Incinerator dioxin precursor sensing and inhibiting system and method

Abstract

The invention relates to the technical field of waste incineration pollution control, and discloses an incinerator dioxin precursor sensing and inhibiting system and method.The method comprises the steps that multi-mode original data are synchronously collected through a sensor array and input into a pre-trained multi-mode large model for inversion, and a three-dimensional initial chlorine concentration distribution field of a hearth is obtained; substituting into an initial risk field formula to calculate a whole-hearth initial risk field; predicting chlorine concentration distribution in future set time through a physical and chemical constraint model in combination with numerical data, and substituting the chlorine concentration distribution into a prediction risk field formula to calculate a dynamic prediction risk field; monitoring the dynamic prediction risk field in real time, screening out a dioxin precursor generation hotspot, calling a preset suppression strategy library to generate a suppression instruction, and outputting the suppression instruction to execution equipment; and acquiring suppressed multi-modal data to invert actual CI distribution and a risk field so as to evaluate the effectiveness of a suppression instruction, and completing model parameter optimization. According to the method, in-furnace CI distribution situation ultra-early perception, dioxin generation risk prediction and active inhibition can be carried out.

Description

Technical Field

[0001] This invention relates to the field of waste incineration pollution control technology, and in particular to a dioxin precursor sensing and inhibition system and method for incinerators. Background Technology

[0002] Dioxins are highly toxic byproducts produced during waste incineration, and their formation mechanism is closely related to specific reaction conditions within the furnace. Dioxins are mainly generated from precursors such as chlorobenzene and chlorophenol through heterogeneous catalytic reactions within a specific temperature window. Due to the complex and variable composition of waste, the migration, transformation, and spatial distribution of chlorine carried within the waste exhibit significant heterogeneity, easily forming localized chlorine-rich micro-regions. These micro-regions provide the necessary material basis and reaction environment for dioxin formation.

[0003] Currently, industry control of dioxins mainly relies on offline or online monitoring of flue gas concentrations. This method is a typical "post-event control," meaning it can only detect dioxins after they have been generated and emitted with the flue gas, and cannot intervene during their formation stage. Furthermore, traditional technologies also include methods to indirectly infer dioxin formation risk by monitoring the temperature of key temperature zones or the concentration of specific gases at the outlet. However, these methods fail to directly perceive the real-time spatial distribution of chlorine, the fundamental substance in dioxin formation, resulting in a lack of direct evidence for risk assessment.

[0004] The aforementioned existing technologies generally suffer from serious lag and blindness, failing to achieve ultra-early warning of dioxin formation and making it difficult to implement precise suppression at the source of formation. They cannot meet the needs of efficient and precise pollution control for waste incineration. Therefore, there is an urgent need to develop a technical solution that can directly sense the distribution of chlorine, predict the formation risk in advance, and intervene precisely. Summary of the Invention

[0005] This invention provides a dioxin precursor sensing and suppression system and method for incinerators, which can perform ultra-early sensing of CI distribution in the furnace, predict the risk of dioxin formation, and actively suppress it.

[0006] This invention provides a method for sensing and inhibiting dioxin precursors in incinerators, comprising:

[0007] S1. Synchronously acquire multimodal raw data through a sensor array and preprocess the multimodal raw data; wherein, the multimodal raw data includes visual data, numerical data and text data;

[0008] S2. Input the pre-processed multimodal raw data into the pre-trained multimodal large model for inversion to obtain the three-dimensional initial chlorine concentration distribution field CCI0(x,y,z) of the furnace. Substitute the initial chlorine concentration distribution field CCI0(x,y,z) into the initial risk field formula to calculate the initial risk field Ψ0(x,y,z) of the entire furnace.

[0009] S3. Based on the initial risk field Ψ0(x,y,z) and numerical data, and combined with a preset physicochemical constraint model, predict the chlorine concentration distribution (CCI) within a set time period using a transfer function. pred (t+Δt), the chlorine concentration distribution CCI pred Substituting (t+Δt) into the formula for predicting the risk field, we can calculate the dynamic predicted risk field Ψ. pred (t+Δt);

[0010] S4. Real-time monitoring of the dynamically predicted risk field Ψ pred (t+Δt), and screen out the predicted risk fields that exceed the preset safety threshold as dioxin precursor generation hotspots, and generate suppression instructions by calling the preset suppression strategy library according to the type of the dioxin precursor generation hotspots, such as the main combustion zone, transition zone or wide zone, and output them to the execution device; wherein, the suppression instructions include precise air curtain intervention, inhibitor injection or overall furnace temperature increase;

[0011] S5. Collect suppressed multimodal data through a sensor array, and invert the actual CI distribution and risk field Ψ using a closed-loop control evaluation formula. actual , in contrast to Ψ pred With Ψ actual The effectiveness of the suppression instructions is evaluated, and data is fed back to the multimodal large model and the physicochemical constraint model to complete parameter optimization, and the preset suppression strategy library is updated synchronously.

[0012] Furthermore, S1 specifically includes:

[0013] S101. Acquire real-time images of the flame shape, brightness, color distribution, and material spill trajectory inside the incinerator as image sequence data V.

[0014] S102. Real-time reading of numerical data N of key operating parameters in each area of ​​the furnace; wherein, the key operating parameters include temperature T, primary / secondary air volume and velocity F, furnace pressure P, and flue gas oxygen content O2;

[0015] S103. Obtain the component analysis report of the waste entering the incinerator as text data. If it is a batch feeding, the text data is updated synchronously according to the feeding batch.

[0016] S104. Perform data preprocessing on the image sequence data V, numerical data N, and text data Text, including noise filtering, image enhancement, size normalization, outlier removal, data smoothing, and format standardization.

[0017] Furthermore, S2 specifically includes:

[0018] S201. Input the preprocessed image sequence data V, numerical data N, and text data Text into the pre-trained multimodal large model. The multimodal large model performs feature extraction, fusion, and inference operations on the input data to invert and obtain the initial chlorine concentration values ​​corresponding to each coordinate point (x, y, z) in the three-dimensional space of the furnace, forming a complete three-dimensional initial chlorine concentration distribution field CCI0(x, y, z) of the furnace.

[0019] S202. Obtain the initial temperature T(x,y,z) and initial flue gas residence time τ(x,y,z) corresponding to each coordinate point (x,y,z) in the initial chlorine concentration distribution field CCI0(x,y,z). Combine the preset catalytic factor function (Catal(x,y,z)) and fixed adjustment coefficients α, β, γ, substitute the above parameters into the formula for the initial risk field of the entire furnace, calculate the initial risk value of each coordinate point (x,y,z) in the three-dimensional space of the furnace, and form the initial risk field Ψ0(x,y,z) of the entire furnace.

[0020] Furthermore, in S202, the formula for the initial risk field of the entire furnace is:

[0021] Ψ0(x,y,z)=(CCI0(x,y,z) α )·exp(-β / T(x,y,x))·(τ(x,y,z) γ )·η(Catal(x,y,z))

[0022] Wherein, Ψ0(x,y,z) is the initial risk field of each coordinate point (x,y,z) in the three-dimensional space of the furnace, which quantifies the initial risk of dioxin precursor formation at each location in the entire furnace; CCI0(x,y,z) is the initial chlorine concentration distribution field of each coordinate point (x,y,z) in the three-dimensional space of the furnace; α is the chlorine concentration influence adjustment coefficient, which reflects the nonlinear influence of chlorine concentration on risk;

[0023] T(x,y,x) represents the initial temperature at each coordinate point (x,y,z) in the three-dimensional space of the furnace; β is the activation energy-related parameter, reflecting the influence of temperature on the reaction rate; τ(x,y,z) represents the initial flue gas residence time at each coordinate point (x,y,z) in the three-dimensional space of the furnace; γ is the time influence coefficient, characterizing the degree of influence of residence time on precursor formation; η(Catal(x,y,z)) represents the catalytic factor at each coordinate point (x,y,z) in the three-dimensional space of the furnace, which is related to the concentration and type of metal catalyst in the fly ash at the corresponding location.

