A combustion optimization control method and system for a waste incinerator based on machine vision

Abstract

The present invention discloses a combustion optimization control method and system for a waste incinerator based on machine vision, which relates to the technical field of waste incineration. The method comprises: capturing multispectral flame images of each wind chamber in the waste incinerator; generating a panoramic flame image of each wind chamber according to the multispectral flame images of each wind chamber; determining the combustion state of the flame in each wind chamber and the air flow state between each wind chamber based on the recognition result of the flame area; determining the residence state of the garbage in each wind chamber based on the recognition result of the garbage area; adjusting the air supply parameters of each wind chamber according to the combustion state of the flame in each wind chamber, the air flow state between each wind chamber and the residence state of the garbage in each wind chamber; real-time monitoring of the multispectral flame images of each wind chamber in the waste incinerator after adjustment, and predicting the future state of each wind chamber in the waste incinerator through a deep learning network, so as to optimize the air volume distribution and combustion coordinated control parameters between each wind chamber.

Description

Technical Field

[0001] The present application relates to the technical field of waste incineration, and in particular to a combustion optimization control method and system for a waste incinerator based on machine vision. Background Art

[0002] As a key facility for treating municipal solid waste, the efficient and stable operation of waste incinerators is crucial for environmental protection and energy recovery. However, traditional waste incinerators face numerous challenges during the combustion process, including low combustion efficiency, difficulty in controlling pollutant emissions, and complex operation.

[0003] Traditional waste incinerator control methods rely primarily on manual operation and empirical judgment, which is subject to significant subjectivity and error. Operators must constantly monitor the incinerator's flame image and make adjustments based on factors such as flame color and combustion status. This method is not only labor-intensive but also difficult to achieve real-time, accurate control.

[0004] To address the above-mentioned problems, no effective solutions have been proposed so far. Summary of the Invention

[0005] The embodiments of the present application provide a combustion optimization control method and system for a waste incinerator based on machine vision to solve the above-mentioned technical problems.

[0006] The present application provides a combustion optimization control method for a waste incinerator based on machine vision, comprising: capturing a multispectral flame image of each wind chamber in the waste incinerator; generating a flame panoramic image of each wind chamber based on the multispectral flame image of each wind chamber; using an adaptive threshold segmentation algorithm to perform flame area identification and garbage area identification on the flame panoramic image of each wind chamber; determining the combustion state of the flame in each wind chamber and the air flow state between each wind chamber based on the identification result of the flame area; determining the residence state of the garbage in each wind chamber based on the identification result of the garbage area; adjusting the air supply parameters of each wind chamber based on the combustion state of the flame in each wind chamber, the air flow state between each wind chamber and the residence state of the garbage in each wind chamber; real-time monitoring of the adjusted multispectral flame images of each wind chamber in the waste incinerator, and predicting the future state of each wind chamber in the waste incinerator through a deep learning network, so as to optimize the air volume distribution and combustion coordinated control parameters between each wind chamber.

[0007] The present application provides a combustion optimization control system for a waste incinerator based on machine vision, comprising: a flame image capture module for capturing multispectral flame images of each wind chamber in the waste incinerator; a flame panoramic image generation module for generating a flame panoramic image of each wind chamber based on the multispectral flame images of each wind chamber; a flame region recognition module for performing flame region recognition and garbage region recognition on the flame panoramic image of each wind chamber using an adaptive threshold segmentation algorithm; a wind chamber state determination module for determining the combustion state of the flame in each wind chamber and the air flow state between the wind chambers based on the flame region recognition results; and determining the residence state of garbage in each wind chamber based on the garbage region recognition results; a wind chamber air supply parameter adjustment module for adjusting the air supply parameters of each wind chamber based on the combustion state of the flame in each wind chamber, the air flow state between the wind chambers, and the residence state of garbage in each wind chamber; and a prediction optimization module for real-time monitoring the adjusted multispectral flame images of each wind chamber in the waste incinerator, and predicting the future state of each wind chamber in the waste incinerator through a deep learning network, so as to optimize the air volume distribution between the wind chambers and the combustion coordinated control parameters.

