Layered temperature control pyrolysis gasification intelligent monitoring system

Abstract

The application discloses a kind of layered temperature control pyrolysis gasification intelligent monitoring systems, belong to domestic waste pyrolysis processing technical field.The system includes the pyrolysis gasification furnace body of vertical structure, monitoring module, execution module and control module.The following core steps are executed by intelligent algorithm: based on flue gas image extraction state index of flue gas representing combustion sufficiency;Based on the image in the furnace extraction layered definition index representing pyrolysis layered stability;Combining other data, using multivariate fusion control model, real-time collaborative calculation and transmission to execution module adjust air supply and auxiliary heating amount.The application solves the problem that prior art monitoring dimension is single, control relies on artificial, weak emission guarantee, significantly improves the operation efficiency, stability and environmental protection standard reliability of distributed garbage pyrolysis equipment.

Description

Technical Field

[0001] This invention belongs to the field of municipal solid waste pyrolysis treatment technology, specifically relating to a layered temperature-controlled intelligent monitoring system for pyrolysis gasification. Background Technology

[0002] Distributed waste management scenarios refer to application models that involve on-site or nearby processing at the source of waste generation or in nearby areas, such as towns, rural areas, islands, and remote communities. These scenarios typically face core constraints such as small and dispersed waste generation, and high or even impractical long-distance transportation costs. Therefore, technologies suitable for this scenario must meet specific requirements such as miniaturization, modularity, low operating energy consumption, strong environmental adaptability, ease of operation, and high cost-effectiveness.

[0003] Municipal solid waste pyrolysis gasification technology, especially the non-powered stratified temperature control technology suitable for distributed scenarios, has become an effective way to solve the waste disposal problems in towns and villages due to its advantages such as low investment and operating costs and no need for auxiliary fuel. Its core equipment—the vertical stratified pyrolysis gasifier—relies on the weight of the waste falling and forming functional layers for drying, pyrolysis, and combustion within the furnace, achieving a self-sustaining and clean conversion of waste. However, in terms of the actual operation and control of the equipment, existing technologies still have the following significant shortcomings: 1) Monitoring dimensions are singular and isolated, failing to achieve process collaborative optimization: Existing monitoring solutions mostly focus on simple monitoring of local temperatures in the furnace or final flue gas emissions, such as installing thermocouples only at the outlet of the secondary combustion chamber or installing simple flue gas analyzers on the chimney. Due to the lack of synchronous acquisition and correlation analysis of multiple parameters such as pressure gradients in the furnace, temperature distribution in each layer, and oxygen content at key locations, operators find it difficult to accurately judge the pyrolysis progress, combustion completeness, and gasification efficiency of materials. This leads to lagging and blind operation adjustments, with equipment operating in suboptimal conditions for extended periods, resulting in large fluctuations in processing efficiency and persistently high energy consumption and operating costs.

[0004] 2) Low level of automation and intelligence in control, over-reliance on manual experience: Currently, key operating parameters of most distributed pyrolysis equipment, such as air supply and feeding speed, still rely mainly on on-site operators to manually adjust based on limited instrument readings and personal experience. The equipment lacks automatic control algorithms based on multivariable coupling models, and when faced with disturbances such as fluctuations in waste composition and weather changes, it cannot automatically maintain the stability and efficiency of the pyrolysis process, easily leading to problems such as incomplete combustion, increased pollutant generation, or temperature runaway.

[0005] 3) Weak capacity to ensure compliance with environmental emission standards and lack of risk early warning: Existing technologies lack real-time, online process monitoring capabilities for pollutants generated during pyrolysis. Typically, only periodic sampling and testing are conducted at the end of the process, making online intervention impossible at key stages of pollutant generation (such as the pyrolysis and combustion layers). When emission indicators fluctuate or approach exceed limits, the system lacks effective real-time early warning mechanisms and automatic feedback control strategies, resulting in a risk of delayed compliance with environmental emission standards and difficulty in meeting increasingly stringent environmental regulatory requirements.

[0006] For example, in patent 202110642601.2, entitled "An Intelligent Flue Gas Control System for a Municipal Solid Waste Incineration Equipment," the temperature detection instrument is located in the air chamber area corresponding to the oxygen supply pipeline system; the flue gas oxygen content detection instrument, negative pressure instrument, and CO detector are respectively installed at the flue gas outlet of the incinerator and the flue gas outlet of the secondary combustion chamber; the negative pressure fan is located at the upper end of the secondary combustion chamber; the flow indicator and flow control valve are installed on each oxygen supply pipeline system; and the in-furnace material weight feedback instrument is located below the incinerator. This point-based monitoring cannot comprehensively and in real time reflect the complete thermal state of each functional layer in the furnace. Summary of the Invention

[0007] In view of this, in order to solve the problems mentioned in the background art, a layered temperature-controlled pyrolysis gasification intelligent monitoring system is proposed. The objective of this invention can be achieved through the following technical solutions.

