Coal as fired quantity real-time detection method based on electrostatic method and BP neural network

Through the full-section non-contact electrostatic method and the BP neural network model, combining the electrostatic signal strength, coal powder flow rate and temperature, the measurement error and hysteresis of the coal inlet volume are solved, real-time and accurate detection of the coal inlet volume is achieved, and the improvement of the coal inlet volume is improved. The unit variable load control rate and energy utilization rate are provided.

CN119086374APending Publication Date: 2024-12-06NORTH CHINA ELECTRIC POWER UNIV
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202411105310.X
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-08-13
Publication Date
2024-12-06

AI Technical Summary

Technical Problem

The prior art has large errors when measuring the amount of coal entering the furnace, especially affected by the coal pile humidity, surface state and light source environment, and the measurement results are hysteresis, making it difficult to meet the optimization needs of unit coordination control.

Method used

The full-section non-contact electrostatic method is used to measure the electrostatic signal intensity, coal powder flow rate and temperature, combined with the BP neural network model, a real-time detection method of the amount of coal entering the furnace is established, and the auxiliary variables are verified through data noise reduction and normalization processing. and use the average relative error, average absolute error, and root mean square error as model evaluation index.

Benefits of technology

Real-time and accurate detection of the amount of coal in furnace is achieved, the response characteristics of coal in furnace in dynamic situations are optimized, and the variable load control performance and energy utilization rate of the unit are improved.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119086374A_ABST
    Figure CN119086374A_ABST
Patent Text Reader

Abstract

The invention belongs to the technical field of unit coal as fired quantity real-time measurement, and particularly relates to a coal as fired quantity real-time detection method based on an electrostatic method and a BP neural network, and the method comprises the steps: selecting auxiliary variables in a BP neural network model, and the auxiliary variables comprise electrostatic signal intensity, pulverized coal flow velocity and temperature; measuring an air powder parameter value through a total cross-section non-contact electrostatic method, and verifying the rationality of auxiliary variable selection by calculating and analyzing the Pearson correlation between the auxiliary variable and the coal as fired; performing data noise reduction processing and normalization processing to establish a BP neural network model; and comparing the coal feeding quantity under the stable load and the variable load with the coal as fired quantity predicted by the BP neural network, and introducing an average relative error, an average absolute error and a root-mean-square error as evaluation indexes of the model to verify the accuracy. The real-time accurate detection of the coal as fired is realized, the control performance of response load and variable load is improved, and the control rate and the energy utilization rate of the variable load of the unit are improved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The invention belongs to the technical field of real-time measurement of coal quantity fed into a furnace of a unit, and in particular relates to a real-time detection method of coal quantity fed into a furnace based on an electrostatic method and a BP neural network. Background Art

[0002] In the coordinated control of the unit, accurate detection of the amount of coal entering the furnace and optimization of the response characteristics are of great significance to improving the load control performance of the unit. If the changes in the amount of coal entering the furnace can be accurately detected in real time, the coordinated control of the boiler can be optimized by reasonably adjusting the parameters such as the primary air volume and primary air pressure, improving the control performance of the response unit load and variable load, and improving the control rate of the unit variable load.

[0003] However, the actual coal quantity signal measurement point on site is generally located before the coal mill, and the uncertainty of the pulverizing system makes it difficult to determine the real-time coal quantity entering the furnace. In recent years, various coal quantity detection technologies and products have been continuously developed and have been applied in some domestic power plants. However, their measurement accuracy and reliability still cannot meet the requirements of the current optimization unit coordinated control system.

[0004] The following are common measurement methods in the prior art:

[0005] ●Acoustic wave measurement method

[0006] The basic idea of ​​acoustic wave measurement is that when the acoustic wave signal is transmitted through the coal pile, the speed of sound is a function of the solid phase concentration. According to the speed and amplitude of the acoustic wave propagation, the density and shape of the coal pile can be estimated. The advantage of the acoustic wave measurement method is that it can be applied to coal piles of various irregular shapes. The disadvantage is that the sound propagation speed is fast, the humidity and surface state of the coal pile have a greater impact, and the measurement will produce a large error, so it is rarely used in practice.