[0008] Based on the embodiments provided in the present application, by capturing the multispectral flame images of each wind chamber in the waste incinerator and generating a flame panoramic image, the combustion state and garbage retention state of each wind chamber can be comprehensively and accurately reflected; the flame area and garbage area of the flame panoramic image are identified by using an adaptive threshold segmentation algorithm, which can effectively overcome the subjectivity and error problems of traditional manual identification methods and improve the accuracy and reliability of identification; based on the identification results, the combustion state, air leakage state and garbage retention state of the flame in each wind chamber can be accurately determined, thereby providing a scientific basis for adjusting the air supply parameters of each wind chamber and realizing precise control of the waste incinerator; in addition, by real-time monitoring of the adjusted multispectral flame image and using a deep learning network to predict the future state of each wind chamber, the air volume distribution and combustion coordinated control parameters between the wind chambers can be further optimized, the combustion efficiency and stability of the waste incinerator can be improved, and pollutant emissions can be reduced, with significant economic and environmental benefits. BRIEF DESCRIPTION OF THE DRAWINGS

[0009] The drawings described herein are used to provide a further understanding of the embodiments of the present invention and constitute a part of this application. The illustrative embodiments of this application and their descriptions are used to explain this application and do not constitute an improper limitation of this application. In the drawings:

[0010] Figure 1 This is a flow chart of an optional method for optimizing combustion control of a waste incinerator based on machine vision according to an embodiment of the present application;

[0011] Figure 2This is a structural diagram of an optional machine vision-based combustion optimization control system for a waste incinerator according to an embodiment of the present application.

[0012] The purpose, features and advantages of the present invention will be further described with reference to the accompanying drawings and in conjunction with the embodiments. DETAILED DESCRIPTION

[0013] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0014] Alternatively, as Figure 1 As shown, the present application provides a combustion optimization control method for a waste incinerator based on machine vision, comprising:

[0015] S101, capturing multispectral flame images of each wind chamber in the waste incinerator;

[0016] S102, generating a panoramic flame image of each wind chamber based on the multispectral flame image of each wind chamber;

[0017] S103, using an adaptive threshold segmentation algorithm to perform flame area recognition and garbage area recognition on the flame panoramic image of each wind chamber;

[0018] S104: Based on the identification results of the flame area, the burning state of the flame in each air chamber and the air flow state between the air chambers are determined; based on the identification results of the garbage area, the retention state of the garbage in each air chamber is determined;

[0019] S105, adjusting the air supply parameters of each air chamber according to the burning state of the flame in each air chamber, the air flow state between the air chambers, and the retention state of the garbage in each air chamber;

[0020] S106, real-time monitoring of the multispectral flame images of each wind chamber in the waste incinerator after adjustment, and prediction of the future state of each wind chamber in the waste incinerator through a deep learning network to optimize the air volume distribution and combustion coordinated control parameters between the wind chambers.

[0021] Based on the embodiments provided in the present application, by capturing the multispectral flame images of each wind chamber in the waste incinerator and generating a flame panoramic image, the combustion state and garbage retention state of each wind chamber can be comprehensively and accurately reflected; the flame area and garbage area of the flame panoramic image are identified by using an adaptive threshold segmentation algorithm, which can effectively overcome the subjectivity and error problems of traditional manual identification methods and improve the accuracy and reliability of identification; based on the identification results, the combustion state, air leakage state and garbage retention state of the flame in each wind chamber can be accurately determined, thereby providing a scientific basis for adjusting the air supply parameters of each wind chamber and realizing precise control of the waste incinerator; in addition, by real-time monitoring of the adjusted multispectral flame image and using a deep learning network to predict the future state of each wind chamber, the air volume distribution and combustion coordinated control parameters between the wind chambers can be further optimized, the combustion efficiency and stability of the waste incinerator can be improved, and pollutant emissions can be reduced, with significant economic and environmental benefits.

[0022] Alternatively, as Figure 2 As shown, the present application provides a combustion optimization control system for a waste incinerator based on machine vision, comprising:

[0023] The flame image capturing module 201 is used to capture the multi-spectral flame image of each wind chamber in the waste incinerator;

[0024] The flame panoramic image generating module 202 is used to generate a flame panoramic image of each wind chamber based on the multispectral flame image of each wind chamber;

[0025] The flame region identification module 203 is used to identify the flame region and the garbage region of the flame panoramic image of each wind chamber using an adaptive threshold segmentation algorithm;

[0026] The air chamber state determination module 204 is used to determine the combustion state of the flame in each air chamber and the air flow state between the air chambers based on the identification result of the flame area; and to determine the retention state of the garbage in each air chamber based on the identification result of the garbage area;

[0027] The air supply parameter adjustment module 205 is used to adjust the air supply parameters of each air chamber according to the combustion state of the flame in each air chamber, the air flow state between each air chamber, and the retention state of garbage in each air chamber;

[0028] The prediction and optimization module 206 is used to monitor the multispectral flame images of each wind chamber in the waste incinerator in real time after adjustment, and predict the future state of each wind chamber in the waste incinerator through a deep learning network to optimize the air volume distribution and combustion coordinated control parameters between the wind chambers.

[0029] The air supply parameters for each plenum primarily focus on the air supply characteristics within a single plenum. These parameters refer to specific air supply indicators set for each plenum, such as air volume, air pressure, and wind speed. These parameters determine the air supply intensity and effectiveness of a single plenum, directly impacting the combustion of waste within that plenum. For example, when waste in a particular plenum burns vigorously, the air supply volume may need to be increased to provide sufficient oxygen to support combustion, or the air pressure may need to be adjusted to allow the wind to penetrate deeper into the waste layer, promoting even combustion.