[0008] A layered temperature-controlled intelligent monitoring system for pyrolysis gasification includes: The pyrolysis furnace has an exhaust pipe at the center of its top and a feed inlet on the side wall of the exhaust pipe. Its main body is a cylindrical structure, with the top and bottom connected to the cylinder by upper and lower conical structures, respectively. An observation port for observing the state inside the furnace is provided in the lower half of the cylinder. The monitoring module includes multiple temperature and pressure sensors installed at different heights within the furnace body, a flue gas analyzer for monitoring flue gas composition, and at least one high-temperature camera device positioned aligned with the observation port to capture images of the pyrolysis and gasification process of materials inside the furnace. The execution module includes an air supply system for supplying air into the furnace and an auxiliary burner for auxiliary heating; The control module is communicatively connected to both the monitoring module and the execution module, and is configured as follows: S1: Acquire real-time data from the monitoring module, including combustion zone temperature T, CO concentration C, flue gas image I_s, and pyrolysis furnace image I_f; S2: Extract and classify features from the flue gas image I_s, and calculate the flue gas state index I_smoke, which represents the degree of incomplete combustion; S3: Perform image analysis on the image I_f inside the pyrolysis furnace, identify the layering interface of the waste layer, and calculate the layer clarity index I_layer, which characterizes the quality of layering; S4: Based on the CO concentration C, the flue gas state index I_smoke, and the preset target CO concentration C_target, calculate the target air supply volume Q_a of the air supply system through the first control model. S5: Based on the combustion zone temperature T, CO concentration C, stratification clarity index I_layer, and preset target temperature T_target, the target heating amount Q_b of the auxiliary combustion system is calculated through the second control model; S6: Output the calculated Q_a and Q_b to the air supply system and auxiliary combustion system for execution, so as to maintain the temperature of the oxygen-overflow combustion zone at no lower than the preset high temperature threshold and maintain the stable distribution of the pyrolysis and gasification reaction of the material in the furnace in each zone.

[0009] Preferably, in step S2, the flue gas state index I_smoke, which characterizes the degree of incomplete combustion, is calculated as follows: S21: Perform color space conversion and region segmentation on the smoke image I_s, and extract the overall color histogram features of the smoke region; S22: Input the color histogram features into a pre-trained classification model. The classification model outputs the probability distribution of flue gas belonging to a preset combustion state category. The combustion state category includes at least complete combustion, incomplete combustion, and excessive oxygen or high humidity. S23: Based on the probability distribution and the preset weight vector, the scalar form of the flue gas state index I_smoke is calculated, with the incomplete combustion category having the highest weight.

[0010] Preferably, in step S3, the layer sharpness index I_layer, which characterizes the quality of layering, is calculated as follows: S31: Perform grayscale conversion and vertical gradient calculation on the furnace image I_f to obtain the gradient intensity image; S32: In the gradient intensity image, find the gradient extremum points along the vertical direction, the extremum points corresponding to the potential interfaces between different garbage layers; S33: Statistically analyze the number of extreme points, average gradient intensity, and uniformity of distribution within the preset height range; S34: Based on the number of extreme points, average gradient strength, and distribution uniformity, the layered sharpness index I_layer is calculated using a weighted fusion algorithm.

[0011] Preferably, the first control model in step S4 is:

[0012] Where K_p1, K_i1, and K_d1 are PID control parameters for CO concentration deviation, and K_s is the compensation coefficient for the flue gas state index.

[0013] Preferably, the second control model in step S5 is:

[0014] Where K_p2 and Ki2 are PI control parameters for temperature deviation. This is the compensation coefficient for CO concentration. This is the compensation coefficient for the layered sharpness index. When the layered quality is poor and the I_layer value is low, the compensation amount is increased. Increase.

[0015] Preferably, the control module is further configured to execute step S7: S7: Based on the current Q_a, Q_b, T, C, I_smoke and I_layer data, calculate a comprehensive furnace health score H using the third evaluation model; When H is below the first health threshold, an optimization suggestion signal is generated, prompting adjustment of at least one control parameter in the first control model and / or the second control model. When H falls below the second health threshold, which is below the first health threshold, a fault alarm signal is generated and a safety protocol is executed.

[0016] Preferably, the control module includes an edge computing module and a cloud collaboration module; The edge computing module is deployed near the pyrolysis gasifier body and is used to execute steps S1 to S6 to achieve real-time closed-loop control at the millisecond to second level. The cloud collaboration module communicates with the edge computing module to receive historical operating data, perform offline optimization of the control parameters in the first control model and the second control model, and send the optimized parameters to the edge computing module for updating.

