Industrial waste circulating fluidized bed feeding control method based on stock bin image

The motor controller frequency is optimized through silo image recognition and semantic segmentation algorithm, which solves the problem of inaccurate measurement of silo material level, realizes real-time optimization and automated control of feeding volume, and improves the feeding stability and automation level of industrial waste circulation fluidized beds.

CN120563833APending Publication Date: 2025-08-29龙游县金怡热电有限公司 +1
View PDF 0 Cites 1 Cited by

Patent Information

Application Number
CN202510658447.6
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2025-05-21
Publication Date
2025-08-29

AI Technical Summary

Technical Problem

In the prior art, the material level radar measurement of the silo is inaccurate, and the feeding control cannot be related to the material level changes in real time, resulting in fluctuations in the feeding volume, affecting the working conditions in the furnace, and lacking automated control methods.

Method used

The recognition method based on silo image is adopted, and the level recognition model is established in combination with semantic segmentation algorithm. The silo image is obtained through the monitoring camera, and the evolutionary learning algorithm is used to optimize the frequency of spiral and revolution motor controllers to realize real-time calculation and automatic control of feeding volume.

Benefits of technology

It realizes accurate detection of the material level of the silo and real-time optimization of the feed quantity, improves the degree of automation of feed control, reduces dust interference, and ensures the stability of feed and the stability of the working conditions in the furnace.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN120563833A_ABST
    Figure CN120563833A_ABST
Patent Text Reader

Abstract

The invention discloses an industrial waste circulating fluidized bed feeding control method based on a stock bin image, and the method comprises the steps: obtaining a fuel image in a stock bin through a monitoring camera with light supplement, building a fuel region recognition model through a semantic segmentation algorithm, and obtaining a stock bin material level recognition method. A feeding amount calculation model is established by using feeding amount data under different material levels, different revolution motor controller frequencies and different spiral motor controller frequencies, and the optimal revolution motor controller frequency and the optimal spiral motor controller frequency are output in real time by taking stable feeding as a target. Therefore, the standard feeding amount can be updated in real time, meanwhile, the optimal control scheme under the current material level is optimized in real time, equipment is automatically controlled, and the automation degree of feeding is improved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of industrial waste circulating fluidized bed control, and in particular to an industrial waste circulating fluidized bed feeding control method based on a silo image. Background Art

[0002] Circulating fluidized bed incineration for power generation and heat production is the primary method for treating industrial solid waste. The characteristics of circulating fluidized bed incineration require that incoming fuel be crushed and mixed to achieve a relatively uniform consistency. In relatively stable industrial parks, industrial solid waste thermal power plants / power plants can maintain a relatively stable fuel source. Compared to domestic waste, industrial solid waste has a more stable source, lower moisture content, and higher calorific value. Maintaining a stable feedstock ensures the normal operation of the unit.

[0003] The feed system in these power plants often follows a process flow of crushing, silo mixing, chain conveyor transport, and star valve discharge. Throughout the entire feeding process, the chain conveyor and star valve remain nearly fully open or at a high speed to prevent blockage. Therefore, the primary means of controlling feed is during the silo mixing stage. Fuel accumulates in the silo, and the orbiting feeder stirs and mixes the fuel throughout the silo, while the screw feeder simultaneously transports the fuel to the discharge port. Therefore, the frequency of the orbiting and screw motor controllers is the primary control variable determining feed rate.

[0004] In fact, when adjusting the feeding amount, the material level of the silo is also an extremely important factor. Due to the accumulation of fuel, the stacking density at different material levels is also different. At the same revolution motor controller frequency and spiral motor controller frequency, the feeding amount is also different at different material level heights.

[0005] At present, installing radar ranging in the silo is a conventional technical route, but its results are greatly affected by dust, the measurement results have a large offset, and when controlling the feeding, the influence of the material level is basically not considered, or the feeding level is roughly estimated artificially, resulting in feeding fluctuations and affecting the working conditions in the furnace. Therefore, there is currently a lack of a means to accurately measure the silo material level and a feeding control method that simultaneously considers the material level, orbital feeder, and screw feeder. Summary of the Invention

[0006] The present invention provides a feeding control method for a circulating fluidized bed of industrial waste based on silo images, which solves the problems of inaccurate material level radar measurement in existing silos, the inability of feeding control to correlate material level changes in real time, and the inability to achieve automation. The method has a simple process, comprehensive consideration of influencing factors, and efficient control.