[0030] The air volume distribution and combustion coordinated control parameters between each air chamber mainly focus on the coordinated relationship and overall optimization between multiple air chambers. The air volume distribution and combustion coordinated control parameters between each air chamber include the coordination of the air volume distribution ratio and the combustion control strategy. The air volume distribution ratio refers to the relative size relationship of the air volume between different air chambers to achieve the balance of the overall combustion and maximize the efficiency; the combustion coordinated control parameters involve the coordinated control of multiple air chambers during the combustion process, such as synchronously adjusting the air supply parameters and combustion strategies of each air chamber according to the distribution and combustion status of the garbage to achieve the best overall combustion effect. For example, at a certain stage of the garbage incinerator, it may be necessary to allocate more air volume to the air chamber where combustion is more difficult, while reducing the air volume of other air chambers to concentrate resources to solve the combustion problem; or after the garbage enters the furnace, the air supply parameters of each air chamber are dynamically adjusted according to the movement and combustion status of the garbage, so that the combustion process smoothly transitions between different air chambers to avoid local overheating or incomplete combustion.

[0031] The air supply parameters for each plenum, as well as the air volume distribution and combustion coordination control parameters between plenums, describe the system's optimized control strategy for the incinerator's combustion process from both a local and global perspective. By simultaneously considering the air supply characteristics of individual plenums and the coordination between multiple plenums, more comprehensive combustion optimization control can be achieved, improving the incinerator's combustion efficiency and stability, reducing pollutant emissions, and extending the equipment's service life.

[0032] Furthermore, the flame image capturing module includes at least two light sources of different wavelengths; wherein the at least two light sources of different wavelengths include a light source of visible light band and a light source of near infrared band;

[0033] The flame image capture module is used to capture images of each wind chamber in the waste incinerator under different spectra using at least two light sources with different wavelengths, thereby obtaining a multispectral flame image of each wind chamber;

[0034] The flame panoramic image generation module is used to fuse the multispectral flame images of each wind chamber through an image fusion algorithm to obtain a flame panoramic image of each wind chamber; wherein the flame panoramic image of each wind chamber contains multiple spectral information.

[0035] Furthermore, the flame area recognition result includes the flame area image corresponding to each wind chamber; the garbage area recognition result includes the garbage area image corresponding to each wind chamber; the flame area recognition module uses an adaptive threshold segmentation algorithm to perform flame area recognition and garbage area recognition on the flame panoramic image of each wind chamber, and is configured as follows:

[0036] Extract the red component and saturation characteristics of the flame in each wind chamber and the reflectance spectrum characteristics of the garbage in each wind chamber from the flame panoramic image of each wind chamber containing multiple spectral information;

[0037] Based on the color distribution characteristics of flames and the spectral reflectance characteristics of garbage, the red component and saturation characteristics of the flames in each wind chamber, as well as the reflectance spectral characteristics of the garbage in each wind chamber, are processed to establish color models for the flame and garbage areas in each wind chamber.

[0038] It should be understood that light sources in the visible light band emit wavelengths between 400 and 700 nanometers, a range that can be directly perceived by the human eye. The RGB color space represents color based on the human eye's ability to perceive red, green, and blue light. Therefore, the RGB color space is primarily used to represent colors in the visible light band. The HSV (hue, saturation, value) and HSI (hue, saturation, intensity) color spaces are alternative representations of the RGB color space, decomposing color into three independent components: hue, saturation, and value / intensity.

[0039] Based on the color distribution characteristics of the flame and the spectral reflectance characteristics of the garbage, the red component and saturation characteristics of the flame in each wind chamber are analyzed, including: in the RGB color space, the red component of the flame, that is, the pixel value of the red channel is extracted. Flames usually have a higher red component because the color of the flame is mainly red and orange. By setting the threshold of the red component, the flame area can be preliminarily screened out; in the HSV or HSI color space, the saturation feature is extracted, that is, the purity or vividness of the color. The saturation of the flame is usually high, which means that the color of the flame is very bright. The saturation feature can help further distinguish flames from other objects with similar colors, such as red walls or red clothes;

[0040] The color models of the flame area and garbage area are used to distinguish the flame area from the non-flame area and the garbage area from the non-garbage area in each air chamber;

[0041] A gray-level co-occurrence matrix is used to extract texture features of the flame and garbage regions in each air chamber, such as contrast, energy, and uniformity. Texture features help distinguish the dynamic changes of flames from the static distribution of garbage, and identify different types of garbage.

[0042] The Sobel operator is used to determine the shape boundaries of the flame area and the shape boundaries of the garbage area in each air chamber based on the texture characteristics of the flame area and the texture characteristics of the garbage area in each air chamber.