[0017] Compared with the prior art, the beneficial effects of the present invention are as follows: 1) Achieved comprehensive and precise monitoring and diagnosis of the pyrolysis process: By constructing a multi-dimensional sensing network integrating furnace temperature, carbon monoxide concentration, flue gas visual characteristics, and internal pyrolysis stratification images, the system can perform three-dimensional and real-time status monitoring and correlation analysis of the entire pyrolysis and combustion process. This enables operators to comprehensively and accurately grasp the material pyrolysis progress, combustion sufficiency, and operational stability, solving the problems of isolated information and delayed judgment in traditional monitoring methods, and providing a reliable data foundation for process optimization.

[0018] 2) Improved system automation and intelligence, ensuring stable and efficient operation: Based on a multivariable coupled control model, the system can automatically integrate multi-source monitoring data and adjust the air supply and auxiliary heating in real time. This intelligent closed-loop control mechanism can adaptively respond to external disturbances, such as fluctuations in waste composition and changes in operating conditions, thereby continuously maintaining the pyrolysis process in a highly efficient and stable state. This significantly reduces reliance on manual experience, improves the stability of processing efficiency, and effectively reduces energy consumption and operating costs.

[0019] 3) Enhanced pollutant process control and emission risk early warning capabilities: By shifting environmental monitoring from end-of-pipe detection to process stages, the system enables real-time assessment and dynamic control of key influencing factors such as combustion completeness. Combined with an intelligent assessment and early warning mechanism based on multi-parameter trend analysis, the system can identify process risks before emission indicators become abnormal and trigger corresponding controls or alarms, achieving a shift from passive treatment to proactive prevention and control, effectively ensuring the continuous and stable compliance of environmental emissions. Attached Figure Description

[0020] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0021] Figure 1 This is a schematic diagram of the pyrolysis gasification furnace body structure of the present invention; Figure 2 This is the core control flowchart of the system of the present invention; Figure 3 This is a detailed flowchart of step S2, flue gas state index extraction, of the present invention. Figure 4 This is a detailed flowchart of step S3, layered clarity index extraction, of the present invention.

[0022] Explanation of the labels in the diagram: 100 - Pyrolysis gasification furnace body, 110 - Main body, 120 - Top funnel-shaped structure, 130 - Bottom funnel-shaped structure, 140 - Exhaust pipe, 150 - Feed inlet, 160 - Observation port. Detailed Implementation

[0023] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0024] This invention provides a layered temperature-controlled intelligent monitoring system for pyrolysis gasification, such as... Figure 1 As shown, The pyrolysis gasification furnace body 100 is a vertical structure, comprising a cylindrical main body 110, a funnel-shaped structure 120 welded to the top of the main body, and a funnel-shaped structure 130 welded to the bottom of the main body. The main body 110 has an inner diameter of 4.85 meters, a total height of 6.8 meters, an effective volume of approximately 28 cubic meters, and a designed daily processing capacity of 15 tons of municipal solid waste. A flue pipe 140 is fixedly connected to the center of the top of the furnace body. The flue pipe 140 has an inner diameter of 0.6 meters, is made of 310S stainless steel, and has a wall thickness of 12 mm. A feed inlet 150 is opened on the side of the flue pipe 140. The feed inlet has a size of 2.7 meters × 2.4 meters and is equipped with a hydraulically driven sealing cover. The cover is surrounded by high-temperature resistant ceramic fiber sealing strips. The sealing pressure is adjustable from 0 to 0.5 MPa to ensure that it is open when feeding and sealed after feeding. An observation port 160 is provided in the lower half of the cylindrical main body 110, at a vertical distance of 1.8 meters from the furnace bottom. The observation port 160 uses a double-layer high-temperature resistant borosilicate sight glass with a diameter of 120 mm, a thickness of 15 mm, and a light transmission band of 400-1100 nanometers. The observation port 160 is equipped with an independent compressed air curtain cleaning system with an air source pressure of 0.4-0.6 MPa, which can periodically or manually blow away the dust adhering to the surface of the sight glass. The central axis of the observation port 160 forms a downward angle of 15° with the horizontal plane, and its field of view covers the furnace space from the horizontal line of the observation port upward to the bottom of the feed port, with a vertical coverage angle of about 60° and a horizontal coverage angle of about 45°. This location was chosen based on a large amount of experimental data, because the 0-1.0 meter area at the bottom of the furnace is mainly a slag accumulation layer, and the solid material is dense and has poor fluidity. If the observation port is set here, the window is easily buried by slag. At a height of 1.8 meters, the slag layer has become more stable, and above it is the active area of ​​the oxygen-overflow combustion layer and the oxygen-deficient pyrolysis layer. The material is loose and porous, and the window can clearly observe the temperature radiation zone from bottom to top and the material falling interface from top to bottom.