[0007] Total light scattering method

[0008] The principle of the total light scattering method is that when a light beam passes through an experimental medium, the solid particles in the medium will scatter and absorb the light beam, causing the intensity of the light beam to decrease. The mass of the solid particles can be measured by the degree of attenuation of the light beam intensity after passing through the medium. The total light scattering method is widely used in offline measurements in laboratories, but in actual measurements, the measurement results will also be greatly affected because the light source is greatly affected by the environment.

[0009] Thermal balance method

[0010] The heat balance method is based on the principle of heat conservation before and after the primary air and coal powder are mixed, assuming that the specific heat capacity of the coal powder is relatively fixed. The temperature and specific heat capacity of the primary air and coal powder before and after the mixing are calculated to obtain the coal volume at the powder pipe outlet. This method may cause errors when the coal type changes or the coal types are mixed, and the measurement results will have a certain lag due to the rapid temperature change of the air-powder mixture.

[0011] Since the existing technology is greatly affected by the humidity and surface state of the coal pile, the measurement will produce large errors; the light source is greatly affected by the environment, which has a great impact on the measurement results; since the temperature of the air-powder mixture changes rapidly, the measurement results have a lag problem. There is an urgent need for a real-time detection method for the amount of coal entering the furnace based on the electrostatic method and BP neural network, which can realize real-time and accurate detection of the amount of coal entering the furnace based on the electrostatic method, optimize the response characteristics of the coal entering the furnace under dynamic conditions, improve the control performance of the response load and variable load, and improve the control rate and energy utilization rate of the unit's variable load. Summary of the invention

[0012] The object of the present invention is to provide a real-time detection method for the amount of coal fed into a furnace based on an electrostatic method and a BP neural network, characterized in that the method comprises the following steps:

[0013] Step S1: selecting auxiliary variables in the BP neural network model, wherein the auxiliary variables include: electrostatic signal intensity, coal powder flow rate and temperature;

[0014] Step S2: The air-powder parameter value of the primary air-powder duct at the outlet of the coal mill is measured by a full-section non-contact electrostatic method, and the rationality of the selection of the auxiliary variables is verified by calculating and analyzing the Pearson correlation between the auxiliary variables and the amount of coal entering the furnace;

[0015] Step S3: performing data noise reduction and normalization processing;

[0016] Step S4: Establishing a BP neural network model between the amount of coal entering the furnace and the electrostatic signal intensity, coal powder flow rate, and temperature;

[0017] Step S5: Compare the coal feeding amount under stable load and variable load with the coal feeding amount predicted by BP neural network, and introduce the average relative error, average absolute error and root mean square error as evaluation indicators of the model to verify the accuracy of the real-time detection method of coal feeding amount.

[0018] The Pearson correlation between the auxiliary variables and the amount of coal fed into the furnace in the calculation and analysis in step S2 includes:

[0019] The Pearson correlation coefficient between the electrostatic signal intensity and the amount of coal fed into the furnace is 0.727;

[0020] The Pearson correlation coefficient between the electrostatic signal intensity and the coal powder flow rate is 0.863;

[0021] The Pearson correlation coefficient between electrostatic signal intensity and temperature is 0.689.

[0022] The data noise reduction process in step S3 adopts a sliding average filtering method.

[0023] The normalization process in step S3 is:

[0024]

[0025] Among them, x and y represent the data before and after normalization, respectively. min and x max are the minimum and maximum values ​​in the sample data, respectively.

[0026] The basic function of the BP neural network model in step S4 is:

[0027] Q=f(E,V,T)

[0028] Among them, Q represents the amount of coal entering the furnace, E represents the electrostatic signal intensity, V represents the coal powder flow rate, and T represents the primary air duct temperature.