[0043] Among them, shape boundary features include not only contour features, but also the description of the entire shape area. It involves the geometric properties and spatial distribution characteristics of the shape, such as the direction, curvature, and length of the boundary. Methods such as Fourier shape descriptors can be used to describe shape boundaries.

[0044] The maximum inter-class variance algorithm is used to calculate the adaptive threshold corresponding to the flame panoramic image of each wind chamber. The maximum inter-class variance algorithm is used to automatically determine the optimal threshold based on the grayscale histogram of the image to separate the flame area and the garbage area from the background.

[0045] Based on the color models of the flame and garbage areas, the texture features of the flame and garbage areas in each wind chamber, and the shape boundaries of the flame and garbage areas in each wind chamber, the adaptive threshold corresponding to the flame panoramic image of each wind chamber is adjusted. For example, for flame areas with obvious color features, the threshold is appropriately lowered to retain more details; while for garbage areas with complex textures, the threshold is appropriately increased to reduce noise.

[0046] Based on the adjusted adaptive threshold corresponding to the flame panoramic image of each wind chamber, the flame panoramic image of each wind chamber is segmented to obtain a flame area image and a garbage area image corresponding to each wind chamber.

[0047] Furthermore, based on the following formula, the adaptive threshold corresponding to the flame panoramic image of each wind chamber is adjusted according to the color models of the flame area and the garbage area, the texture features of the flame area and the garbage area in each wind chamber, and the shape boundaries of the flame area and the garbage area in each wind chamber:

[0048]

[0049] Among them, T initial is the adaptive threshold calculated using the maximum inter-class variance algorithm; T adjust is the adjusted adaptive threshold; Contrast flame and Contrast total Represent the texture features of the flame area and the texture features of the flame panoramic image respectively; C flame and C gabb They represent the eigenvalues of the color model of the flame area and the eigenvalues of the color model of the garbage area respectively; P flame and P gabb Represent the shape boundary features of the flame area and the shape boundary features of the garbage area respectively; Ptotal Represents the shape boundary features of the flame panoramic image.

[0050] Furthermore, based on the adjusted adaptive threshold corresponding to the flame panoramic image of each wind chamber, the flame panoramic image of each wind chamber is segmented to obtain a flame area image and a garbage area image corresponding to each wind chamber, which is configured as follows:

[0051] Applying the adjusted adaptive threshold corresponding to the flame panoramic image of each wind chamber to the flame panoramic image of each wind chamber to perform image segmentation to obtain a binary image of the flame area and a binary image of the garbage area;

[0052] Morphological processing is performed on the dynamic characteristics of the image to optimize the boundaries of the binary image in the flame area and the boundaries of the binary image in the garbage area. The morphological processing includes erosion and dilation operations to eliminate noise and fill small holes in the area.

[0053] The optimized binary image of the flame area is determined as the flame area image corresponding to each wind chamber, and the optimized binary image of the garbage area is determined as the garbage area image corresponding to each wind chamber.

[0054] Furthermore, the combustion state of the flame in each wind chamber includes the area change and flickering frequency of the flame in each wind chamber; the retention state of the garbage in each wind chamber includes the area fluctuation and accumulation; the wind chamber state determination module determines the combustion state of the flame in each wind chamber and the air flow state between each wind chamber based on the identification result of the flame area; and determines the retention state of the garbage in each wind chamber based on the identification result of the garbage area, and is configured as follows:

[0055] Acquire multiple frames of flame area images corresponding to each wind chamber and multiple frames of garbage area images corresponding to each wind chamber;

[0056] Frame difference analysis is performed on multiple consecutive frames of flame area images corresponding to each wind chamber to identify the difference areas between adjacent frames. The identified difference areas are then used to determine the dynamic change areas of the flame in each wind chamber. The dynamic change areas are used to represent the movement and expansion of the flame in each wind chamber.

[0057] Based on the dynamic change area of the flame in each wind chamber, the motion vector of the flame area is analyzed using the optical flow method to calculate the area change and flicker frequency of the flame in each wind chamber; the area change includes the change speed and change direction;

[0058] If the flame area increases rapidly in a short period of time, it indicates that the combustion intensity is increasing; if the area changes slightly or tends to be stable, it indicates that the combustion is relatively stable. Analyze the flicker frequency of the flame area. The flicker frequency of the flame is closely related to the stability of the combustion. A high flicker frequency may indicate unstable combustion.

[0059] Analyze the area change and flicker frequency of the flame in each plenum to determine whether there is flame connectivity between adjacent plenums. If there is significant movement or overlap between adjacent plenums, this indicates wind crosstalk.