[0025] The monitoring module includes the following devices: The preferred temperature sensor is a type K armored thermocouple with a probe diameter of 6 mm and an inconel 600 high-temperature alloy sheath. The temperature range is 0-1200℃. Three thermocouples are installed: one at a height of 1.8 meters from the furnace bottom, corresponding to the central area of ​​the oxygen-rich combustion layer, embedded 150 mm into the furnace wall; another at a height of 2.8 meters from the furnace bottom, corresponding to the lower area of ​​the oxygen-deficient pyrolysis layer, embedded 100 mm into the furnace wall; and the third at a height of 4.2 meters from the furnace bottom, corresponding to the boundary between the drying layer and the pyrolysis layer, embedded 80 mm into the furnace wall. The signal lines of the three thermocouples use high-temperature resistant compensated wires, which are collected in a terminal box and then connected to the analog input module of the control unit. The CO sensor preferably uses a non-dispersive infrared (NDIR) gas analysis module with a measurement range of 0-5000ppm, a resolution of 1ppm, and a response time of ≤15 seconds. The sampling probe is installed on the vertical section of the 140mm exhaust pipe, 0.5 meters above the connection between the furnace body and the exhaust pipe. The entire sampling pipeline is heated and insulated, with the heating temperature maintained at 160-180℃ to prevent water vapor condensation from dissolving acidic gases and causing measurement errors. The sampling pretreatment system includes a ceramic filter, a peristaltic pump drainage unit, and a gas-water separator. The flue gas image acquisition device preferably uses an industrial area array CCD color camera, equipped with an electric zoom lens, a stainless steel water-cooled jacket and an air purging device. The installation position is located on the side observation window of the exhaust pipe 140, 1.0 meter away from the flue gas outlet flange. The lens axis is at a 90° angle with the flue gas flow direction, and the flue gas image is captured through a 45° reflector. The high-temperature camera device is preferably an endoscope-type camera specifically designed for high-temperature environments; it is equipped with an independent circulating water cooling system and compressed air positive pressure explosion-proof purging; the high-temperature camera device 230 is fixedly installed on the flange of the observation port 160, and the lens can be adjusted to penetrate 0-20 mm into the inner wall of the furnace. After installation, it is locked with a sealing cover; its video signal is converted into an optical fiber signal through an optical transceiver and transmitted to the optical fiber transceiver in the control room, and then connected to the control module; The pressure sensor is preferably a diffused silicon pressure transmitter, which is installed on the side pressure tap of the exhaust pipe 140 at the top of the furnace body to monitor the negative pressure status of the furnace.

[0026] The control module adopts a layered distributed architecture, including an industrial computer, a field control layer, and a remote monitoring layer. The field control layer is located in the operator's room near the furnace body, and the core device is a programmable logic controller (PLC). The rack extensions are distributed near various measuring points on the furnace body and communicate with the CPU via a PROFINET bus. The remote monitoring layer is deployed on a cloud server. A 5G wireless network or wired broadband maintains a long-term connection with the cloud server, and encrypted data transmission is performed using the MQTT protocol.

[0027] The execution module configuration includes an air supply system and an auxiliary burner. The air supply system comprises a primary air fan, a secondary air fan, and an induced draft fan. The primary air fan is a Type 472 centrifugal fan with a power of 7.5 kW, an air volume of 4500-6500 cubic meters per hour, and a total pressure of 1200-1600 Pa. The secondary air fan is a Type 919 high-pressure centrifugal fan with a power of 5.5 kW, an air volume of 2000-3000 cubic meters per hour, and a total pressure of 2500-3000 Pa. Both fans are equipped with Danfoss FC302 series frequency converters, with a control signal of 420mA and a response time of ≤200 milliseconds. The auxiliary burner uses a light diesel fuel burner.

[0028] The monitoring module is used to collect key physicochemical parameters and visual information within the reactor in real time, specifically including: Temperature sensor: preferably a type K thermocouple, at least one of which is installed in the area corresponding to the oxygen-permeable combustion layer inside the furnace to monitor the core combustion temperature T; Smoke image acquisition device: preferably an industrial color CCD with a protective cover, installed at the outlet of the exhaust pipe 140, with its lens aimed at the direction of smoke flow, for continuously capturing smoke images I_s; High-temperature imaging device: This is an endoscope specifically designed for high-temperature environments, equipped with a high-temperature resistant camera and a forced air-cooled jacket, installed at the observation port 160. Its lens angle is adjusted so that its main field of view covers the area inside the furnace from the horizontal line of the observation port to below the feed inlet, specifically for acquiring internal images I_f reflecting the stratification state of waste pyrolysis; CO sensor: preferably a non-dispersive infrared (NDIR) sensor, whose sampling probe extends into the upper flue gas collection area of ​​the furnace body for real-time detection of carbon monoxide concentration C; Optional pressure sensor: installed on the side wall of the furnace body to monitor the negative pressure status inside the furnace.