[0024] Furthermore, S3 specifically includes:

[0025] S301. Extract the three-dimensional initial chlorine concentration distribution field CCI0(x,y,z) of the furnace and the initial risk field Ψ0(x,y,z) of the entire furnace. At the same time, retrieve the preprocessed current numerical data N and perform consistency verification on the extracted data.

[0026] S302. Activate the preset simplified computational fluid dynamics and chemical reaction kinetics model as the physicochemical constraint model. Input the validated initial chlorine concentration distribution field CCI0(x,y,z) and the current numerical data N into the physicochemical constraint model. Perform calculations through the transfer function G to predict the chlorine concentration values ​​at each coordinate point (x,y,z) in the three-dimensional space of the furnace at time t+Δt within a future set time, forming the predicted chlorine concentration distribution field CCI0 at time t+Δt. pred (t+Δt), its calculation formula is:

[0027] CCI pred (t+Δt)=G(CCI0,F,T,...)

[0028] Among them, CCI pred (t+Δt) represents the predicted chlorine concentration distribution in the furnace at time t+Δt; G is the transfer function, describing the transport and transformation of chlorine under the influence of airflow, diffusion, and reaction; CCI0 is the initial chlorine concentration distribution field; F is the airflow rate and velocity of the primary / secondary airflow; T is the current furnace temperature field; "..." represents other parameters that affect the evolution of chlorine concentration.

[0029] S303, based on CCI pred (t+Δt), the temperature distribution T at time t+Δt predicted by the physicochemical constraint model. pred (t+Δt), flue gas residence time distribution τ pred (t+Δt), Catalyst parameter distribution Catal pred (t+Δt), using fixed adjustment coefficients α, β, γ and the catalytic factor function η(·), are substituted into the predicted risk field formula to calculate the predicted risk value of each coordinate point in the three-dimensional space of the furnace at time t+Δt, forming a dynamic predicted risk field Ψ.pred (t+Δt).

[0030] Furthermore, in S303, the formula for predicting the risk field is:

[0031] Ψ pred (t+Δt)=(CCI pred (t+Δt) α )·exp(-β / T pred (t+Δt))·(τ pred (t+Δt) γ )·η(Catal pred (t+Δt))

[0032] Among them, Ψ pred (t+Δt) represents the predicted risk field at the future time t+Δt, reflecting the risk of dioxin precursor formation in various regions of the furnace in the short term; CCI pred (t+Δt) represents the predicted chlorine concentration distribution at time t+Δt; α is the chlorine concentration influence adjustment coefficient; T pred (t+Δt) represents the predicted furnace temperature distribution at the future time t+Δt; β represents the activation energy-related parameter; τ pred (t+Δt) represents the predicted flue gas residence time distribution at future time t+Δt; γ is the time influence coefficient; Catal pred (t+Δt) represents the predicted distribution of catalyst-related parameters at the future time t+Δt, η(Catal) pred (t+Δt)) represents the catalytic factor at the corresponding position.

[0033] Furthermore, S4 specifically includes:

[0034] S401, Read the dynamically predicted risk field Ψ pred (t+Δt) performs a real-time, no-dead-angle scan of the entire furnace space, synchronously recording Ψ at each coordinate point. pred Numerical values ​​and corresponding spatial location information (x, y, z);

[0035] S402. Set a critical risk value for dioxin precursor formation as a safety threshold Ψ threshold Ψ will be monitored in real time pred (t+Δt) and Ψ threshold Compare them one by one and filter out all Ψ pred >Ψ threshold The region serves as a hotspot for the formation of dioxin precursors;

[0036] S403. Call the preset furnace partition parameters of the incinerator, match the spatial coordinates (x,y,z) of the dioxin precursor generation hotspots with the partition boundaries, and classify the hotspot types: wherein, the furnace partition parameters are the predefined physical range coordinate boundaries of the main combustion zone and the transition zone.

[0037] S404. According to the hotspot type, call the preset suppression strategy library to match the corresponding strategy, generate a suppression command, and transmit the suppression command to the execution device corresponding to the incinerator in real time to trigger the corresponding suppression operation; wherein, the suppression strategy library stores the corresponding mapping rules of hotspot type-suppression strategy-execution parameter in advance.

[0038] Furthermore, S403 specifically includes:

[0039] If the hot spot is entirely or mainly located in the preset main combustion zone, it is determined to be a hot spot in the main combustion zone; wherein, the main combustion zone is the core combustion area in the middle of the furnace, and the corresponding coordinate range is a preset fixed value;

[0040] If all or most of the hot spots are located in the preset transition zone, they are determined to be transition zone hot spots; wherein, the transition zone is the connection area between the main combustion zone and the tail flue, and the corresponding coordinate range is a preset fixed value.

[0041] If a hotspot's coverage area spans a single partition boundary, or is scattered across multiple local areas and its overall coverage area exceeds the range of a single partition, it is considered a widespread hotspot.

[0042] Furthermore, S5 specifically includes:

[0043] S501. After the device executing the suppression command completes its operation, a new round of multimodal data acquisition is initiated. The acquisition method is the same as in step S1 and maintains real-time performance.

[0044] S502. Input the newly acquired multimodal data into the pre-trained multimodal large model to invert and obtain the actual chloride concentration distribution field (CCI) after the suppression operation. actual (x,y,z), then CCI actual Substituting the corresponding actual parameters into the evaluation formula of the closed-loop control version, the suppressed actual risk field Ψ is calculated. actual The formula is as follows:

[0045]

[0046] Where A is the quantification value of the specific suppression action performed, and K is the preset suppression efficiency coefficient;

[0047] S503, Extract the dynamically predicted risk field Ψ pred The risk value of the corresponding hotspot region in (t+Δt) is compared with the actual risk field Ψ obtained in step S502.actual The risk values ​​of the same spatial coordinates (x, y, z) are compared one by one:

[0048] If Ψ actual Below Ψ pred (t+Δt) and below the safety threshold Ψ threshold The current suppression strategy is deemed effective.

[0049] If Ψ actual Not reduced or still higher than Ψ threshold The current suppression strategy is deemed ineffective or invalid.

[0050] S504. Combine the comparison results from step S503, the corresponding suppression action A, the multimodal data acquired after suppression, and the actual CI distribution field CCI. actual As new training samples, they are fed back to the multimodal large model and the physicochemical constraint model, and the inhibition strategy library is automatically updated based on the model learning results and the inhibition effect evaluation conclusions.

[0051] The present invention also provides a dioxin precursor sensing and suppression system for incinerators, based on the dioxin precursor sensing and suppression method for incinerators as described above, the system comprising:

[0052] The acquisition module is used to synchronously acquire multimodal raw data through a sensor array and preprocess the multimodal raw data; wherein the multimodal raw data includes visual data, numerical data and text data;

[0053] The first calculation module is used to input the pre-processed multimodal raw data into the pre-trained multimodal large model for inversion to obtain the three-dimensional initial chlorine concentration distribution field CCI0(x,y,z) of the furnace, and substitute the initial chlorine concentration distribution field CCI0(x,y,z) into the initial risk field formula to calculate the initial risk field Ψ0(x,y,z) of the entire furnace.

