the unique name of the discharge port, region, monitoring time, oxygen content, and NOx concentration data.

[0008] Step 2. Clean and regularize the extracted data, extract the data with oxygen content greater than 0 and less than 16%, and according to expert experience, the oxygen content interval of 16% - 21% is recognized as the data of kiln ignition, which is not the case of normal process production. Perform value range verification to remove outliers (extreme values, negative values, etc.).

[0009] Step 3. Determine the value range of NOx, set the minimum value to 0 and the maximum value to x.

[0010] Step 4. Using the oxygen content as the abscissa and the NOx concentration as the ordinate, calculate the probability density function of the cement industry using the non-parametric method of kernel density estimation.

[0011] Step 5. Division of the independent variable value range, the oxygen content is divided according to a compensation of 0.1%; the NOx concentration is divided in steps of 0.1 mg / m³ into 10x intervals; each independent variable value area is labeled and encoded with a number.

[0012] Step 6. Through the mathematical integration method, obtain the cumulative probability density of each rectangle of the independent variable domain, and also obtain the probability density distribution of the normal class. Obviously, the cumulative probability density of the entire independent variable domain is 1.

[0013] Step 7. Sort all the independent variable domain rectangles and their corresponding cumulative probability densities in descending order according to the cumulative probability density, and obtain the set of independent variable domain rectangles with a cumulative probability density greater than 95% through loop calculation.

[0014] Step 8. For the data of the kiln tail exhaust port of other cement enterprises to be inspected, estimate its PDF by the kernel density method with the same parameters, and use the set of independent variable domain rectangles with a cumulative probability density greater than 95% for mathematical integration to obtain the cumulative probability density yi of this area of the kiln tail exhaust port of this cement enterprise. The larger this cumulative probability density yi is, the closer it is to the normal class; the smaller this cumulative probability density yi is, the more deviated it is from the normal class.

[0015] Therefore, this cumulative probability density yi can be used as a determination index for the deviation of the kiln tail exhaust port of this cement enterprise from the normal class. By applying this yi, the data of the continuous monitoring system for flue gas emissions at the kiln tail exhaust port of cement enterprises can be divided into the normal class and the deviated class. The data characteristics of the normal class are the cement industry characteristics of the data of the continuous monitoring system for flue gas emissions at the kiln tail exhaust port of cement enterprises.

[0016] Specifically, Step 9. Integrate the "independent variable value range greater than 95%" to obtain the cumulative probability density yi of this cement enterprise;

[0017] Step 10. Repeat Step 8 and Step 9 to obtain the cumulative probability density yi of the kiln tail exhaust port of all individual cement enterprises. This cumulative probability density is the degree of deviation of an individual cement enterprise from the normal cement enterprise, that is, the characteristic score of the abnormal class of the kiln tail exhaust port of this cement enterprise. When the cumulative probability density deviates from the normal class, it is classified as the deviated class;

[0018] Step 11. Sort all the kiln tail exhaust ports of individual cement enterprises according to their cumulative probability density yi, and screen out the deviated class;

[0019] Step 12. For the cumulative probability density yi of the deviated class, conduct offline inspection and online verification, remove the deviated data of the kiln tail exhaust port of this cement enterprise from the set c, and then repeat Steps 1 to 11 for iteration;

[0020] Step 13. After multiple iterations, the set c obtained by removing the data of the kiln tail exhaust ports of the deviated class cement enterprises is the normal cement industry characteristic data set. Repeat Steps 2 to 4 to obtain the probability density function, which is the normal cement industry characteristic probability density function.

[0021] In summary, the present invention has the following beneficial effects: The advantages of the process and device involved in the present invention are as follows: 1. The present invention provides a supervision index, which helps the ecological environment supervision department to supervise whether the state of the kiln tail exhaust port of cement enterprises is normal.