[0029] The control module is communicatively connected to all sensors and actuators in the monitoring module via data cables. The control module internally stores a control program configured to execute a continuously running control loop; for example... Figure 2 The diagram shown is the core control flowchart of the system. The algorithm program of this invention is encapsulated in the "pyrolysis cloud control platform" software module of the industrial control computer. It is implemented by mixed programming of C# and Python. The image processing part calls the OpenCV library version 4.5, and the deep learning inference calls TensorFlow.NET.

[0030] In step S1, the control module synchronously triggers and reads all real-time data from the monitoring module at a fixed period, including combustion zone temperature T, CO concentration C, flue gas image I_s, and furnace image I_f; the system adds a unified timestamp to each data packet to ensure the temporal consistency of subsequent analysis. The system operates on a 1-second cycle, with the PLC's timer interrupt triggering the industrial control computer to execute a data acquisition command. Simultaneously, the industrial control computer sends a software trigger signal to the flue gas image acquisition device via the GigE interface to capture the current frame of flue gas image I_s. The image resolution is 1920×1080, with RGB three channels and 8-bit depth. The system automatically adds a timestamp to the image and stores it in association with temperature and CO data from the same period. The high-temperature camera device 230 continuously acquires images I_f of the furnace interior at a 1-second interval, with an original resolution of 1280×720 and a frame rate of 25fps. Due to the high-temperature radiation environment inside the furnace, the images exhibit significant uneven lighting—overexposed central areas and underexposed edge areas. The system selects a central 800×600 pixel area for processing to avoid distorted edge areas.

[0031] like Figure 3 As shown, step S2: extraction of flue gas state index I_smoke; First, the smoke image I_s is preprocessed: ROI cropping: Since about 60% of the area in the flue gas image is the background flue wall, in order to improve processing efficiency, the system fixedly cropped an 800×600 pixel rectangular area in the center of the image as the region of interest (ROI), which covers the main flue gas flow channel; Color space conversion: Use OpenCV's `cvtColor()` function to convert the ROI from RGB space to HSV space; H channel range 0180, S channel 0255, V channel 0255; Background modeling and smoke segmentation: During initialization, the system acquires 100 consecutive frames of images with no or very little smoke as background modeling samples; a Gaussian mixture model (GMM) is used, with the number of Gaussian components set to 3 and the background threshold set to 2.5 times the standard deviation; during runtime, each frame of image is subtracted from the background model, and pixels with a foreground probability greater than 0.7 are marked as smoke regions; the segmented binary mask is multiplied with the original image to extract a clean smoke foreground image; Then, extract color statistical features from the foreground region: Color characteristics: Histograms for the H and S channels are calculated separately within the foreground smoke region. The number of bins in the histogram is set to 32. Extracted features include the H channel mean, H channel variance, S channel mean, and S channel variance. In actual operation, the S channel mean of black smoke samples is usually below 80, and the H channel distribution is divergent; the S channel mean of gray smoke samples is between 80 and 150; the S channel mean of white smoke samples is above 150, and the V channel is close to 255.

[0032] Texture features: The foreground image of the flue gas is converted into a grayscale image, and the gray-level co-occurrence matrix (GLCM) is calculated. The parameters are set as follows: distance d=1, angle θ=0°, 45°, 90°, 135°, and grayscale levels are compressed to 16 levels. Four feature values ​​are extracted: contrast, energy, homogeneity, and entropy. Experience shows that when combustion is complete, the flue gas texture is delicate, with low contrast and low entropy; when combustion is incomplete, the flue gas turbulence is intense, with high contrast and high entropy. Optical thickness characteristic: The ratio of the average V channel value of the 10% of pixels with the lowest brightness in the foreground smoke area to the average V channel value of the background area (smoke-free area) is used as the optical thickness index. The closer this index is to 0, the denser the smoke and the worse the light transmittance. This embodiment uses the lightweight convolutional neural network MobileNetV2 as the classification model. The network input layer size is 224×224×3, and the output layer has 3 neurons, corresponding to the three categories of complete combustion, slightly incomplete combustion, and severely incomplete combustion. The model has been trained under laboratory conditions using 5000 manually labeled smoke images, achieving an accuracy of 94.2% on the training set and 91.7% on the validation set.

[0033] During runtime, the system scales the smoke foreground image to 224×224, normalizes it to the [0,1] interval, and inputs it into the TensorFlow model. Forward inference takes approximately 35 milliseconds. The model output is a three-class probability vector P = [p_1, p_2, p_3].