[0054] The second calculation module is used to predict the chlorine concentration distribution (CCI) within a set future time period based on the initial risk field Ψ0(x,y,z) and numerical data, combined with a preset physicochemical constraint model, using a transfer function. pred (t+Δt), the chlorine concentration distribution CCI pred Substituting (t+Δt) into the formula for predicting the risk field, we can calculate the dynamic predicted risk field Ψ. pred (t+Δt);

[0055] The generation module is used to monitor the dynamically predicted risk field Ψ in real time. pred(t+Δt), and screen out the predicted risk fields that exceed the preset safety threshold as dioxin precursor generation hotspots, and generate suppression instructions by calling the preset suppression strategy library according to the type of the dioxin precursor generation hotspots, such as the main combustion zone, transition zone or wide zone, and output them to the execution device; wherein, the suppression instructions include precise air curtain intervention, inhibitor injection or overall furnace temperature increase;

[0056] The update module is used to collect suppressed multimodal data through a sensor array and invert the actual CI distribution and risk field Ψ using a closed-loop control evaluation formula. actual , in contrast to Ψ pred With Ψ actual The effectiveness of the suppression instructions is evaluated, and data is fed back to the multimodal large model and the physicochemical constraint model to complete parameter optimization, and the preset suppression strategy library is updated synchronously.

[0057] The present invention also provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps of the above-described method.

[0058] The present invention also provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the steps of the above-described method.

[0059] The beneficial effects of this invention are as follows:

[0060] This invention overcomes the "black box" problem of furnace operating conditions by sensing the distribution of chlorine, the fundamental precursor to dioxin formation, and utilizing a multimodal large-scale model. It achieves precise sensing of chlorine distribution within the furnace and ultra-early warning of dioxin formation risks, successfully promoting the shift of dioxin control from post-event treatment to pre-event prevention. Simultaneously, based on dynamically predicted risk fields, it generates and executes precise spatially oriented suppression strategies, effectively avoiding the blindness and resource waste of traditional extensive control methods, achieving proactive and precise pollution suppression. Furthermore, the system continuously self-corrects and optimizes through a closed-loop feedback mechanism, possessing self-learning and evolutionary capabilities, and can adapt to complex and ever-changing waste incineration conditions over the long term, ensuring the robustness and stability of the control effect. Attached Figure Description

[0061] Figure 1 This is a schematic diagram of a method flow according to an embodiment of the present invention.

[0062] Figure 2 This is a schematic diagram of the device structure according to an embodiment of the present invention.

[0063] Figure 3 This is a schematic diagram of the internal structure of a computer device according to an embodiment of the present invention.

[0064] The realization of the objective, functional features and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0065] It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.

[0066] like Figure 1 As shown, the present invention provides a method for sensing and inhibiting dioxin precursors in an incinerator, comprising:

[0067] S1. Synchronously acquire multimodal raw data through a sensor array and preprocess the multimodal raw data; wherein, the multimodal raw data includes visual data, numerical data and text data.

[0068] S101, Visual Modal Data Acquisition: Real-time images of flame shape, brightness, color distribution, and material spill trajectory inside the incinerator are captured by a high-temperature industrial endoscope and CCD / infrared camera installed at the observation port of the incinerator. These images are used as image sequence data V and stored synchronously according to timestamps.

[0069] S102. Numerical Modal Data Acquisition: Through the data interface of the Distributed Control System (DCS), the numerical data N of key operating parameters in each area of ​​the furnace are read in real time to ensure that the data acquisition interval does not exceed the preset threshold (e.g., 1 second / time); wherein, the key operating parameters include temperature T, primary / secondary air volume and velocity F, furnace pressure P, and flue gas oxygen content O2.

[0070] S103, Text Modal Data Acquisition: Connect to the garbage crane weighing and recognition system to automatically obtain the composition analysis report of the garbage entering the incinerator (including key information such as the proportion of chlorinated plastics and the proportion of kitchen waste) as text data. If it is a batch feeding, the text data will be updated synchronously according to the feeding batch.

[0071] S104. Perform data preprocessing on the image sequence data V, numerical data N, and text data Text, including noise filtering, image enhancement, size normalization, outlier removal, data smoothing, and format standardization. Specifically,

[0072] ① Visual modal data preprocessing: The acquired image sequence V is subjected to noise filtering (removing image noise under high temperature environment), image enhancement (improving the contrast between flame and material areas), size normalization (unifying image resolution) and time series alignment processing to output a standardized visual data subset;

[0073] ② Numerical modal data preprocessing: The read numerical data N is subjected to outlier removal (filtering outlier data caused by sensor failure through a preset threshold range), data smoothing (eliminating instantaneous fluctuation interference), and unit unification (converting numerical data from different sources into preset standard units), and a standardized numerical data subset is output.

[0074] ③ Text modal data preprocessing: The waste component analysis report Text is processed by structure extraction (extracting key quantitative information such as chlorine content and component proportion from unstructured text) and format standardization (converting to a unified data format) to output a standardized text data subset.

[0075] Finally, the preprocessed standardized visual data, numerical data, and text data are linked and integrated according to timestamps to form a multimodal data set in a unified format, which is stored in the system's designated database. At the same time, a data validity verification report is output to provide qualified and standardized data support for the input of the multimodal large model in step S2.

[0076] S2. Input the pre-processed multimodal raw data into the pre-trained multimodal large model for inversion to obtain the three-dimensional initial chlorine concentration distribution field CCI0(x,y,z) of the furnace. Substitute the initial chlorine concentration distribution field CCI0(x,y,z) into the initial risk field formula to calculate the initial risk field Ψ0(x,y,z) of the entire furnace.

[0077] S201. Extract the standardized multimodal dataset that has been preprocessed and integrated from the specified database output in step S1. At the same time, retrieve the data validity verification report to confirm that there are no missing or abnormal data in each subset of visual, numerical, and text data, and that they meet the input format requirements of the multimodal large model. Start the pre-trained multimodal large model, load the optimal parameters of the model training (including the multimodal information learned from historical data and the CI distribution mapping parameters), and ensure that the model is in a ready state.

[0078] The preprocessed image sequence data V, numerical data N, and text data Text are input into a pre-trained multimodal large model. The multimodal large model learns the complex nonlinear mapping relationship between multimodal information and CI distribution based on massive historical data (including CI content data sampled and measured under specific working conditions). For example, a high-brightness yellow flame combined with a specific wind field corresponds to a high CCI region, and a high chlorine content in the waste causes the CCI baseline to shift upward. Feature extraction, fusion, and inference operations are performed on the input data to invert the initial chlorine concentration value corresponding to each coordinate point (x,y,z) in the three-dimensional space of the furnace, forming a complete three-dimensional initial chlorine concentration distribution field CCI0(x,y,z) of the furnace.

[0079] The pre-trained multimodal large model is an artificial intelligence model trained in advance based on massive historical multimodal data under incinerator operation scenarios (including visual modal flame / material image sequences, numerical modal operating parameters such as temperature / airflow / pressure, textual modal waste composition analysis reports, and CI content data sampled and measured under specific operating conditions). The core of this model is to learn the complex nonlinear mapping relationship between multi-source heterogeneous information such as visual, numerical, and textual data and the chlorine element CI distribution in the furnace (such as the high-brightness yellow flame + specific air field corresponding to the high CCI area of ​​concentrated PVC combustion, and the overall CCI baseline shifting upward when the waste has a high chlorine content). It has the ability to accurately invert the initial chlorine concentration distribution field CCI0(x,y,z) in the three-dimensional space of the furnace after receiving real-time pre-processed multimodal data, providing core input data for subsequent initial risk field calculations. This model is a pre-trained multimodal fusion model customized for incinerator operating conditions. Its core is a dedicated artificial intelligence model adapted to the complex physicochemical environment inside the waste incinerator. Essentially, it is based on a deep learning architecture and is specifically optimized for the fusion capability of multi-source heterogeneous data (visual, numerical, and textual). It is trained using massive historical operating data of incinerators (including measured samples of chlorine content). Its core function is to accurately establish the nonlinear mapping relationship between multimodal information and the distribution of chlorine in the furnace, realize the dynamic inversion and prediction of the three-dimensional chlorine distribution field in the furnace, and provide core data support for the risk assessment of dioxin precursor formation.