[0022] 2. The present invention helps cement enterprises quickly monitor the process status, reveals the underlying data value, assists enterprises and administrative management departments in managing the continuous flue gas emission monitoring system and production process, and timely adjusts the pollution control facilities to keep the cement enterprises operating within the normal operating range.

[0023] 3. The present invention can quickly check the status of a large number of cement discharge ports in a short period of time, improve the effectiveness of the method through an iterative approach, with low cost and high speed. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] Figure 1 is a flow chart of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0025] The following further elaborates on the present invention in conjunction with the Figure 1 drawings and embodiments:

[0026] Embodiment 1

[0027] Step 1. Data extraction, obtaining the monitoring data of the continuous flue gas emission monitoring system at the discharge ports of all cement industry enterprises (set as set c), the data including the unique name of the discharge port, region, monitoring time, oxygen content, NOx concentration data.

[0028] Step 2. Clean and regularize the extracted data, extract the data with oxygen content greater than 0 and less than 16%, and according to expert experience, the oxygen content range of 16% - 21% is recognized as the data of kiln ignition, which is not the case of normal process production. Conduct range verification to remove outliers (extreme values, negative values, etc.).

[0029] Step 3. Determine the value range of NOx, set the minimum value to 0 and the maximum value to x.

[0030] Step 4. Using the oxygen content as the abscissa and the NOx concentration as the ordinate, calculate the probability density function of the cement industry using the non-parametric method of kernel density estimation.

[0031] Step 5. Division of the independent variable value range, the oxygen content is divided at a compensation of 0.1%, divided into 160 intervals; the NOx concentration is divided at a step of 0.1 mg / m³, divided into 10x intervals; thus, the independent variable value range is divided into 160 * 10x independent variable value regions, and each independent variable value region is labeled with a number, encoded from 1 to 1600x.

[0032] Step 6. Use the mathematical integration method to integrate each independent variable value region to obtain the cumulative probability density of each independent variable value region, and obtain 1600x cumulative probability densities.

[0033] Step 7. Sort the 1600x independent variable value regions according to their cumulative probability density. Calculate the independent variable value region with a cumulative probability density of at least 95%, which is called the "independent variable value region greater than 95%".

[0034] Step 8. Obtain the monitoring data of the flue gas continuous monitoring system at the discharge port of a single cement industry enterprise. The data includes the unique name of the discharge port, region, monitoring time, oxygen content, and NOx concentration data. Clean and regularize the extracted data, extract the data with an oxygen content greater than 0 and less than 16%. According to expert experience, the oxygen content range of 16% - 21% is considered data for kiln ignition and not a normal process production situation. Conduct a value range check to remove outliers (extreme values, negative values, etc.). Determine the value range of NOx, set the minimum value to 0, and the maximum value to x. Using the oxygen content as the abscissa and the NOx concentration as the ordinate, calculate the probability density function of the cement industry of a single enterprise using the non-parametric method of kernel density estimation.

[0035] Step 9. Integrate the "independent variable value region greater than 95%" using the mathematical integration method to obtain the cumulative probability density yi of the kiln tail discharge port of this cement enterprise in the "independent variable value region greater than 95%".

[0036] Step 10. Repeat Step 8 and Step 9 to obtain the cumulative probability density yi of the kiln tail discharge ports of all single cement enterprises. This cumulative probability density is the degree of deviation of a single cement enterprise from a normal cement enterprise, that is, the characteristic score of the kiln tail discharge port of this cement enterprise being an abnormal type. The larger this cumulative probability density, the closer it is to the normal type; the smaller this cumulative probability density, the more deviated it is from the normal type, and the more it should be classified as a deviated type.

[0037] Step 11. Sort all the kiln tail discharge ports of single cement enterprises according to their cumulative probability density yi. The lower its cumulative probability density, the more it should be classified as a deviated type, and a prompt is required to further inspect the kiln tail discharge port of this cement enterprise.

[0038] Step 12. Through on-site inspection and online verification, if it is found that there is indeed a problem with the kiln tail discharge port of a single cement enterprise, the kiln tail discharge port of this cement enterprise needs to be removed from the set c, and then repeat Steps 1 to 11 for iteration.