[0007] A feeding control method for a circulating fluidized bed of industrial waste based on a silo image comprises the following steps:

[0008] (1) Based on the video image captured by the surveillance camera with fill light, the material level is calculated based on the silo material level recognition method, which is specifically expressed as: L = model(img), where L is the material level, model(·) is the silo material level recognition formula, and img is the silo image;

[0009] (2) Based on the feed rate calculation model, the feed rate M is calculated according to the material level, the frequency of the screw motor controller, and the frequency of the revolution motor controller. Specifically, it is expressed as: M = f(L, C1, C2), where f(·) is the feed rate calculation model, C1 is the frequency of the screw motor controller, and C2 is the frequency of the revolution motor controller;

[0010] (3) After the combustion in the circulating fluidized bed is stable for at least half an hour (based on the main steam flow rate fluctuating within 5 T / H), the average feed rate within half an hour is calculated and used as the standard feed rate M':

[0011] Among them, M i is the calculated feeding amount at minute i, M i =f(L i ,C1 i ,C2 i );

[0012] (4) Set upper and lower safety limits for C1 and C2, and use evolutionary learning algorithms (such as genetic algorithms, particle optimization algorithms, simulated annealing algorithms, etc.) within the safety limits to obtain the most suitable C1 and C2 values. The most suitable C1 and C2 values ​​are determined by calculating the absolute error between the feed amount M” and the standard feed amount M’, that is, |M”-M’| min ;

[0013] (5) The obtained C1 and C2 values ​​are written into the DCS system through the OPC DA protocol, and then the screw feeder and the orbital feeder are controlled to achieve feeding control.

[0014] In the above technical solution, further, the method of identifying the material level by using the silo material level identification method in step (1) can be more comprehensive than the single-point measurement of the existing radar ranging, and can avoid the error caused by dust on the measurement result (the monitoring camera with fill light can be continuously purged during the shooting process to avoid dust interference); in step (2), the feeding amount calculation model takes the silo material level into consideration, and a more accurate feeding amount can be calculated; in step (3), the standard feeding amount is based on the statistical mean under the actual stable working conditions, and changes according to real-time statistical data, which is more in line with the actual working conditions; in step (4), the feeding amount stability is used as the benchmark, providing a new standard for feeding control.

[0015] Furthermore, the silo level identification method L=model(img) is established as follows:

[0016] (1) Obtain a video of a certain period of time (≥24 hours) in the silo and save the video as an image at 1-second intervals;

[0017] (2) Use an image labeling tool (such as Image Labeler in MATLAB) to mark the fuel part of each image as 1 and the rest as 0, and form a data set;

[0018] (3) The dataset is trained using a semantic segmentation algorithm to obtain a fuel region recognition model, which is expressed as model_S(·). Due to the different colors of materials in the silo and poor lighting conditions, conventional methods (such as edge detection and threshold detection) are completely unable to accurately identify the fuel region. The use of a deep learning algorithm for semantic segmentation can effectively avoid the problem of inaccurate recognition.

[0019] (4) Record the pixel area S of the fuel area when the tank is full max , record the pixel area S of the fuel area when the tank is empty min , record the current fuel area pixel area S current (S current =model_S(img current ), specifically based on the current image img current , obtained by using the fuel area identification model), the silo height is recorded as I, and the calculation formula for the silo level is: Right now

[0020]

[0021] Furthermore, the establishment process of the feed rate calculation model is as follows:

[0022] (1) When the furnace is shut down, conduct a fuel feeding test. min and maximum material level L max Between, evenly select n (specifically 10) test material sites;

[0023] (2) At each material level test point, respectively at the minimum safety screw motor controller frequency C1 min To the maximum safe screw motor controller frequency C1 max Take m1 (specifically 10) test points between the minimum safe revolution motor controller frequency C2 min To the minimum safe revolution motor controller frequency C2 maxTake m2 (specifically 10) test points to obtain test points corresponding to n·m1·m2 working conditions (each test point can last for 1 minute), and record the feeding amount M under each working condition (L, C1, C2) respectively;

[0024] (3) Using different material levels L, screw motor controller frequency C1, and revolution motor controller frequency C2 as variables and the feed rate M as the target value, a feed rate calculation model is established using various methods (such as linear fitting and supervised machine learning algorithms). At different material levels, the density of the silo material will change. Even at the same screw motor controller frequency and the same revolution motor controller frequency, the feed rate will also change. Therefore, when establishing the feed rate calculation model, the material level is also used as a variable parameter.