[0080] S202. From the numerical data subset of the multimodal dataset, obtain the initial temperature T(x,y,z) and initial flue gas residence time τ(x,y,z) corresponding to each coordinate point (x,y,z) in the initial chlorine concentration distribution field CCI0(x,y,z). Combine the preset catalytic factor function η(Catal(x,y,z)) (which is related to the concentration and type of metal catalysts such as Cu and Fe in the fly ash at the corresponding coordinate point) and fixed adjustment coefficients α, β, and γ. Substitute the above parameters into the formula for the initial risk field of the entire furnace to calculate the initial risk value of each coordinate point (x,y,z) in the three-dimensional space of the furnace, thus forming the initial risk field Ψ0(x,y,z) of the entire furnace.

[0081] The basic formula for the initial CI distribution trend index is as follows:

[0082] Ψ0=f(CCI,T,τ,Catal)≈(CCI α )·exp(-β / T)·(τ γ )·η(Catal)

[0083] Among them, Ψ0 is the initial CI distribution trend index, which quantifies the potential risk of dioxin precursor formation in a specific area of ​​the furnace at a specific time; the higher the value, the higher the risk. CCI is the local chlorine concentration, which is the core variable for perception and inversion, reflecting the degree of chlorine accumulation in a local area of ​​the furnace. α is the chlorine concentration influence adjustment coefficient, usually α≥1, reflecting the nonlinear influence of chlorine concentration on risk; local enrichment will significantly amplify the risk. T is the local temperature, which affects the rate of dioxin precursor formation reaction. β is the activation energy related parameter, used to reflect the Arrhenius relationship between reaction rate and temperature within a specific temperature window (250-450℃). τ is the flue gas residence time, characterizing the length of time the flue gas stays in this area of ​​the furnace; under incomplete combustion conditions, the longer the residence time, the more complete the precursor formation. γ is the time influence coefficient, characterizing the degree of influence of flue gas residence time on dioxin precursor formation. η is the catalytic factor, which is a function related to the concentration and type of metal catalysts such as Cu and Fe in fly ash, affecting the catalytic efficiency of the reaction.

[0084] The formula for the initial risk field of the entire furnace is:

[0085] Ψ0(x,y,z)=(CCI0(x,y,z) α )·exp(-β / T(x,y,x))·(τ(x,y,z) γ )·η(Catal(x,y,z))

[0086] Wherein, Ψ0(x,y,z) is the initial risk field of each coordinate point (x,y,z) in the three-dimensional space of the furnace, which quantifies the initial risk of dioxin precursor formation at each location in the entire furnace; CCI0(x,y,z) is the initial chlorine concentration distribution field of each coordinate point (x,y,z) in the three-dimensional space of the furnace; α is the chlorine concentration influence adjustment coefficient, which reflects the nonlinear influence of chlorine concentration on risk;

[0087] T(x,y,x) represents the initial temperature at each coordinate point (x,y,z) in the three-dimensional space of the furnace; β is the activation energy-related parameter, reflecting the influence of temperature on the reaction rate; τ(x,y,z) represents the initial flue gas residence time at each coordinate point (x,y,z) in the three-dimensional space of the furnace; γ is the time influence coefficient, characterizing the degree of influence of residence time on precursor formation; η(Catal(x,y,z)) represents the catalytic factor at each coordinate point (x,y,z) in the three-dimensional space of the furnace, which is related to the concentration and type of metal catalyst in the fly ash at the corresponding location.

[0088] Finally, the rationality of the inverted CCI0(x,y,z) and the calculated Ψ0(x,y,z) is verified to ensure that the chlorine concentration values ​​are within the actual physical range and that the risk values ​​do not have abnormal abrupt changes. After the verification is passed, the initial chlorine concentration distribution field CCI0(x,y,z) and the initial risk field Ψ0(x,y,z) are stored in the system database in a three-dimensional grid data format. At the same time, a data output report (including information such as data range, accuracy, and validity) is generated to provide core input data for the spatiotemporal evolution prediction in step S3.

[0089] S3. Based on the initial risk field Ψ0(x,y,z) and numerical data, and combined with a preset physicochemical constraint model, predict the chlorine concentration distribution (CCI) within a set time period using a transfer function. pred (t+Δt), the chlorine concentration distribution CCI pred Substituting (t+Δt) into the formula for predicting the risk field, we can calculate the dynamic predicted risk field Ψ. pred (t+Δt).

[0090] S301. Extract the three-dimensional initial chlorine concentration distribution field CCI0(x,y,z) of the furnace and the initial risk field Ψ0(x,y,z) of the entire furnace, and simultaneously retrieve the preprocessed current numerical data N. satandard (Including real-time wind field, temperature field, pressure, flue gas oxygen content and other parameters), the extracted data are checked for consistency to ensure that the timestamps of the initial field data and the current numerical field data match, the coordinate system is the same, and the parameter units are consistent, so as to avoid data deviation from affecting the prediction results.

[0091] S302. Activate the preset simplified computational fluid dynamics and chemical reaction kinetics model as the physicochemical constraint model, load the model's preset parameters (including fluid diffusion coefficient, reaction rate constant, catalyst activity coefficient, etc.), and use this model as the physical constraint condition for the evolution of chlorine concentration and risk field. Define the core parameters of the transfer function G in the model (characterizing the influence weights of wind transport, diffusion, and chemical reactions on the migration and transformation of chlorine). Use the validated initial chlorine concentration distribution field CCI0(x,y,z) and the current numerical data N... satandard (Focusing on extracting primary / secondary air volume and velocity F, and current temperature field T) Input the aforementioned physicochemical constraint model, and perform calculations through the transfer function G to predict the chlorine concentration values ​​at each coordinate point (x, y, z) in the three-dimensional space of the furnace at time t+Δt within a future set time (preset to 30-60 seconds), forming the predicted chlorine concentration distribution field CCI at time t+Δt. pred (t+Δt), its calculation formula is:

[0092] CCI pred (t+Δt)=G(CCI0,F,T,...)

[0093] Among them, CCI pred (t+Δt) represents the predicted chlorine concentration distribution in the furnace at time t+Δt; G is the transfer function, describing the transport and transformation of chlorine under the influence of airflow, diffusion, and reaction; CCI0 is the initial chlorine concentration distribution field; F is the airflow rate and velocity of the primary / secondary airflow; T is the current furnace temperature field; "..." represents other parameters that affect the evolution of chlorine concentration.

[0094] The simplified computational fluid dynamics and chemical reaction kinetics model is a simplified multiphysics coupling model adapted to incinerator scenarios. Its core is the integration of the fundamental principles of computational fluid dynamics (CFD) and chemical reaction kinetics, while simplifying complex calculation processes to meet real-time prediction requirements. The CFD component primarily simulates the flow, mixing, diffusion, and transport processes of flue gas and material particles within the furnace, accurately depicting the influence of wind and temperature fields on the spatial migration of chlorine (corresponding to the core logic of the transfer function G in step S3). The chemical reaction kinetics component focuses on chlorine-related reactions (such as the formation reactions of dioxin precursors like chlorobenzene and chlorophenol), quantifying the correlation between reaction rates and parameters such as temperature and catalyst concentration, closely aligning with the reaction characteristics within the 250-450℃ danger temperature window. The simplified features, by omitting secondary physical / chemical processes and simplifying complex control equations, improve computational speed while maintaining prediction accuracy, ensuring the output of future chlorine concentration and risk field predictions within a short timeframe, adapting to the system's real-time early warning and control requirements.