[0060] If it is determined whether there is flame connectivity between adjacent plenums, the area change and flickering frequency of the flame in each plenum are analyzed to identify the airflow disturbance characteristics between the plenums and determine the crossflow status between the plenums. Airflow disturbances may cause irregular changes in flame shape and positional shifts.

[0061] Frame difference analysis is performed on the garbage area images corresponding to each wind chamber to track the shape and distribution characteristics of the garbage area;

[0062] The shape and distribution characteristics of the garbage area are used to determine the area fluctuation and accumulation of garbage in each wind chamber.

[0063] Monitor the fluctuation of the garbage area: if the garbage area remains stable in multiple consecutive frames, it means that the garbage is in a good state; if the area fluctuates greatly, it may mean that the garbage is in an unstable state;

[0064] Garbage accumulation detection: Detects garbage accumulation by identifying the shape and distribution of garbage areas. Garbage accumulation can lead to incomplete combustion and obstructed airflow within the furnace.

[0065] Furthermore, the air supply parameters of each air chamber include air volume, air pressure, wind speed, and air temperature. The air chamber air supply parameter adjustment module adjusts the air supply parameters of each air chamber according to the combustion state of the flame in each air chamber, the air flow state between each air chamber, and the state of garbage accumulation in each air chamber, and is configured as follows:

[0066] Adopting the adaptive feature fusion algorithm, a temporal convolution and bidirectional LSTM hybrid model is constructed; the temporal convolution and bidirectional LSTM hybrid model is the TCN-ABiLSTM hybrid deep neural network model;

[0067] in,

[0068] X=W×softmax[A×(p ref ; F flame ; F leak ; F stay )]

[0069] Where X is the input vector of the temporal convolution and bidirectional LSTM hybrid model. The input vector integrates the reference air supply parameters, the characteristic sequence of the combustion state of the flame in each air chamber, the characteristic sequence of the air flow state between each air chamber, and the characteristic sequence of the garbage retention state in each air chamber. The reference air supply parameters are the air supply parameters of each air chamber before adjustment; W is a learnable weight matrix used to adjust the importance of different features in the model input; A is the feature association matrix used to capture the relationship between different features; softmax is the normalization function; p ref is the reference air supply parameter vector; F flame is the flame combustion state characteristic sequence matrix; F leak is the characteristic sequence matrix of the wind-blowing state, F stay is the garbage residence state characteristic sequence matrix; (p ref ; F flame ; F leak ; F stay ) is the characteristic vector formed by sequentially concatenating the reference air supply parameter vector and various state characteristic sequence matrices;

[0070] The output of the temporal convolution and bidirectional LSTM hybrid model is the time series prediction value of the garbage residence time, the time series prediction value of the turbulence degree, and the time series prediction value of the air volume ratio;

[0071] Among them, the residence time of garbage: ensure that the residence time of garbage in the furnace is greater than the total time required for theoretical drying, thermal decomposition and combustion, so as to ensure that the gaseous combustibles in the flue gas are completely burned;

[0072] Turbulence: An indicator that characterizes the degree of mixing between garbage and air. The greater the turbulence, the better the mixing and the more complete the combustion reaction.

[0073] Air volume ratio: the ratio of actual air volume to theoretical air volume, also known as excess air coefficient, has a great influence on the combustion condition of garbage;

[0074] The temporal convolution and bidirectional LSTM hybrid model includes a temporal convolutional network layer and a bidirectional LSTM layer; the temporal convolutional network layer is used to extract the sequence features of the data; the bidirectional LSTM layer is used to mine the temporal features of the data using the attention mechanism;

[0075] Initialize the gray wolf population, where each individual gray wolf in the gray wolf population represents a set of candidate solutions for air supply parameters;

[0076] The parameters of the improved gray wolf optimization algorithm are set, including the population size and the number of iterations; a chaotic initialization strategy is introduced to make the gray wolf individuals evenly distributed in the search domain; the improved gray wolf optimization algorithm is Improved Grey Wolf Optimizer (IGWO);

[0077] The fitness value of each individual gray wolf is calculated based on a fitness function, wherein the fitness function is determined based on the following parameters: target values for garbage residence time, turbulence, and air volume ratio, and the time series prediction values of garbage residence time, turbulence, and air volume ratio output by a hybrid model of temporal convolution and bidirectional LSTM.

[0078] Sort the gray wolf individuals according to their fitness values, and select the three individuals with the highest fitness as the leading wolves;

[0079] Update the position of each individual gray wolf so that it moves towards the position of the leader wolf to simulate hunting behavior;

[0080] By updating the position of each individual gray wolf, we explore the optimal solution of air supply parameters;

[0081] The optimal solution of the air supply parameters explored is determined as the air supply parameters of each air chamber after adjustment.