[0034] The system presets a baseline value W = [w_1, w_2, w_3] for each category, with 0, 0.6, and 1.0 being preferred. The flue gas state index is calculated by weighted summation. , The index is a continuous scalar between 0 and 1, with higher values ​​indicating more severe incomplete combustion. For example, if the model outputs P=[0.15, 0.70, 0.15] at a certain moment, then I_smoke=0.15×0+0.70×0.6+0.15×1.0=0.57. The index is updated in real time with a period of 0.1 seconds, and after being filtered by a moving average (window length 5), it is stored in the real-time database for subsequent decision-making modules to use.

[0035] like Figure 4 As shown, step S3: Extraction of the layered sharpness index I_layer; First, image enhancement is performed; Illumination correction: The homomorphic filtering algorithm is used to convert the image to the logarithmic domain. A high-pass filter is used to suppress low-frequency illumination components and enhance high-frequency reflection components. The filter cutoff frequency is set to 30 pixels and the gain coefficient is 0.8. After processing, the brightness difference between the center and edge of the image is reduced from 120 gray levels in the original image to less than 30 gray levels. Noise reduction: Bilateral filtering is used with a diameter d=9, color space sigmaColor=50, and coordinate space sigmaSpace=50. Bilateral filtering can smooth noise while preserving edge information well. Compared with Gaussian filtering, the edge preservation is improved by about 40%. Then gradient calculation and projection are performed; Grayscale conversion: The RGB image is converted to grayscale using a weighted average method, with weight coefficients R: 0.299, G: 0.587, and B: 0.114; Vertical gradient calculation: The vertical gradient G_y is calculated using the Sobel operator; the convolution kernel size is 3×3, and the absolute value of the calculation result is taken and normalized to the range of 0-255; Horizontal projection: The vertical gradient image is integrally projected horizontally, that is, the gradient values ​​of each row of pixels are summed; the horizontal axis of the projection curve Profile(y) is the image row coordinate (corresponding to the vertical height of the furnace), and the vertical axis is the cumulative gradient intensity.

[0036] Next, peak detection is performed; the system adopts a peak detection algorithm based on continuous wavelet transform. The Mexican Hat wavelet is selected as the mother wavelet, with a scale factor a=5. The Profile(y) curve is subjected to wavelet transform, and the position corresponding to the local maximum value of the wavelet coefficient is the candidate peak; further threshold filtering is set: the peak amplitude must be greater than 1.5 times the mean of the entire curve, and the distance between adjacent peaks must not be less than 20 pixels; in actual operation, when the stratification in the furnace is clear, the system usually detects 34 significant peaks, corresponding to: the interface between the original waste layer and the drying layer, the interface between the drying layer and the pyrolysis layer, the interface between the pyrolysis layer and the oxygen-permeable combustion layer, and the interface between the oxygen-permeable combustion layer and the slag layer.

[0037] Re-feature quantization: The system extracts the following three quantization features: Number of effective peaks N: The total number of peaks detected within a preset vertical range (corresponding to a furnace height of 1.0-4.5 meters); when N=3 or 4, it indicates good layering; Average peak intensity A_avg: The average amplitude of all detected peaks; This value reflects the interlayer contrast and ranges from 0 to 255; Under normal operating conditions, A_avg is usually between 80 and 150; Peak distribution uniformity U: Calculates the standard deviation of the distance between adjacent peaks; The smaller the standard deviation, the more uniform the thickness distribution of each layer; Operating experience shows that layering uniformity is better when the standard deviation is less than 30 pixels.

[0038] Finally, a single exponent is synthesized through a fusion function: , Where f(N) is an evaluation function for the number of peaks, taking its maximum value at the ideal number of layers; A_max and U_max are normalization coefficients; α, β, and γ are weighting coefficients; a higher I_layer value indicates clearer and more stable pyrolysis layering; 1.0 is taken when N=3, 0.9 when N=4, 0.5 when N=2, and 0.2 when N=1 or N≥5; the denominator 150 is the empirical value of the expected maximum value of \(A_{avg}\), and the denominator 100 is the empirical value of the expected maximum value of U. This index is also stored in the real-time database after being filtered by a moving average (window length 5).

[0039] Step S4: Air supply volume decision; Feedback control channel: based on CO concentration deviation As input. The value is set at 100 ppm, which balances combustion completeness and airflow economy, and employs an incremental digital PID algorithm.