[0029] The average relative error in step S5 is:

[0030]

[0031] The mean absolute error in step S5 is:

[0032]

[0033] The root mean square error in step S5 is:

[0034]

[0035] Wherein, MAPE is the mean relative error, MAE is the mean absolute error, and RMSE is the root mean square error.

[0036] The beneficial effects of the present invention are:

[0037] Based on the full-section non-contact electrostatic measurement technology, combined with the BP neural network model, the real-time and accurate detection of the amount of coal entering the furnace in the primary air duct of the unit is realized. By selecting the principle that the electrostatic measurement signal of the wind powder in the primary air duct, the wind powder flow rate in the primary air duct, the mill outlet temperature and other parameters are correlated with the amount of coal entering the furnace, a coal entering the furnace identification model based on the BP neural network is established, and the real-time and accurate detection of the amount of coal entering the furnace based on the electrostatic method is realized, which plays a significant role in optimizing the response characteristics of the coal entering the furnace under dynamic conditions, improving the control performance of the response load and variable load, and improving the control rate of the unit's variable load and energy utilization. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] Figure 1 This is a flow chart of a method for real-time detection of coal quantity entering a furnace based on an electrostatic method and a BP neural network according to the present invention;

[0039] Figure 2 This is a schematic diagram of auxiliary variable data before sliding average filtering in an embodiment of the present invention;

[0040] Figure 3 This is a schematic diagram of auxiliary variable data after sliding average filtering in an embodiment of the present invention;

[0041] Figure 4 This is a comparison chart of the actual value of the amount of coal fed into the furnace under 60% stable load and the amount of coal fed into the furnace predicted by the BP neural network in an embodiment of the present invention;

[0042] Figure 5 This is a comparison chart of the actual value of the amount of coal entering the furnace at 60% to 80% load and the amount of coal entering the furnace predicted by the BP neural network in an embodiment of the present invention. DETAILED DESCRIPTION

[0043] The present invention provides a real-time detection method for the amount of coal fed into a furnace based on an electrostatic method and a BP neural network. The present invention will be further described in detail below in conjunction with the accompanying drawings.

[0044] like Figure 1 The embodiment of the present invention shown discloses a real-time measurement method for the air-powder flow rate in the primary air-powder pipeline of a thermal power plant, specifically a real-time detection method for the amount of coal entering the furnace based on an electrostatic method and a BP neural network, comprising the following steps:

[0045] Step S1: selecting auxiliary variables in the BP neural network model, wherein the auxiliary variables include: electrostatic signal intensity, coal powder flow rate and temperature;

[0046] Step S2: The air-powder parameter value of the primary air-powder duct at the outlet of the coal mill is measured by a full-section non-contact electrostatic method, and the rationality of the selection of the auxiliary variables is verified by calculating and analyzing the Pearson correlation between the auxiliary variables and the amount of coal entering the furnace;

[0047] Step S3: performing data noise reduction and normalization processing;

[0048] Step S4: Establishing a BP neural network model between the amount of coal entering the furnace and the electrostatic signal intensity, coal powder flow rate, and temperature;

[0049] Step S5: Compare the coal feeding amount under stable load and variable load with the coal feeding amount predicted by BP neural network, and introduce the average relative error, average absolute error and root mean square error as evaluation indicators of the model to verify the accuracy of the real-time detection method of coal feeding amount.

[0050] The specific implementation process of each step is described in detail below.

[0051] Step S1: selecting auxiliary variables in the BP neural network model, wherein the auxiliary variables include: electrostatic signal intensity, coal powder flow rate and temperature;

[0052] In this embodiment, the auxiliary variables in the BP neural network model are selected. According to the correlation between the amount of coal entering the furnace and the air-powder parameters in the primary air duct, the electrostatic signal intensity, coal powder flow rate, and temperature are comprehensively selected as auxiliary variables for establishing the coal amount detection model.