[0082] Furthermore, in the process of updating the position of each individual gray wolf, a chaotic perturbation mechanism based on Logistic mapping is introduced;

[0083]

[0084] E ij =μ×E ij ×(1-E ij )

[0085] in, is the updated air supply parameter vector of the i-th gray wolf individual, indicating the updated value of the air supply parameter of the i-th gray wolf individual after iterative update, reflecting the new position of the i-th gray wolf individual in the search domain; q i is the air supply parameter vector of the i-th gray wolf individual, which represents the air supply parameter combination of the i-th gray wolf individual in the current iteration; q leader is the air supply parameter vector of the leader wolf, representing the current optimal solution and serving as the target direction for other individuals to update their positions; B is the coefficient vector, used to control the step size of the individual moving toward the leader wolf's position, which gradually decreases during the iteration process; D is the weight vector, used to adjust the degree of influence of the difference between the individual and the leader wolf; E i is the chaotic disturbance vector, which is generated by the Logistic map and provides an independent disturbance for each air supply parameter; μ is the control parameter of the Logistic map; , which is usually taken between [3.57, 4] to ensure that the system is in a chaotic state, thereby generating a chaotic sequence with ergodicity and randomness; E ijThe jth element of the chaotic perturbation vector represents the perturbation intensity for the jth air supply parameter and is generated iteratively through the logistic map. This formula, by introducing chaotic perturbations, enhances the global search capability of the Grey Wolf optimization algorithm, helping to find more optimal air supply parameter combinations and thus improve the combustion efficiency of the waste incinerator.

[0086] Furthermore, the future state of each wind chamber in the waste incinerator includes the future combustion state of the flame in each wind chamber, the future air flow state between each wind chamber, and the future retention state of the garbage in each wind chamber. The prediction optimization module monitors the multispectral flame images of each wind chamber in the waste incinerator in real time after adjustment, and predicts the future state of each wind chamber in the waste incinerator through a deep learning network to optimize the air volume distribution and combustion coordinated control parameters between each wind chamber. It is configured as follows:

[0087] Real-time monitoring of the multispectral flame images of each wind chamber in the waste incinerator after adjustment;

[0088] Based on the adjusted multispectral flame images of each wind chamber in the waste incinerator, the reference combustion state of the flame in each wind chamber, the reference blowby state between the wind chambers, and the reference residence state of the waste in each wind chamber are determined;

[0089] Through a deep learning network, based on the reference combustion state of the flame in each air chamber, the reference crossflow state between the air chambers, and the reference residence state of the garbage in each air chamber, the future combustion state of the flame in each air chamber, the future crossflow state between the air chambers, and the future residence state of the garbage in each air chamber are predicted to optimize the air volume distribution and combustion coordinated control parameters between the air chambers.

[0090] It should be noted that in this application, the embodiments implemented on the combustion optimization control system side of the waste incinerator based on machine vision can be referenced with the embodiments implemented on the combustion optimization control method side of the waste incinerator based on machine vision, and this application will not go into details one by one.

[0091] The above are only preferred embodiments of the present invention and are not intended to limit the patent scope of the present invention. Any equivalent structure or equivalent process transformation made using the contents of the present invention description and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present invention.

Claims

1. A combustion optimization control method for a waste incinerator based on machine vision, characterized in that: include: Capturing multispectral flame images of each wind chamber in a waste incinerator; Generate a panoramic flame image of each wind chamber based on the multispectral flame image of each wind chamber; The flame area and garbage area of each wind chamber flame panoramic image are identified using an adaptive threshold segmentation algorithm, which is configured as follows: Extract the red component and saturation characteristics of the flame in each wind chamber and the reflectance spectrum characteristics of the garbage in each wind chamber from the flame panoramic image of each wind chamber containing multiple spectral information; Based on the color distribution characteristics of flames and the spectral reflectance characteristics of garbage, the red component and saturation characteristics of the flames in each wind chamber, as well as the reflectance spectral characteristics of the garbage in each wind chamber, are processed to establish color models for the flame and garbage areas in each wind chamber. The color models of the flame area and the garbage area are used to distinguish the flame area from the non-flame area and the garbage area from the non-garbage area in each air chamber; Gray-level co-occurrence matrix is used to extract the texture features of the flame area and the garbage area in each air chamber; The Sobel operator is used to determine the shape boundaries of the flame area and the shape boundaries of the garbage area in each air chamber based on the texture characteristics of the flame area and the texture characteristics of the garbage area in each air chamber. Calculating an adaptive threshold corresponding to the flame panoramic image of each wind chamber using a maximum inter-class variance algorithm; wherein the maximum inter-class variance algorithm is used to automatically determine an optimal threshold based on a grayscale histogram of the image to separate the flame area and the garbage area from the background; Adjusting the adaptive threshold corresponding to the flame panoramic image of each wind chamber based on the color model of the flame region and the garbage region, the texture characteristics of the flame region and the garbage region in each wind chamber, and the shape boundary of the flame region and the garbage region in each wind chamber; Based on the adjusted adaptive threshold corresponding to the flame panoramic image of each wind chamber, the flame panoramic image of each wind chamber is segmented to obtain a flame area image and a garbage area image corresponding to each wind chamber; Based on the identification results of the flame area, the combustion state of the flame in each air chamber and the air flow state between the air chambers are determined; based on the identification results of the garbage area, the retention state of the garbage in each air chamber is determined; Adjust the air supply parameters of each air chamber according to the burning state of the flame in each air chamber, the air leakage state between each air chamber and the retention state of the garbage in each air chamber; The multispectral flame images of each wind chamber in the waste incinerator are monitored in real time after adjustment, and the future state of each wind chamber in the waste incinerator is predicted through a deep learning network to optimize the air volume distribution and combustion coordinated control parameters between the wind chambers.