[0040] In this embodiment, after engineering tuning, the PID parameters are: K_p=0.8, Ki=0.05, K_d=0.2, and the sampling period T_s=2 seconds. The control quantity u_PID has the dimension of cubic meters per hour, and the output limit is 0-2000 cubic meters per hour. Feedforward compensation channel: The flue gas state index I_smoke(k) is used as the feedforward signal. The feedforward gain K_s was determined to be 150 through on-site step response testing; that is:

[0041] When black smoke appears in the flue gas, i.e., I_smoke is close to 1.0, the feedforward compensation can reach up to 150 cubic meters per hour, and the system can increase the air volume in advance without waiting for the CO concentration to rise. The baseline air volume Q_a_base is set to 500 cubic meters per hour to maintain a basic negative pressure in the furnace (50 Pa to 100 Pa); the final target air supply volume Q_a(k) is: , The system converts the calculated Q_a) value into a 420mA signal through the analog output module and sends it to the frequency converters of the primary and secondary air fans respectively; the air volume distribution strategy is: primary air accounts for 60% and secondary air accounts for 40%; Step S5: Assist in determining the amount of heat required; A multivariable decoupling compensation structure is adopted, with a control cycle of 2 seconds; the temperature deviation of the oxygen-permeable combustion layer is used. As the main control input, T_target is set to 880℃ (higher than the standard requirement of 850℃ to allow for margin). A digital PI controller is used.

[0042] After tuning, K_p=12, Ki=0.8, the output u_PI is in kilowatts, and the range is limited to 0-200 kilowatts; The real-time CO concentration C(k) is directly used as the disturbance compensation input; the compensation coefficient lambda is set to 0.015, that is:

[0043] When the CO concentration reaches 1000ppm, the compensation is 15 kW, which effectively suppresses the formation of incomplete combustion products; The layered sharpness index I_layer(k) is used as the stability compensation input. The compensation coefficient mu is set to 80, i.e.:

[0044] When the stratification is good (I_layer=0.9), the compensation is 8 kW; when the stratification becomes blurred (I_layer=0.5), the compensation increases to 40 kW, and the system actively increases the heat input to stabilize the pyrolysis process. Final target heating amount for: , This model achieves synergistic control of temperature stability, combustion completeness, and pyrolysis process stability; Step S6: Control Execution and Cycle; The control module converts the calculated Q_a and Q_b into specific control signals, which are sent to the frequency converters of the primary / secondary air fans and the fuel regulating valves of the auxiliary burners, respectively; The actuators respond to these signals to achieve precise adjustment, and the system then returns to step S1 to start the next control cycle, forming a closed loop.

[0045] Step 7: Furnace Health Scoring Model; The control module periodically runs an evaluation module; This module extracts features based on the recent time-series data of T, C, I_smoke, and I_layer, including mean, variance, trend slope, and frequency of exceeding limits, and inputs them into a scoring model trained based on a random forest regression algorithm, outputting a furnace health score H of 0-100. When H is below the first threshold, such as 80 points, the system provides optimization suggestions on the human-machine interface; when H is below the second threshold, such as 60 points, an audible and visual alarm is triggered and a predefined safety protocol may be executed. Parameter self-optimization mechanism: The system adopts an edge-cloud collaborative architecture; the on-site control module acts as an edge node, responsible for executing the real-time closed-loop control of S1-S6 mentioned above; at the same time, it uploads encrypted anonymous operating data to the cloud server; a reinforcement learning (RL) agent is deployed in the cloud, which takes process state characteristics as its state, fine-tuning of the edge controller parameters (K_p,K_i,K_s,λ,μ) as its "action", and comprehensive operating effects, including temperature stability, low emissions, and low energy consumption, as its "reward", and performs offline training in a simulation environment or historical data; after training, the cloud sends the optimized parameter set to the edge node for updates, thereby realizing the continuous autonomous evolution of the control model and adapting to different garbage characteristics and operating conditions.

[0046] This system underwent a 72-hour continuous operation test at the municipal solid waste treatment project in Bairinzuo Banner, Inner Mongolia (with a daily processing capacity of 15 tons). Recorded data showed that the temperature of the superoxide combustion chamber remained stable between 865 and 895℃, with a standard deviation of 12.3℃, a 62% decrease compared to the 32.7℃ observed during manual operation. The average CO concentration was 215 ppm, a 50% decrease compared to the 430 ppm observed during manual operation. The average auxiliary fuel consumption was 4.8 kg / hour, a 33% decrease compared to 7.2 kg / hour for manual operation; 98.5% of the smoke opacity was below Ringelmann level 1, and no black smoke above Ringelmann level 3 was observed. The average loss on ignition of the slag was 3.2%, which is better than the national standard requirement of 5%.

[0047] The above data fully demonstrates that the algorithm of this invention operates stably and has significant control effects in actual engineering environments.