[0025] The above control method relies on a silo image-based industrial waste circulating fluidized bed feeding control system, which includes:

[0026] The surveillance camera with fill light is facing the materials in the silo and obtains video images of the current materials in the silo through the observation hole directly above the silo. When the light is poor, the camera will automatically turn on the fill light to provide light to ensure the quality of the video image. At the same time, the surveillance camera has a compressed air purge function to avoid the impact of dust.

[0027] Video communication cable, which can be a network cable, optical fiber, or a combination of the two, is used to transmit video images captured by a surveillance camera with fill light to an industrial computer;

[0028] The industrial computer receives the video images inside the silo transmitted by the video communication line, and outputs the frequency setting values ​​of the screw motor controller and the orbital motor controller in real time. It also writes the frequencies of the two motor controllers back to the DCS system through the OPCDA protocol to realize the control of the screw feeder and the orbital feeder, and thus realize the feeding control.

[0029] The present invention solves the problems of inaccurate material level measurement in the silo of the industrial solid waste circulating fluidized bed, the inability of feeding control to be associated with material level changes in real time, and the inability to achieve automation. Through this industrial solid waste circulating fluidized bed feeding control method based on silo images, the material level of the silo can be accurately detected, the standard feeding amount can be updated in real time, and the optimal control scheme under the current material level can be optimized in real time, and the equipment can be automatically controlled.

[0030] Beneficial effects of the present invention:

[0031] The industrial waste circulating fluidized bed feeding control method based on silo images of the present invention obtains the fuel image in the silo through a monitoring camera with fill light, uses a semantic segmentation algorithm to establish a material level recognition model, uses feeding amount experimental data under different material levels, different revolution motor controller frequencies, and different spiral motor controller frequencies to establish a feeding amount calculation model, and with the goal of stable feeding, outputs the optimal revolution motor controller frequency and spiral motor controller frequency in real time, thereby improving the degree of automation of feeding. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] Figure 1 This is a schematic diagram of a circulating fluidized bed feeding device for industrial solid waste;

[0033] In the figure, the silo feed port 1, silo 2, revolution feeder 3, screw feeder 4, monitoring camera with fill light 5, video communication line 6, industrial computer 7, DCS system 8, revolution motor controller 9, screw motor controller 10;

[0034] Figure 2 The flowchart of the feeding control method of the industrial waste circulating fluidized bed based on the silo image is shown;

[0035] Figure 3 Establish a process for silo level identification methods;

[0036] Figure 4 Establish a process for the feed rate calculation model. DETAILED DESCRIPTION

[0037] like Figure 1 As shown, this embodiment provides an industrial solid waste circulating fluidized bed feeding device, including: a silo feed port 1, a silo 2, a revolution feeder 3, a screw feeder 4, a monitoring camera with fill light 5, a video communication line 6, an industrial computer 7, a DCS system 8, a revolution motor controller 9, and a screw motor controller 10. The silo feed port 1 is located at the center position directly above the silo 2. The silo 2 is cylindrical. The revolution feeder 3 is at the bottom of the silo 2. The screw feeder 4 is on the side of the revolution motor. The monitoring camera with fill light 5 is installed on the top of the silo 2. The monitoring camera with fill light 5 and the industrial computer 7 are connected through a video communication line 6. The data processed by the industrial computer 5 is transmitted to the DCS system 8 through the OPCDA protocol. The DCS system 8 controls the frequency of the revolution motor controller 9 and the screw motor controller 10. When the frequency of the revolution motor controller 9 changes, it will control the revolution feeder 3. When the frequency of the screw motor controller 10 changes, it will control the screw feeder 4.