[0095] Predicting chlorine concentration distribution over a predetermined time period using a transfer function relies on integrating the initial chlorine concentration state with the physicochemical interactions within the furnace using the transfer function G. This enables dynamic extrapolation from the initial field to the future field. The specific process is as follows:

[0096] Determine the input parameters of the transfer function: take the initial chlorine concentration distribution field CCI0(x,y,z) (initial state baseline) obtained in step S2 and the real-time numerical field data after preprocessing in step S1 (focusing on primary / secondary air volume and velocity F, current temperature field T, and also including auxiliary parameters such as pressure and flue gas oxygen content) as input to the transfer function G to ensure that the parameters cover the key influencing factors of chlorine migration and transformation.

[0097] The core operational logic of the transfer function: The transfer function G incorporates the core rules of a simplified model of computational fluid dynamics and chemical reaction kinetics. On the one hand, through fluid dynamics-related sub-modules, it simulates the convective transport of chlorine under wind-driven conditions (such as the flow trajectory of chlorine-rich gas masses with airflow) and diffusion processes (the diffusion range of chlorine to the surrounding area). On the other hand, through chemical reaction kinetics-related sub-modules, it quantifies the conversion rate of chlorine under the influence of factors such as temperature and catalysts (such as the efficiency of chlorine production / consumption in the reaction of chlorine with other substances within a temperature window of 250-450℃).

[0098] Setting the time step and iterative calculation: Define the future prediction time (e.g., 30-60 seconds, i.e., t+Δt), divide the total prediction time into several small time steps, and iterate the transfer function G step by step according to the time step. Each step is based on the chlorine concentration distribution result of the previous step, combined with the real-time updated numerical field parameters (e.g., dynamic changes in air volume and temperature), to correct the spatial distribution of chlorine concentration, and accumulate to complete the extrapolation of the entire time.

[0099] Outputting the chlorine concentration distribution at a future set time: After iterative calculation to the set time node, the transfer function G outputs the final result, that is, the chlorine concentration value at each coordinate point (x,y,z) in the three-dimensional space of the furnace at the future time t+Δt, forming a complete predicted chlorine concentration distribution field CCI. pred (t+Δt).

[0100] Throughout the process, the transfer function G encapsulates simplified physicochemical laws, transforming the complex migration and transformation process of chlorine into a computable mapping relationship, enabling accurate deduction from the initial state to the future state. Furthermore, the "simplification" characteristic ensures computational speed, meeting the needs of real-time prediction.

[0101] S303, based on CCI pred (t+Δt), the temperature distribution T at time t+Δt predicted by the physicochemical constraint model. pred (t+Δt), flue gas residence time distribution τ pred (t+Δt), Catalyst parameter distribution Catal pred (t+Δt), using fixed adjustment coefficients α, β, γ and the catalytic factor function η(·), are substituted into the predicted risk field formula to calculate the predicted risk value of each coordinate point in the three-dimensional space of the furnace at time t+Δt, forming a dynamic predicted risk field Ψ. pred (t+Δt) completes the update from the initial static risk field to the future dynamic risk field. The formula for predicting the risk field is:

[0102] Ψ pred (t+Δt)=(CCI pred (t+Δt) α ·exp(-β / T pred (t+Δt))·(τ pred (t+Δt) γ )·η(Catal pred (t+Δt))

[0103] Among them, Ψ pred (t+Δt) represents the predicted risk field at the future time t+Δt, reflecting the risk of dioxin precursor formation in various regions of the furnace in the short term; CCI pred(t+Δt) represents the predicted chlorine concentration distribution at time t+Δt; α is the chlorine concentration influence adjustment coefficient; T pred (t+Δt) represents the predicted furnace temperature distribution at the future time t+Δt; β represents the activation energy-related parameter; τ pred (t+Δt) represents the predicted flue gas residence time distribution at future time t+Δt; γ is the time influence coefficient; Catala pred (t+Δt) represents the predicted distribution of catalyst-related parameters at the future time t+Δt, η(Catal) pred (t+Δt)) represents the catalytic factor at the corresponding position.

[0104] Finally, regarding CCI pred (t+Δt) and Ψ pred (t+Δt) Physical rationality verification is performed (e.g., the predicted chlorine concentration value must be within a reasonable range under actual combustion conditions, and the trend of risk value change must conform to the influence law of parameters such as temperature and wind field); after the verification is passed, the predicted chlorine concentration distribution field and the dynamic predicted risk field are stored in the system's real-time database in a three-dimensional grid time series data format, and a prediction result report (including prediction duration, data accuracy, key influencing parameters, etc.) is output to provide real-time dynamic input data for the risk identification and suppression strategy generation in step S4.

[0105] S4. Real-time monitoring of the dynamically predicted risk field Ψ pred (t+Δt), and select the predicted risk fields that exceed the preset safety threshold as dioxin precursor generation hotspots, and generate suppression instructions by calling the preset suppression strategy library according to the type of the dioxin precursor generation hotspots, namely the main combustion zone, transition zone or wide zone, and output them to the execution device; wherein, the suppression instructions include precise air curtain intervention, inhibitor injection or overall furnace temperature increase.

[0106] S401, Read the dynamically predicted risk field Ψ pred (t+Δt) represents the potential risk value of dioxin precursor formation at each coordinate point in the three-dimensional space of the furnace within the next 30-60 seconds. A real-time, no-dead-angle scan of the entire furnace space is performed, and the Ψ value at each coordinate point is recorded synchronously. pred Numerical values ​​and corresponding spatial location information (x, y, z);

[0107] S402. Set a critical risk value for dioxin precursor formation as a safety threshold Ψ threshold Ψ will be monitored in real time pred (t+Δt) and Ψ threshold Compare them one by one and filter out all Ψ pred >Ψ threshold The region serves as a hotspot for the formation of dioxin precursors;

[0108] S403. Call the pre-defined furnace partition parameters (predefined physical range coordinate boundaries of the main combustion zone, transition zone, etc.) of the incinerator, and match the spatial coordinates (x, y, z) of the dioxin precursor generation hotspots with the partition boundaries to classify the hotspot types: wherein, the furnace partition parameters are the pre-defined physical range coordinate boundaries of the main combustion zone and transition zone, specifically including:

[0109] ① If all or most of the hot spots are located in the preset main combustion zone (the core combustion area in the middle of the furnace, with the corresponding coordinate range being a preset fixed value), they are determined to be hot spots in the main combustion zone.

[0110] ② If all or most of the hot spots are located in the preset transition zone (the area connecting the main combustion zone and the tail flue, with the corresponding coordinate range being a preset fixed value), they are determined to be hot spots in the transition zone.

[0111] ③ If the hotspot coverage area crosses the boundary of a single zone (such as involving both the main combustion zone and the transition zone), or is scattered across multiple local areas and the overall coverage area exceeds the range of a single zone, it is judged as a widespread hotspot.

[0112] S404. Based on the hotspot type, a preset suppression strategy library is invoked to match the corresponding strategy, generating a suppression command. This suppression command is then transmitted in real-time to the corresponding execution equipment (primary air regulation system, inhibitor injection system, furnace temperature control system) of the incinerator to trigger the corresponding suppression operation. The suppression strategy library pre-stores corresponding mapping rules between hotspot type, suppression strategy, and execution parameters. The suppression command specifically includes:

[0113] ① For hot spots in the main combustion zone: Match the "precise air curtain intervention" strategy. The instructions include the target primary air nozzle number (corresponding to the preset nozzle position based on the hot spot coordinates), the adjusted air volume value, and the nozzle angle parameters. The core purpose is to destroy the local chlorine-rich environment by blowing chlorine-rich gas masses in the heat dissipation area through a directional air curtain.