[0053] In this embodiment, the selection of auxiliary variables in the BP neural network model specifically includes:

[0054] First, in the primary air duct, friction and collision occur continuously between coal powder particles and between coal powder and air and the inner wall of the duct. Coal powder particles carry static charge, so the flow rate of coal powder can be reflected by measuring the strength of electrostatic signal. Therefore, the strength of electrostatic signal is selected as the auxiliary variable for establishing the real-time detection model of coal entering the furnace.

[0055] Secondly, the mass flow rate of coal entering the furnace is equal to the product of the pipeline cross-sectional area, coal powder density, coal powder flow rate and coal powder concentration. The amount of coal entering the furnace is directly related to the coal powder flow rate. The greater the coal powder flow rate, the greater the amount of coal entering the furnace. Therefore, the coal powder flow rate is selected as the auxiliary variable for establishing the real-time detection model of the coal amount entering the furnace.

[0056] Finally, the temperature of the primary air duct will change the charge of the coal powder particles. As the temperature rises, the thermal motion of the coal powder particles and air molecules intensifies, and the frequency of collision and friction increases, thereby enhancing the charging ability of the particles and increasing the intensity of the electrostatic signal. In addition, the change in temperature will affect the relative humidity of the air. At high temperatures, the relative humidity of the air is low, and the accumulation of static electricity is enhanced, while at low temperatures, the opposite is true. Therefore, temperature is selected as an auxiliary variable for the establishment of a real-time detection model for the amount of coal entering the furnace.

[0057] In summary, according to the correlation between the amount of coal entering the furnace and the air-powder parameters in the primary air duct, the electrostatic signal intensity, coal powder flow rate and temperature were selected as auxiliary variables for establishing the coal amount detection model.

[0058] Step S2: The air-powder parameter value of the primary air-powder duct at the outlet of the coal mill is measured by a full-section non-contact electrostatic method, and the rationality of the selection of the auxiliary variables is verified by calculating and analyzing the Pearson correlation between the auxiliary variables and the amount of coal entering the furnace;

[0059] The Pearson correlation between the auxiliary variables and the amount of coal fed into the furnace in the calculation and analysis in step S2 includes:

[0060] The Pearson correlation coefficient between the electrostatic signal intensity and the amount of coal fed into the furnace is 0.727;

[0061] The Pearson correlation coefficient between the electrostatic signal intensity and the coal powder flow rate is 0.863;

[0062] The Pearson correlation coefficient between electrostatic signal intensity and temperature is 0.689.

[0063] In this embodiment, the air-powder parameter value of the primary air-powder pipeline at the outlet of the coal mill is measured by a full-section non-contact electrostatic method, and the required auxiliary variables are obtained after calculation. The rationality of the selection of auxiliary variables is verified by calculating and analyzing the Pearson correlation between the electrostatic signal intensity, coal powder flow rate, temperature and the amount of coal entering the furnace.

[0064] The air-powder parameters of the primary air-powder duct at the outlet of the coal mill were measured by the full-section non-contact electrostatic method, and the required raw data of the auxiliary variables were obtained after calculation. The rationality of the selection of auxiliary variables was verified by calculating and analyzing the Pearson correlation between the electrostatic signal intensity, coal powder flow rate, temperature and the amount of coal entering the furnace.

[0065] The Pearson correlation calculation formula is:

[0066]

[0067] The Pearson correlation coefficient is expressed as:

[0068]

[0069] Step S3: performing data noise reduction and normalization processing;

[0070] The data noise reduction process in step S3 adopts a sliding average filtering method.

[0071] The normalization process in step S3 is:

[0072]

[0073] Among them, x and y represent the data before and after normalization, respectively.min and x max are the minimum and maximum values ​​in the sample data, respectively.

[0074] In this embodiment, data noise reduction processing and normalization processing are performed.