[0038] The embodiment of the present invention further provides an industrial waste circulating fluidized bed feeding control system based on silo images, and its working process is as follows:

[0039] (1) The monitoring camera 5 with fill light is facing the material in the silo 2 and obtains the video image of the current material in the silo 2 through the observation hole directly above the silo 2. When the light is not good, the camera will automatically turn on the fill light to provide light source to ensure the quality of the video image. At the same time, the monitoring camera 5 with fill light has a compressed air purge function to avoid the influence of dust;

[0040] (2) The video signal obtained by the monitoring camera 5 with fill light is transmitted to the industrial computer 7 via the video communication line 6 (which can be a network cable, an optical fiber, or a combination of the two);

[0041] (3) The industrial computer 7 receives the video image signal in the silo transmitted by the video communication line 6, and outputs the frequency setting values ​​of the screw motor controller 10 and the revolution motor controller 9 in real time, and writes the frequencies of the two motor controllers back to the DCS system 8 through the OPCDA protocol. The DCS system 8 adjusts the revolution feeder 3 and the screw feeder 4 by controlling the screw motor controller 10 and the revolution motor controller 9, thereby realizing the control of feeding.

[0042] like Figure 2 The figure shows a feeding control method for a circulating fluidized bed of industrial waste based on a silo image. The specific steps are as follows:

[0043] (1) Based on the video image captured by the surveillance camera with fill light, the material level is calculated based on the silo material level recognition method, which is specifically expressed as: L = model(img), where L is the material level, model(·) is the silo material level recognition formula, and img is the silo image;

[0044] (2) Based on the feed rate calculation model, the feed rate M is calculated according to the material level, the frequency of the screw motor controller, and the frequency of the revolution motor controller. Specifically, it is expressed as: M = f(L, C1, C2), where f(·) is the feed rate calculation model, C1 is the frequency of the screw motor controller, and C2 is the frequency of the revolution motor controller;

[0045] (3) After the combustion in the circulating fluidized bed is stable for at least half an hour (based on the main steam flow rate not fluctuating by more than 5 T / H), the average feed rate within half an hour is calculated and used as the standard feed rate M':

[0046] Among them, M i is the calculated feeding amount at minute i, M i =f(L i ,C1 i ,C2 i );

[0047] (4) Set upper and lower safety limits for C1 and C2, and use evolutionary learning algorithms (such as genetic algorithms, particle optimization algorithms, simulated annealing algorithms, etc.) within the safety limits to obtain the most suitable C1 and C2 values. The most suitable C1 and C2 values ​​are determined by calculating the absolute error between the feed amount M” and the standard feed amount M’, that is, |M”-M’| min ;

[0048] (5) The obtained C1 and C2 values ​​are written from the industrial computer 7 to the DCS system 8 through the OPC DA protocol, and then the screw feeder 4 and the orbital feeder 3 are controlled to realize the control of feeding.

[0049] like Figure 3 As shown, the silo material level identification method L=model(img) is established as follows:

[0050] (1) Obtain a video of a certain period of time (≥24 hours) in the silo and save the video as an image at 1-second intervals;

[0051] (2) Use an image labeling tool (such as Image Labeler in MATLAB) to label the fuel part of each image as 1 and the rest as 0, and form a data set;

[0052] (3) The dataset is trained using semantic segmentation algorithm to obtain the fuel area recognition model, which is expressed as model_S(·);

[0053] (4) Record the area S of the fuel area when the tank is full max , record the fuel area S when the tank is empty min , record the current fuel area S current (S current =model_S(img current ), specifically based on the current image img current , obtained by using the fuel area identification model), the silo height is recorded as I, and the calculation formula for the silo level is: Right now

[0054]

[0055] like Figure 4 As shown, the feed rate calculation model M = f(L, C1, C2) is established as follows:

[0056] (1) When the furnace is shut down, conduct a fuel feeding test. min and maximum material level L max Between, 10 test material sites were evenly selected;

[0057] (2) At each material level test point, respectively at the minimum safety screw motor controller frequency C1 min To the maximum safe screw motor controller frequency C1 max Take 10 test points between the minimum safe revolution motor controller frequency C2 min To the maximum safe revolution motor controller frequency C2 max Take 10 test points, each test point lasts for 1 minute, so as to obtain 1000 test points corresponding to working conditions, and record the feeding amount M under each working condition respectively;

[0058] (3) Taking different material levels L, the frequency C1 of the screw motor controller 10, and the frequency C2 of the revolution motor controller 9 as variables, and the feeding amount M as the target value, various means (such as linear fitting and supervised machine learning algorithms) are used to establish a feeding amount calculation model.