[0114] ② Targeting hotspots in the transition zone: Match the "Urea / Ammonium Sulfate Precision Injection" strategy. The instructions include the target spray gun number (corresponding to the preset spray gun position based on the hotspot coordinates), the amount of inhibitor injected, and the injection duration. The core purpose is to inhibit the formation of dioxin precursors by injecting urea or ammonium sulfate, using competitive adsorption or passivation of metal catalysts such as Cu and Fe in fly ash.

[0115] ③ Targeting a wide range of hotspots: Matching the "overall furnace temperature increase" strategy, the instructions include the overall target temperature value of the furnace and the heating rate parameters. The core purpose is to allow the materials in the furnace to quickly cross the dangerous temperature window of dioxin formation of 250-450℃, thereby blocking the formation of precursors from the temperature condition.

[0116] S5. After the corresponding equipment executing the suppression commands (precise air curtain intervention, precise urea / ammonium sulfate injection, overall furnace temperature increase, etc.) completes its operation, the sensor array collects the suppressed multimodal data, and the actual CI distribution and risk field Ψ are inverted through the closed-loop control evaluation formula. actual , in contrast to Ψ pred With Ψ actual The effectiveness of the suppression instructions is evaluated, and data is fed back to the multimodal large model and the physicochemical constraint model to complete parameter optimization, and the preset suppression strategy library is updated synchronously.

[0117] S501. After the device executing the suppression command completes its operation, a new round of multimodal data acquisition is initiated. The acquisition method is the same as in step S1 and maintains real-time performance.

[0118] ①Visual Modal: Continuously acquire image sequences of flame morphology, brightness, color distribution, and material spill trajectory after furnace suppression operations using a high-temperature industrial endoscope and CCD / infrared camera. actual ;

[0119] ② Numerical mode: The temperature T of each region after the suppression operation is read in real time from the DCS system. actual Primary / secondary air volume and velocity F actual Pressure P actual Oxygen content in flue gas 2actual Equal numerical data N actual ;

[0120] ③ Text Modality: If the suppression operation involves adjustments related to waste composition (such as optimization of subsequent batches of waste feeding), the updated waste composition analysis report from the waste crane weighing and identification system will be simultaneously accessed as text data. actual ;

[0121] For the collected V actual N actual Text actual Perform the same preprocessing as in step S1 (such as noise filtering, data normalization, format standardization, etc.) to ensure data availability.

[0122] S502, Transfer the newly acquired multimodal data V actual N actual Text actual The data is input into a pre-trained multimodal large model (consistent with the model used in step S2). The model inverts the actual chloride concentration distribution field (CCI) after the suppression operation based on the real-time input data. actual (x,y,z), then CCI actual and the corresponding actual parameters (T) actual τ actual Catal actualSubstituting the values ​​into the evaluation formula of the closed-loop control version, the actual risk field Ψ after suppression is calculated. actual The formula is as follows:

[0123]

[0124] Among them, Ψ actual This is used to evaluate the effectiveness of the suppression strategy by identifying the actual risk field after the suppression operation. (CCI) actual To suppress the actual chlorine concentration distribution after the operation. α is the chlorine concentration influence adjustment coefficient, consistent with the meaning of the aforementioned formula. T actual To suppress the actual furnace temperature after operation; β is a parameter related to activation energy, consistent with the meaning of the aforementioned formula. τ actual To suppress the actual flue gas residence time after operation. γ is the time influence coefficient, consistent with the meaning of the aforementioned formula. η(Catal) actual The catalytic factor after the suppression operation is related to the state of the metal catalyst in the fly ash at the corresponding location. K is the suppression efficiency coefficient, used to quantify the impact of the suppression action on risk reduction. A is the quantitative value of the suppression action, such as the quantitative indicators of specific suppression operations like the air volume adjustment amount of precise air curtain intervention, the urea / ammonium sulfate injection amount, and the furnace temperature increase value.

[0125] S503, Extract the dynamically predicted risk field Ψ pred The risk value of the corresponding hotspot region in (t+Δt) is compared with the actual risk field Ψ obtained in step S502. actual The risk values ​​of the same spatial coordinates (x, y, z) are compared one by one:

[0126] ①If Ψ actual Below Ψ pred (t+Δt) and below the safety threshold Ψ threshold The current suppression strategy is deemed effective.

[0127] ②If Ψ actual Not reduced or still higher than Ψ threshold The current suppression strategy is deemed ineffective or invalid.

[0128] S504, Combine the comparison results from step S503, the corresponding suppression action A, and the multimodal data V collected after suppression. actual N actual Text actual and actual CI distribution field CCI actual As new training samples, they are fed back to the multimodal large model and the physicochemical constraint model, specifically including:

[0129] ① The multimodal large model optimizes its learning of the mapping relationship between "multimodal information-CI distribution" based on new samples, thereby improving the accuracy of CI distribution inversion;

[0130] ② The spatiotemporal evolution prediction model is combined with new samples to adjust the physical and chemical constraint parameters (such as the coefficients of the transfer function G) and optimize the prediction accuracy of CCI and Ψ;

[0131] ③ Synchronously update the suppression efficiency coefficient K in the model to make it more closely match the actual suppression effect and optimize the calculation accuracy of the closed-loop control formula.

[0132] Finally, based on the model learning results and the evaluation conclusions of the inhibition effect, the inhibition strategy library is automatically updated, specifically including:

[0133] ① For effective strategies, record their corresponding hotspot type, operating conditions, and optimal execution parameters (such as the optimal air volume adjustment value corresponding to a specific main combustion zone hotspot) and optimize the strategy mapping rules;

[0134] ② For strategies that are ineffective or have poor results, mark the corresponding scenarios and adjust the parameters (such as increasing the amount of inhibitor sprayed, adjusting the angle of the air curtain, etc.), or supplement new suppression schemes;

[0135] ③ Save all data (input data, calculation results, suppression actions, evaluation conclusions) in this closed-loop process to form a system operation log, providing data support for subsequent model iteration and strategy optimization, and completing the complete closed loop of execution-monitoring-evaluation-optimization.

[0136] The present invention also provides a dioxin precursor sensing and suppression system for incinerators, based on the dioxin precursor sensing and suppression method for incinerators as described above, the system comprising:

[0137] The acquisition module 1 is used to synchronously acquire multimodal raw data through a sensor array and preprocess the multimodal raw data; wherein the multimodal raw data includes visual data, numerical data and text data;

[0138] The first calculation module 2 is used to input the pre-processed multimodal raw data into the pre-trained multimodal large model for inversion to obtain the three-dimensional initial chlorine concentration distribution field CCI0(x,y,z) of the furnace, and substitute the initial chlorine concentration distribution field CCI0(x,y,z) into the initial risk field formula to calculate the initial risk field Ψ0(x,y,z) of the entire furnace.

[0139] The second calculation module 3 is used to predict the chlorine concentration distribution (CCI) within a set time period in the future, based on the initial risk field Ψ0(x,y,z) and numerical data, combined with a preset physicochemical constraint model, through a transfer function. pred (t+Δt), the chlorine concentration distribution CCI pred Substituting (t+Δt) into the formula for predicting the risk field, we can calculate the dynamic predicted risk field Ψ. pred (t+Δt);

[0140] Module 4 is used to monitor the dynamically predicted risk field Ψ in real time. pred (t+Δt), and screen out the predicted risk fields that exceed the preset safety threshold as dioxin precursor generation hotspots, and generate suppression instructions by calling the preset suppression strategy library according to the type of the dioxin precursor generation hotspots, such as the main combustion zone, transition zone or wide zone, and output them to the execution device; wherein, the suppression instructions include precise air curtain intervention, inhibitor injection or overall furnace temperature increase;

[0141] Update module 5 is used to collect suppressed multimodal data through a sensor array, and to invert the actual CI distribution and risk field Ψ using a closed-loop control evaluation formula. actual , in contrast to Ψ pred With Ψ actual The effectiveness of the suppression instructions is evaluated, and data is fed back to the multimodal large model and the physicochemical constraint model to complete parameter optimization, and the preset suppression strategy library is updated synchronously.