[0075] Data noise reduction and normalization are performed, and the noise reduction method here adopts the sliding average filtering method.

[0076] Sliding average filtering method: It regards N consecutive sampling values ​​as a queue, and the length of the queue is fixed to N. Each time a new data is sampled, it is put into the end of the queue, and the original data at the head of the queue is discarded (first-in, first-out principle). The arithmetic average operation of the N data in the queue can obtain a new filtering result. The sliding average filtering method has a good suppression effect on periodic interference and has high smoothness.

[0077] Comparison of the original data of electrostatic signal intensity, coal powder flow rate and temperature with those after sliding average filtering:

[0078] Before filtering Figure 2 As shown, after filtering Figure 3 As shown, yellow is the electrostatic signal intensity, green is the coal powder flow rate, and blue is the temperature.

[0079] The normalization formula:

[0080]

[0081] Among them, x and y represent the data before and after normalization, respectively. min and x max are the minimum and maximum values ​​in the sample data, respectively.

[0082] Step S4: Establishing a BP neural network model between the amount of coal entering the furnace and the electrostatic signal intensity, coal powder flow rate, and temperature;

[0083] The basic function of the BP neural network model in step S4 is:

[0084] Q=f(E,V,T)

[0085] Among them, Q represents the amount of coal entering the furnace, E represents the electrostatic signal intensity, V represents the coal powder flow rate, and T represents the primary air duct temperature.

[0086] In this embodiment, based on the principle that the amount of coal entering the furnace is approximately equal to the amount of coal fed under stable load, a BP neural network model is established between the amount of coal entering the furnace and the electrostatic signal intensity, coal powder flow rate, and temperature.

[0087] According to the principle that the amount of coal entering the furnace under stable load is approximately equal to the amount of coal fed, a BP neural network model is established between the amount of coal entering the furnace and the electrostatic signal intensity, coal powder flow rate, and temperature. The established BP neural network model can be used to predict the amount of coal entering the furnace under stable load and unstable load.

[0088] Basic model functions:

[0089] Q=f(E,V,T)

[0090] Among them, Q represents the amount of coal entering the furnace, E represents the electrostatic signal intensity, V represents the coal powder flow rate, and T represents the primary air duct temperature.

[0091] Step S5: Compare the coal feeding amount under stable load and variable load with the coal feeding amount predicted by BP neural network, and introduce the average relative error, average absolute error and root mean square error as evaluation indicators of the model to verify the accuracy of the real-time detection method of coal feeding amount.

[0092] The average relative error in step S5 is:

[0093]

[0094] The mean absolute error in step S5 is:

[0095]

[0096] The root mean square error in step S5 is:

[0097]

[0098] Wherein, MAPE is the mean relative error, MAE is the mean absolute error, and RMSE is the root mean square error.

[0099] In this embodiment, the coal feed rate under stable load and variable load and the coal entering the furnace predicted by the BP neural network are compared respectively, and the average relative error, average absolute error and root mean square error are introduced as evaluation indicators of the model to verify the accuracy of the coal entering the furnace detection method.

[0100] The actual value of the coal amount entering the furnace under stable load and variable load and the coal amount entering the furnace predicted by the BP neural network are compared respectively, and the average relative error, mean absolute error and root mean square error are introduced as evaluation indicators of the model to verify the accuracy of the coal amount detection method.

[0101] The formulas for mean relative error MAPE, mean absolute error MAE, and root mean square error RMSE are:

[0102]

[0103] like Figure 4As shown, the actual value of the amount of coal entering the furnace under 60% stable load is compared with the amount of coal entering the furnace predicted by the BP neural network: among them, the yellow curve represents the actual value of the amount of coal entering the furnace, and the blue curve represents the predicted value of the BP neural network.