Claims

1. A feeding control method for an industrial waste circulating fluidized bed based on a silo image, characterized in that: The specific steps include: (1) Based on the video image captured by the surveillance camera with fill light, the material level is calculated based on the silo material level recognition method, which is specifically expressed as: L = model(img), where L is the material level, model(·) is the silo material level recognition formula, and img is the silo image; (2) Based on the feed rate calculation model, the feed rate M is calculated according to the material level, the frequency of the screw motor controller, and the frequency of the revolution motor controller. Specifically, it is expressed as: M = f(L, C1, C2), where f(·) is the feed rate calculation model, C1 is the frequency of the screw motor controller, and C2 is the frequency of the revolution motor controller; (3) After the combustion in the circulating fluidized bed is stable for at least half an hour, the average feed amount within half an hour is calculated and used as the standard feed amount M': Among them, M i is the calculated feeding amount at minute i, M i =f(L i ,C1 i ,C2 i ); (4) Set upper and lower safety limits for C1 and C2, and use the evolutionary learning algorithm within the safety limits to obtain the most suitable C1 and C2 values. The most suitable C1 and C2 values ​​are determined by calculating the absolute error between the feed rate M” and the standard feed rate M’, that is, |M”-M’| min ; (5) The obtained C1 and C2 values ​​are written into the DCS system through the OPC DA protocol, and then the screw feeder and the orbital feeder are controlled to achieve feeding control.

2. The method for controlling feeding of an industrial waste circulating fluidized bed based on a silo image according to claim 1, characterized in that: The establishment process of the silo material level identification method is as follows: (1) Obtain at least 24 hours of video footage from the silo and save the video as images at 1-second intervals; (2) Use image labeling tools to mark the fuel part of each image as 1 and the rest as 0, and form a data set; (3) The semantic segmentation algorithm is used to train the dataset to obtain the fuel area recognition model, whose recognition formula is expressed as model_S(·); (4) Record the pixel area S of the fuel area when the tank is full max , record the pixel area S of the fuel area when the tank is empty min , record the current fuel area pixel area S current , S current =model_S(img current ), specifically based on the current image img current , which is obtained by using the fuel area identification model and recording the silo height as I, then the calculation formula for the silo level is:

3. The method for controlling feeding of an industrial waste circulating fluidized bed based on a silo image according to claim 1, wherein: The establishment process of the feed rate calculation model is as follows: (1) When the furnace is shut down, conduct a fuel feeding test. min and maximum material level L max Between, evenly select n material level test points; (2) At each material level test point, respectively at the minimum safety screw motor controller frequency C1 min To the maximum safe screw motor controller frequency C1 max Take m1 test points between the minimum safe revolution motor controller frequency C2 min To the maximum safe revolution motor controller frequency C2 max Take m2 test points between them to obtain the test points corresponding to n·m1·m2 working conditions, and record the feeding amount M under each working condition (L, C1, C2) respectively; (3) Taking different material levels L, screw motor controller frequency C1, and revolution motor controller frequency C2 as variables and feeding amount M as the target value, a feeding amount calculation model is established.

4. The method for controlling feeding of an industrial waste circulating fluidized bed based on a silo image according to claim 1, wherein: The criterion for determining whether the combustion in the circulating fluidized bed is stable is that the main steam flow rate does not fluctuate by more than 5 T / H.

5. The method for controlling feeding of an industrial waste circulating fluidized bed based on a silo image according to any one of claims 1 to 4, characterized in that: The system relies on a silo image-based industrial waste circulating fluidized bed feeding control system; the silo image-based industrial waste circulating fluidized bed feeding control system includes: A surveillance camera with fill light is positioned directly against the material in the silo to capture a video image of the material in the silo through an observation hole directly above the silo. The surveillance camera with fill light also has a compressed air purge function, which is used to purge during filming to avoid the effects of dust. Video communication cable, used to transmit video images captured by a surveillance camera with fill light to an industrial computer; The industrial computer is used to receive video images transmitted by the video communication line and output the frequency setting values ​​of the screw motor controller and the orbital motor controller in real time. It also writes the frequencies of the two motor controllers back to the DCS system through the OPCDA protocol to realize the control of the screw feeder and the orbital feeder, and thus realize the feeding control.

Citation Information

Cited By

  • Treatment process for preparing microporous carbon from waste sulfuric acid

    CN121180974A