[0142] Each of the above modules is used to perform the respective steps in the above-described method for sensing and inhibiting dioxin precursors in incinerators. The specific implementation methods are as described in the above-described method embodiments, and will not be repeated here.

[0143] like Figure 3 As shown, the present invention also provides a computer device, which may be a server, and its internal structure may be as follows: Figure 3 As shown, the computer device includes a processor, memory, network interface, and database connected via a system bus. The processor provides computational and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and database. The internal memory provides the environment for the operation of the operating system and computer programs in the non-volatile storage media. The database stores all data required for the process of the dioxin precursor sensing and suppression method in the incinerator. The network interface is used for communication with external terminals via a network connection. The computer program is executed by the processor to implement the dioxin precursor sensing and suppression method in the incinerator.

[0144] Those skilled in the art will understand that Figure 3 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer equipment on which the present application is applied.

[0145] An embodiment of this application also provides a computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements any of the above-described methods for detecting and suppressing dioxin precursors in incinerators.

[0146] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by hardware related to computer program instructions. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the above method embodiments. Any references to memory, storage, databases, or other media provided in this application and used in the embodiments can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM), such as dynamic RAM (used as main storage) or static RAM (commonly used as cache memory). By way of illustration and not limitation, RAM has various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDR SDRAM), and Rambus DRAM (RDRAM).

[0147] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, apparatus, article, or method that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, apparatus, article, or method. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, apparatus, article, or method that includes that element.

[0148] The above description is merely a preferred embodiment of the present invention and does not limit the patent scope of the present invention. Any equivalent structural or procedural transformations made based on the content of the present invention's specification and drawings, or direct or indirect applications in other related technical fields, are similarly included within the patent protection scope of the present invention.

Claims

1. A method for sensing and inhibiting dioxin precursors in an incinerator, characterized in that, include: S1. Synchronously acquire multimodal raw data through a sensor array and preprocess the multimodal raw data; wherein, the multimodal raw data includes visual data, numerical data and text data; S2. Input the pre-processed multimodal raw data into the pre-trained multimodal large model for inversion to obtain the three-dimensional initial chlorine concentration distribution field CCI0(x,y,z) of the furnace. Substitute the initial chlorine concentration distribution field CCI0(x,y,z) into the initial risk field formula to calculate the initial risk field Ψ0(x,y,z) of the entire furnace. S3. Based on the initial risk field Ψ0(x,y,z) and numerical data, and combined with a preset physicochemical constraint model, predict the chlorine concentration distribution (CCI) within a set time period using a transfer function. pred (t+Δt), the chlorine concentration distribution CCI pred Substituting (t+Δt) into the formula for predicting the risk field, we can calculate the dynamic predicted risk field Ψ. pred (t+Δt); S4. Real-time monitoring of the dynamically predicted risk field Ψ pred (t+Δt), and screen out the predicted risk fields that exceed the preset safety threshold as dioxin precursor generation hotspots, and generate suppression instructions by calling the preset suppression strategy library according to the type of the dioxin precursor generation hotspots, such as the main combustion zone, transition zone or wide zone, and output them to the execution device; wherein, the suppression instructions include precise air curtain intervention, inhibitor injection or overall furnace temperature increase. S5. Collect suppressed multimodal data through a sensor array, and invert the actual CI distribution and risk field Ψ using a closed-loop control evaluation formula. actual , in contrast to Ψ pred With Ψ actual The effectiveness of the suppression instructions is evaluated, and data is fed back to the multimodal large model and the physicochemical constraint model to complete parameter optimization, and the preset suppression strategy library is updated synchronously.

2. The method for sensing and inhibiting dioxin precursors in an incinerator according to claim 1, characterized in that, S1 specifically includes: S101. Acquire real-time images of the flame shape, brightness, color distribution, and material spill trajectory inside the incinerator as image sequence data V. S102. Real-time reading of numerical data N of key operating parameters in each area of ​​the furnace; wherein, the key operating parameters include temperature T, primary / secondary air volume and velocity F, furnace pressure P, and flue gas oxygen content O2; S103. Obtain the component analysis report of the waste entering the incinerator as text data. If it is a batch feeding, the text data is updated synchronously according to the feeding batch. S104. Perform data preprocessing on the image sequence data V, numerical data N, and text data Text, including noise filtering, image enhancement, size normalization, outlier removal, data smoothing, and format standardization.

3. The method for sensing and inhibiting dioxin precursors in an incinerator according to claim 2, characterized in that, S2 specifically includes: S201. Input the preprocessed image sequence data V, numerical data N, and text data Text into the pre-trained multimodal large model. The multimodal large model performs feature extraction, fusion, and inference operations on the input data to invert and obtain the initial chlorine concentration values ​​corresponding to each coordinate point (x, y, z) in the three-dimensional space of the furnace, forming a complete three-dimensional initial chlorine concentration distribution field CCI0(x, y, z) of the furnace. S202. Obtain the initial temperature T(x,y,z) and initial flue gas residence time τ(x,y,z) corresponding to each coordinate point (x,y,z) in the initial chlorine concentration distribution field CCI0(x,y,z). Combine the preset catalytic factor function η(Catal(x,y,z)) and fixed adjustment coefficients α, β, γ, and substitute the above parameters into the formula for the initial risk field of the entire furnace to calculate the initial risk value of each coordinate point (x,y,z) in the three-dimensional space of the furnace, thus forming the initial risk field Ψ0(x,y,z) of the entire furnace.

4. The method for sensing and inhibiting dioxin precursors in an incinerator according to claim 3, characterized in that, In S202, the formula for the initial risk field of the entire furnace is: Ψ0(x,y,z)=(CCI0(x,y,z) α )·exp(−β / T(x,t,x))·(τ(x,y,z) γ )·η(center(x,y,z)) Wherein, Ψ0(x,y,z) represents the initial risk field at each coordinate point (x,y,z) in the three-dimensional space of the furnace, quantifying the initial risk of dioxin precursor formation at each location in the entire furnace; CCI0(x,y,z) represents the initial chlorine concentration distribution field at each coordinate point (x,y,z) in the three-dimensional space of the furnace; α is the chlorine concentration influence adjustment coefficient, reflecting the nonlinear influence of chlorine concentration on risk; T(x,y,x) represents the initial temperature at each coordinate point (x,y,z) in the three-dimensional space of the furnace; β is the activation energy related parameter, reflecting the influence of temperature on the reaction rate; τ(x,y,z) represents the initial flue gas residence time at each coordinate point (x,y,z) in the three-dimensional space of the furnace; γ is the time influence coefficient, characterizing the degree of influence of residence time on precursor formation; η(Catal(x,y,z)) is the catalytic factor at each coordinate point (x,y,z) in the three-dimensional space of the furnace, which is related to the concentration and type of metal catalyst in the fly ash at the corresponding location.