[0104] Mean relative error Mean absolute error Root mean square error 60% stable load 0.0088 0.6237 0.7049

[0105] like Figure 5 As shown, the actual value of the amount of coal entering the furnace under 60% to 80% load and the amount of coal entering the furnace predicted by the BP neural network are compared: among them, the yellow curve represents the actual value of the amount of coal entering the furnace, and the blue curve represents the predicted value of the BP neural network.

[0106] Mean relative error Mean absolute error Root mean square error 60% to 80% liter load 0.0317 2.395 3.138

[0107] In summary, through the above experimental data verification, during the operation of the boiler unit in the power station, due to the influence of various reasons such as the complexity of the measurement environment and the weak adaptability of the device, most of the existing technical methods remain at the theoretical level and have not been widely promoted and applied in power plants. How to accurately and timely measure the amount of coal entering the boiler has always been one of the difficult problems that researchers and engineering and technical personnel have been working hard to solve. The real-time detection method of the amount of coal entering the furnace based on the electrostatic method and the BP neural network disclosed in the present invention can realize the real-time and accurate detection of the amount of coal entering the furnace based on the electrostatic method, which plays a significant role in optimizing the response characteristics of the coal entering the furnace under dynamic conditions, improving the control performance of the response load and variable load, and improving the control rate of the unit variable load and energy utilization.

Claims

1. A real-time detection method for the amount of coal entering a furnace based on an electrostatic method and a BP neural network, characterized in that: The steps include: Step S1: selecting auxiliary variables in the BP neural network model, wherein the auxiliary variables include: electrostatic signal intensity, coal powder flow rate and temperature; Step S2: The air-powder parameter value of the primary air-powder duct at the outlet of the coal mill is measured by a full-section non-contact electrostatic method, and the rationality of the selection of the auxiliary variables is verified by calculating and analyzing the Pearson correlation between the auxiliary variables and the amount of coal entering the furnace; Step S3: performing data noise reduction and normalization processing; Step S4: Establishing a BP neural network model between the amount of coal entering the furnace and the electrostatic signal intensity, coal powder flow rate, and temperature; Step S5: Compare the coal feeding amount under stable load and variable load with the coal feeding amount predicted by BP neural network, and introduce the average relative error, average absolute error and root mean square error as evaluation indicators of the model to verify the accuracy of the real-time detection method of coal feeding amount.

2. The real-time detection method of coal quantity into furnace based on electrostatic method and BP neural network according to claim 1 is characterized in that: The Pearson correlation between the auxiliary variables and the amount of coal fed into the furnace in the calculation and analysis in step S2 includes: The Pearson correlation coefficient between the electrostatic signal intensity and the amount of coal fed into the furnace is 0.727; The Pearson correlation coefficient between the electrostatic signal intensity and the coal powder flow rate is 0.863; The Pearson correlation coefficient between electrostatic signal intensity and temperature is 0.

689.

3. The real-time detection method of coal quantity into furnace based on electrostatic method and BP neural network according to claim 1 is characterized in that: The data noise reduction process in step S3 adopts a sliding average filtering method.

4. The real-time detection method of coal quantity into furnace based on electrostatic method and BP neural network according to claim 1 is characterized in that: The normalization process in step S3 is: Among them, x and y represent the data before and after normalization, respectively. min and x max are the minimum and maximum values ​​in the sample data, respectively.

5. The real-time detection method of coal quantity in a furnace based on electrostatic method and BP neural network according to claim 1 is characterized in that: The basic function of the BP neural network model in step S4 is: Q=f(E,V,T) Among them, Q represents the amount of coal entering the furnace, E represents the electrostatic signal intensity, V represents the coal powder flow rate, and T represents the primary air duct temperature.

6. The real-time detection method of coal quantity fed into furnace based on electrostatic method and BP neural network according to claim 1 is characterized in that: The average relative error in step S5 is: The mean absolute error in step S5 is: The root mean square error in step S5 is: Wherein, MAPE is the mean relative error, MAE is the mean absolute error, and RMSE is the root mean square error.