[0039] Step 13. After multiple iterations, the set c obtained after removing the data of the kiln tail discharge ports of deviated cement enterprises is the normal cement industry characteristic data set. Repeat Steps 2 to 4 to obtain the probability density function, which is the normal cement industry characteristic probability density function.

[0040] This specific embodiment is only an interpretation of the present invention and is not a limitation thereof. After reading this specification, those skilled in the art may make modifications to this embodiment that do not contribute creatively as needed, but as long as they are within the scope of the claims of the present invention, they are protected by the patent law.

Claims

1. A method for classifying data features of a continuous monitoring system for flue gas emissions in the cement industry, characterized in that: The following steps are involved: Step 1. Data extraction, obtain set c; Step 2. Clean and organize the extracted data; Step 3. Determine the value range of NOx, set the minimum value to 0 and the maximum value to x; Step 4. Calculate the probability density function of the cement industry with oxygen content as the horizontal axis and NOx concentration as the vertical axis; Step 5. Divide the value range of the independent variable, and use a numerical label for each independent variable value range; Step 6. Integrate each independent variable value region to obtain the cumulative probability density of each independent variable value region; Step 7. Sort the cumulative probability densities from large to small, and obtain the set of independent variable domain rectangles with cumulative probability densities greater than 95% through cyclic calculation; Step 8. Obtain the monitoring data of the flue gas emission continuous monitoring system of a single cement industry enterprise outlet, and calculate the probability density function yi of the cement industry of a single enterprise through the above steps; Step 9. Integrate the "region of independent variable values ​​greater than 95%" to obtain the cumulative probability density yi of the cement enterprise; Step 10. Repeat steps 8 and 9 to obtain the cumulative probability density yi of the kiln tail outlets of all individual cement enterprises. This cumulative probability density is the degree of deviation of a single cement enterprise from a normal cement enterprise, that is, the characteristic score of the kiln tail outlet of the cement enterprise being an abnormal class. When the cumulative probability density deviates from the normal class, it is classified as a deviation class. Step 11. Sort all the kiln outlets of individual cement enterprises according to their cumulative probability density yi, and filter out the deviation class; Step 12. Check the cumulative probability density yi of the deviation class offline and online, remove the deviation data of the cement enterprise kiln tail outlet from the set c, and then repeat steps 1 to 11 for iteration; Step 13. After multiple iterations, the set c obtained after eliminating the kiln outlet data of the deviant cement enterprises is the normal cement industry characteristic data set. Repeat steps 2 to 4 to obtain the probability density function, which is the normal cement industry characteristic probability density function.

2. The method for data feature classification of a continuous monitoring system for flue gas emissions in cement industry according to claim 1 is characterized in that: The data extraction in step 1 is to obtain the monitoring data of the continuous monitoring system of flue gas emissions of all cement industry enterprises as a set c, and the data includes the unique name of the outlet, region, monitoring time, oxygen content, and NOx concentration data.

3. The method for data feature classification of a continuous monitoring system for flue gas emissions in cement industry according to claim 1 is characterized in that: The step 2 specifically extracts data with oxygen content greater than 0 and less than 16%, and identifies the oxygen content range of 16%-21% as kiln ignition data, which is not a normal process production situation. A value range check is performed to remove abnormal values ​​including maximum values ​​and negative values.

4. The method for data feature classification of a continuous monitoring system for flue gas emissions in cement industry according to claim 1 is characterized in that: The step 5 specifically divides the NOx concentration into 10x intervals in steps of 0.1 mg / m3; each independent variable value region is labeled and coded with a number.

5. The method for data feature classification of a continuous monitoring system for flue gas emissions in cement industry according to claim 1 is characterized in that: The step 6 can obtain the probability density distribution of the normal class, and the cumulative probability density of the entire independent variable definition domain is 1.

Images