5. The method for sensing and inhibiting dioxin precursors in an incinerator according to claim 3, characterized in that, S3 specifically includes: S301. Extract the three-dimensional initial chlorine concentration distribution field CCI0(x,y,z) of the furnace and the initial risk field Ψ0(x,y,z) of the entire furnace. At the same time, retrieve the preprocessed current numerical data N and perform consistency verification on the extracted data. S302. Activate the preset simplified computational fluid dynamics and chemical reaction kinetics model as the physicochemical constraint model. Input the validated initial chlorine concentration distribution field CCI0(x,y,z) and the current numerical data N into the physicochemical constraint model. Perform calculations through the transfer function G to predict the chlorine concentration values ​​at each coordinate point (x,y,z) in the three-dimensional space of the furnace at time t+Δt within a future set time, forming the predicted chlorine concentration distribution field CCI0 at time t+Δt. pred (t+Δt), its calculation formula is: CCI pred (t+Δt)=G(CCI0,F,T,...) Among them, CCI pred (t+Δt) represents the predicted chlorine concentration distribution in the furnace at time t+Δt; G is the transfer function, describing the transport and transformation of chlorine under the influence of airflow, diffusion, and reaction; CCI0 is the initial chlorine concentration distribution field; F is the airflow rate and velocity of the primary / secondary air; T is the current furnace temperature field; ".." represents other parameters that affect the evolution of chlorine concentration. S303, based on CCI pred (t+Δt), the temperature distribution T at time t+Δt predicted by the physicochemical constraint model. pred (t+Δt), flue gas residence time distribution τ pred (t+Δt), Catalyst parameter distribution Catal pred (t+Δt), using fixed adjustment coefficients α, β, γ and the catalytic factor function η(·), are substituted into the predicted risk field formula to calculate the predicted risk value of each coordinate point in the three-dimensional space of the furnace at time t+Δt, forming a dynamic predicted risk field Ψ. pred (t+Δt).

6. The method for sensing and inhibiting dioxin precursors in an incinerator according to claim 5, characterized in that, In S303, the formula for predicting the risk field is: P pred (t+Δt)=(CCI pred (t+Δt) α )·exp(-β / T pred (t+Δt))·(τ pred (t+Δt) γ )·η(Catal pred (t+Δt)) Among them, Ψ pred (t+Δt) represents the predicted risk field at the future time t+Δt, reflecting the risk of dioxin precursor formation in various regions of the furnace in the short term; CCI pred (t+Δt) represents the predicted chlorine concentration distribution at time t+Δt; α is the chlorine concentration influence adjustment coefficient; T pred (t+Δt) represents the predicted furnace temperature distribution at the future time t+Δt; β represents the activation energy-related parameter; τ pred (t+Δt) represents the predicted flue gas residence time distribution at future time t+Δt; γ is the time influence coefficient; Catal pred (t+Δt) represents the predicted distribution of catalyst-related parameters at the future time t+Δt, η(Catal) pred (t+Δt)) represents the catalytic factor at the corresponding position.

7. The method for sensing and inhibiting dioxin precursors in an incinerator according to claim 5, characterized in that, S4 specifically includes: S401, Read the dynamically predicted risk field Ψ pred (t+Δt) performs a real-time, no-dead-angle scan of the entire furnace space, synchronously recording Ψ at each coordinate point. pred Numerical values ​​and corresponding spatial location information (x, y, z); S402. Set a critical risk value for dioxin precursor formation as a safety threshold Ψ threshold Ψ will be monitored in real time pred (t+Δt) and Ψ threshold Compare them one by one and filter out all Ψ pred >Ψ threshold The region serves as a hotspot for the formation of dioxin precursors; S403. Call the preset furnace partition parameters of the incinerator, match the spatial coordinates (x,y,z) of the dioxin precursor generation hotspots with the partition boundaries, and classify the hotspot types: wherein, the furnace partition parameters are the predefined physical range coordinate boundaries of the main combustion zone and the transition zone. S404. According to the hotspot type, call the preset suppression strategy library to match the corresponding strategy, generate a suppression command, and transmit the suppression command to the execution device corresponding to the incinerator in real time to trigger the corresponding suppression operation; wherein, the suppression strategy library stores the corresponding mapping rules of hotspot type-suppression strategy-execution parameter in advance.

8. The method for sensing and inhibiting dioxin precursors in an incinerator according to claim 7, characterized in that, Specifically, S403 includes: If the hot spot is entirely or mainly located in the preset main combustion zone, it is determined to be a hot spot in the main combustion zone; wherein, the main combustion zone is the core combustion area in the middle of the furnace, and the corresponding coordinate range is a preset fixed value; If all or most of the hot spots are located in the preset transition zone, they are determined to be transition zone hot spots; wherein, the transition zone is the connection area between the main combustion zone and the tail flue, and the corresponding coordinate range is a preset fixed value. If a hotspot's coverage area spans a single partition boundary, or is scattered across multiple local areas and its overall coverage area exceeds the range of a single partition, it is considered a widespread hotspot.

9. The method for sensing and inhibiting dioxin precursors in an incinerator according to claim 7, characterized in that, S5 specifically includes: S501. After the device executing the suppression command completes its operation, a new round of multimodal data acquisition is initiated. The acquisition method is the same as in step S1 and maintains real-time performance. S502. Input the newly acquired multimodal data into the pre-trained multimodal large model to invert and obtain the actual chloride concentration distribution field (CCI) after the suppression operation. actual (x,y,z), then CCI actual Substituting the corresponding actual parameters into the evaluation formula of the closed-loop control version, the suppressed actual risk field Ψ is calculated. actual The formula is as follows: Where A is the quantification value of the specific suppression action performed, and K is the preset suppression efficiency coefficient; S503, Extract the dynamically predicted risk field Ψ pred The risk value of the corresponding hotspot region in (t+Δt) is compared with the actual risk field Ψ obtained in step S502. actual The risk values ​​of the same spatial coordinates (x, y, z) are compared one by one: If Ψ actual Below Ψ pred (t+Δt) and below the safety threshold Ψ threshold The current suppression strategy is deemed effective. If Ψ actual Not reduced or still above Ψ threshold The current suppression strategy is deemed ineffective or invalid. S504. Combine the comparison results from step S503, the corresponding suppression action A, the multimodal data acquired after suppression, and the actual CI distribution field CCI. actual As new training samples, they are fed back to the multimodal large model and the physicochemical constraint model, and the inhibition strategy library is automatically updated based on the model learning results and the inhibition effect evaluation conclusions.

10. A dioxin precursor sensing and suppression system for an incinerator, based on the dioxin precursor sensing and suppression method for an incinerator according to any one of claims 1-7, characterized in that, The system includes: The acquisition module is used to synchronously acquire multimodal raw data through a sensor array and preprocess the multimodal raw data; wherein the multimodal raw data includes visual data, numerical data and text data; The first calculation module is used to input the pre-processed multimodal raw data into the pre-trained multimodal large model for inversion to obtain the three-dimensional initial chlorine concentration distribution field CCI0(x,y,z) of the furnace, and substitute the initial chlorine concentration distribution field CCI0(x,y,z) into the initial risk field formula to calculate the initial risk field Ψ0(x,y,z) of the entire furnace. The second calculation module is used to predict the chlorine concentration distribution (CCI) within a set future time period based on the initial risk field Ψ0(x,y,z) and numerical data, combined with a preset physicochemical constraint model, using a transfer function. pred (t+Δt), the chlorine concentration distribution CCI pred Substituting (t+Δt) into the formula for predicting the risk field, we can calculate the dynamic predicted risk field Ψ. pred (t+Δt); The generation module is used to monitor the dynamically predicted risk field Ψ in real time. pred (t+Δt), and screen out the predicted risk fields that exceed the preset safety threshold as dioxin precursor generation hotspots, and generate suppression instructions by calling the preset suppression strategy library according to the type of the dioxin precursor generation hotspots, such as the main combustion zone, transition zone or wide zone, and output them to the execution device; wherein, the suppression instructions include precise air curtain intervention, inhibitor injection or overall furnace temperature increase. The update module is used to collect suppressed multimodal data through a sensor array and invert the actual CI distribution and risk field Ψ using a closed-loop control evaluation formula. actual , in contrast to Ψ pred With Ψ actual The effectiveness of the suppression instructions is evaluated, and data is fed back to the multimodal large model and the physicochemical constraint model to complete parameter optimization, and the preset suppression strategy library is updated synchronously.

Images