Claims

1. A layered temperature-controlled intelligent monitoring system for pyrolysis gasification, characterized in that, include: The pyrolysis furnace has an exhaust pipe at the center of its top and a feed inlet on the side wall of the exhaust pipe. Its main body is a cylindrical structure, with the top and bottom connected to the cylinder by upper and lower conical structures, respectively. An observation port for observing the state inside the furnace is provided in the lower half of the cylinder. The monitoring module includes multiple temperature and pressure sensors installed at different heights within the furnace body, a flue gas analyzer for monitoring flue gas composition, and at least one high-temperature camera device positioned aligned with the observation port to capture images of the pyrolysis and gasification process of materials inside the furnace. The execution module includes an air supply system for supplying air into the furnace and an auxiliary burner for auxiliary heating; The control module is communicatively connected to both the monitoring module and the execution module, and is configured as follows: S1: Acquire real-time data from the monitoring module, including combustion zone temperature T, CO concentration C, flue gas image I_s, and pyrolysis furnace image I_f; S2: Extract and classify features from the flue gas image I_s, and calculate the flue gas state index I_smoke, which represents the degree of incomplete combustion; S3: Perform image analysis on the image I_f inside the pyrolysis furnace, identify the layering interface of the waste layer, and calculate the layer clarity index I_layer, which characterizes the quality of the layering. S4: Based on the CO concentration C, the flue gas state index I_smoke and the preset target CO concentration C_target, the target air supply volume Q_a of the air supply system is calculated through the first control model; S5: Based on the combustion zone temperature T, CO concentration C, stratification clarity index I_layer, and preset target temperature T_target, the target heating amount Q_b of the auxiliary combustion system is calculated through the second control model; S6: Output the calculated Q_a and Q_b to the air supply system and auxiliary combustion system for execution, so as to maintain the temperature of the oxygen-overflow combustion zone at no lower than the preset high temperature threshold and maintain the stable distribution of the pyrolysis and gasification reaction of the material in the furnace in each zone.

2. The stratified temperature-controlled pyrolysis gasification intelligent monitoring system according to claim 2, characterized in that, In step S2, the flue gas state index I_smoke, which characterizes the degree of incomplete combustion, is calculated as follows: S21: Perform color space conversion and region segmentation on the smoke image I_s, and extract the overall color histogram features of the smoke region; S22: Input the color histogram features into a pre-trained classification model. The classification model outputs the probability distribution of flue gas belonging to a preset combustion state category. The combustion state category includes at least complete combustion, incomplete combustion, and excessive oxygen or high humidity. S23: Based on the probability distribution and the preset weight vector, the scalar form of the flue gas state index I_smoke is calculated, with the incomplete combustion category having the highest weight.

3. The stratified temperature-controlled pyrolysis gasification intelligent monitoring system according to claim 1, characterized in that, In step S3, the layer sharpness index I_layer, which characterizes the quality of layering, is calculated as follows: S31: Perform grayscale conversion and vertical gradient calculation on the furnace image I_f to obtain the gradient intensity image; S32: In the gradient intensity image, find the gradient extremum points along the vertical direction, the extremum points corresponding to the potential interfaces between different garbage layers; S33: Statistically analyze the number of extreme points, average gradient intensity, and uniformity of distribution within the preset height range; S34: Based on the number of extreme points, average gradient strength, and distribution uniformity, the layered sharpness index I_layer is calculated using a weighted fusion algorithm.

4. The stratified temperature-controlled pyrolysis gasification intelligent monitoring system according to claim 1, characterized in that, The first control model in step S4 is: ； Where K_p1, K_i1, and K_d1 are PID control parameters for CO concentration deviation, and K_s is the compensation coefficient for the flue gas state index.

5. The stratified temperature-controlled pyrolysis gasification intelligent monitoring system according to claim 1, characterized in that, The second control model in step S5 is: ； Where K_p2 and Ki2 are PI control parameters for temperature deviation. This is the compensation coefficient for CO concentration. This is the compensation coefficient for the layered sharpness index. When the layered quality is poor and the I_layer value is low, the compensation amount is increased. Increase.

6. The stratified temperature-controlled pyrolysis gasification intelligent monitoring system according to claim 1, characterized in that, The control module is also configured to execute step S7: S7: Based on the current Q_a, Q_b, T, C, I_smoke and I_layer data, calculate a comprehensive furnace health score H using the third evaluation model; When H is below the first health threshold, an optimization suggestion signal is generated, prompting adjustment of at least one control parameter in the first control model and / or the second control model. When H falls below the second health threshold, which is below the first health threshold, a fault alarm signal is generated and a safety protocol is executed.

7. The stratified temperature-controlled pyrolysis gasification intelligent monitoring system according to claim 1, characterized in that, The control module includes an edge computing module and a cloud collaboration module; The edge computing module is deployed near the pyrolysis gasifier body and is used to execute steps S1 to S6 to achieve real-time closed-loop control at the millisecond to second level. The cloud collaboration module communicates with the edge computing module to receive historical operating data, perform offline optimization of the control parameters in the first control model and the second control model, and send the optimized parameters to the edge computing module for updating.

Images