2. A combustion optimization control system for a waste incinerator based on machine vision, wherein the system implements the method according to claim 1, characterized in that: include: Flame image capture module, used to capture multispectral flame images of each wind chamber in the waste incinerator; A flame panoramic image generation module is used to generate a flame panoramic image of each wind chamber based on the multispectral flame image of each wind chamber; The flame area recognition module is used to use an adaptive threshold segmentation algorithm to perform flame area recognition and garbage area recognition on the flame panoramic image of each wind chamber; The air chamber state determination module is used to determine the combustion state of the flame in each air chamber and the air flow state between each air chamber based on the identification result of the flame area; and to determine the retention state of the garbage in each air chamber based on the identification result of the garbage area; The air supply parameter adjustment module of the air chamber is used to adjust the air supply parameters of each air chamber according to the combustion state of the flame in each air chamber, the air leakage state between each air chamber and the retention state of the garbage in each air chamber; The prediction and optimization module is used to monitor the multispectral flame images of each wind chamber in the waste incinerator in real time after adjustment, and predict the future state of each wind chamber in the waste incinerator through a deep learning network to optimize the air volume distribution and combustion coordinated control parameters between the wind chambers.

3. The combustion optimization control system of a waste incinerator based on machine vision according to claim 2 is characterized in that: The flame image capturing module includes at least two light sources of different wavelengths; wherein the at least two light sources of different wavelengths include a light source of visible light band and a light source of near infrared band; The flame image capture module is used to capture images of each wind chamber in the waste incinerator under different spectra using at least two light sources with different wavelengths, thereby obtaining a multispectral flame image of each wind chamber; The flame panoramic image generation module is used to fuse the multispectral flame images of each wind chamber through an image fusion algorithm to obtain a flame panoramic image of each wind chamber; wherein the flame panoramic image of each wind chamber contains multiple spectral information.

4. The combustion optimization control system of a waste incinerator based on machine vision according to claim 3 is characterized in that: The flame area recognition result includes the flame area image corresponding to each wind chamber; the garbage area recognition result includes the garbage area image corresponding to each wind chamber; the flame area recognition module uses an adaptive threshold segmentation algorithm to perform flame area recognition and garbage area recognition on the flame panoramic image of each wind chamber.

5. The combustion optimization control system of a waste incinerator based on machine vision according to claim 4 is characterized in that: According to the color models of the flame area and the garbage area, the texture features of the flame area and the garbage area in each wind chamber, and the shape boundaries of the flame area and the garbage area in each wind chamber, the adaptive threshold corresponding to the flame panoramic image of each wind chamber is adjusted.

6. The combustion optimization control system of a waste incinerator based on machine vision according to claim 4 is characterized in that: The flame panoramic image of each wind chamber is segmented based on the adjusted adaptive threshold corresponding to the flame panoramic image of each wind chamber to obtain a flame area image and a garbage area image corresponding to each wind chamber, and is configured as follows: Applying the adjusted adaptive threshold corresponding to the flame panoramic image of each wind chamber to the flame panoramic image of each wind chamber to perform image segmentation to obtain a binary image of the flame area and a binary image of the garbage area; Performing morphological processing on the dynamic characteristics of the image to optimize the boundaries of the binary image of the flame area and the boundaries of the binary image of the garbage area; wherein the morphological processing includes an erosion operation and a dilation operation; The optimized binary image of the flame area is determined as the flame area image corresponding to each wind chamber, and the optimized binary image of the garbage area is determined as the garbage area image corresponding to each wind chamber.

7. The combustion optimization control system of a waste incinerator based on machine vision according to claim 6 is characterized in that: The combustion state of the flame in each wind chamber includes the area change and flickering frequency of the flame in each wind chamber; the retention state of the garbage in each wind chamber includes the area fluctuation and accumulation; the wind chamber state determination module determines the combustion state of the flame in each wind chamber and the air flow state between the wind chambers based on the identification result of the flame area; determines the retention state of the garbage in each wind chamber based on the identification result of the garbage area, and is configured as follows: Acquire multiple frames of flame area images corresponding to each wind chamber and multiple frames of garbage area images corresponding to each wind chamber; Frame difference analysis is performed on multiple consecutive frames of flame area images corresponding to each wind chamber to identify the difference areas between adjacent frames; the dynamically changing areas of the flame in each wind chamber are determined based on the identified difference areas; wherein the dynamically changing areas are used to represent the movement and expansion of the flame in each wind chamber; Based on the dynamic change area of the flame in each wind chamber, the motion vector of the flame area is analyzed using the optical flow method to calculate the area change and flicker frequency of the flame in each wind chamber; wherein the area change includes the change speed and change direction; Analyze the area change and flicker frequency of the flame in each wind chamber to determine whether there is flame connectivity between adjacent wind chambers; If it is determined whether there is flame communication between adjacent air chambers, the area change and flickering frequency of the flame in each air chamber are analyzed to identify the air flow disturbance characteristics between the air chambers to determine the air flow status between the air chambers; Frame difference analysis is performed on the garbage area images corresponding to each wind chamber to track the shape and distribution characteristics of the garbage area; The shape and distribution characteristics of the garbage area are used to determine the area fluctuation and accumulation of garbage in each wind chamber.

8. The combustion optimization control system of a waste incinerator based on machine vision according to claim 7 is characterized in that: The air supply parameters of each air chamber include air volume, air pressure, wind speed, and air temperature. The air supply parameter adjustment module adjusts the air supply parameters of each air chamber according to the combustion state of the flame in each air chamber, the air flow state between each air chamber, and the state of garbage retention in each air chamber. It is configured as follows: Adopting adaptive feature fusion algorithm, constructing a hybrid model of temporal convolution and bidirectional LSTM; The output of the temporal convolution and bidirectional LSTM hybrid model is the time series prediction value of the garbage residence time, the time series prediction value of the turbulence, and the time series prediction value of the air volume ratio; The temporal convolution and bidirectional LSTM hybrid model includes a temporal convolutional network layer and a bidirectional LSTM layer; the temporal convolutional network layer is used to extract the sequence features of the data; the bidirectional LSTM layer is used to mine the temporal features of the data using the attention mechanism; Initializing a gray wolf population, wherein each gray wolf individual in the gray wolf population represents a candidate solution for a set of air supply parameters; Set the parameters of the improved gray wolf optimization algorithm, including the population size and the number of iterations; introduce a chaotic initialization strategy to make the gray wolf individuals evenly distributed in the search domain; Calculating the fitness value of each individual gray wolf according to a fitness function; wherein the fitness function is determined based on the following parameters: a target value for the residence time of garbage, a target value for the turbulence, and a target value for the air volume ratio; and a time series prediction value of the residence time of garbage, a time series prediction value of the turbulence, and a time series prediction value of the air volume ratio output by a hybrid model of temporal convolution and bidirectional LSTM; Sort the gray wolf individuals according to their fitness values, and select the three individuals with the highest fitness as the leading wolves; Updating the position of each individual gray wolf so that it moves toward the position of the leader wolf to simulate hunting behavior; By updating the position of each individual gray wolf, we explore the optimal solution of air supply parameters; The optimal solution of the air supply parameters explored is determined as the air supply parameters of each air chamber after adjustment.

9. The combustion optimization control system of a waste incinerator based on machine vision according to claim 8 is characterized in that: In the process of updating the position of each gray wolf, a chaotic perturbation mechanism based on Logistic mapping is introduced.

10. The combustion optimization control system of a waste incinerator based on machine vision according to claim 2, characterized in that: The future state of each wind chamber in the waste incinerator includes the future combustion state of the flame in each wind chamber, the future blowby state between each wind chamber, and the future retention state of the garbage in each wind chamber; the prediction optimization module monitors the adjusted multispectral flame image of each wind chamber in the waste incinerator in real time, and predicts the future state of each wind chamber in the waste incinerator through a deep learning network to optimize the air volume distribution and combustion coordinated control parameters between each wind chamber, and is configured as follows: Real-time monitoring of the adjusted multispectral flame images of each wind chamber in the waste incinerator; Determining a reference combustion state of the flame in each wind chamber, a reference blowby state between each wind chamber, and a reference residence state of garbage in each wind chamber based on the adjusted multispectral flame image of each wind chamber in the waste incinerator; Through the deep learning network, based on the reference combustion state of the flame in each wind chamber, the reference crossflow state between the wind chambers, and the reference residence state of the garbage in each wind chamber, the future combustion state of the flame in each wind chamber, the future crossflow state between the wind chambers, and the future residence state of the garbage in each wind chamber are predicted to optimize the air volume distribution and combustion coordinated control parameters between the wind chambers.

Images