Dioxin closed-loop control method based on real-time detection and dynamic optimization

Abstract

A dioxin closed-loop control method based on real-time detection and dynamic optimization is proposed. By deploying an online monitoring system, a tunable laser ionization time-of-flight mass spectrometer is configured at the incineration flue gas discharge port to detect the concentration of dioxin indicator trichlorobenzene in real time. Combined with sensor detection of working condition parameters, the concentration of dioxin toxicity equivalent is converted in real time using polynomial regression to form an integrated system of online monitoring and toxicity assessment. A double-layer random forest-hybrid eagle optimization model is constructed. The first layer is based on historical working condition data to train a random forest model to predict the emission trend of dioxin and the potential risk of exceeding the standard. The second layer dynamically optimizes the incineration parameter combination by improving the eagle optimization algorithm and sends it to the incineration system in real time. At the same time, a calcium-based retarder directional spraying device is deployed in the flue gas purification section to achieve uniform coverage and precise dose control of the retarder and inhibit the low-temperature recombination reaction of dioxin. Finally, the parameters of the double-layer optimization model are dynamically updated based on the error between the actual detection value and the predicted value to improve the adaptability of the model to working condition fluctuations, thereby effectively reducing dioxin emissions and improving the environmental protection and stability of the incineration process.

Description

Technical Field

[0001] This invention relates to the field of environmental monitoring and pollution control technology, and in particular to a closed-loop control method for dioxins based on real-time detection and dynamic optimization. Background Technology

[0002] In today's society, with the acceleration of industrialization and the continuous expansion of urbanization, the problem of waste disposal has become increasingly prominent. Waste incineration, as a common and effective waste treatment method, is widely used. However, the incineration process inevitably produces dioxins, highly toxic substances. Dioxins are extremely toxic and pose a serious threat to human health and the ecological environment; for example, they may cause serious consequences such as endocrine disorders, cancer, and birth defects. Therefore, effective monitoring and control of dioxins generated during waste incineration is particularly important.

[0003] Existing dioxin emission control technologies also have several shortcomings. Some waste incineration plants rely solely on experience or simple fixed parameter settings to regulate the incineration process, such as furnace temperature, oxygen content, and activated carbon injection volume, lacking scientific and precise control strategies. This extensive control approach is ill-suited to the complex and variable composition of waste and the constantly fluctuating incineration conditions, resulting in unstable dioxin emissions and difficulty in consistently meeting increasingly stringent environmental standards. Furthermore, methods for inhibiting low-temperature dioxin resynthesis are relatively limited and inefficient, failing to effectively reduce the secondary formation of dioxins during flue gas purification.

[0004] Furthermore, existing monitoring and control processes are often independent and fail to form an organic whole. Monitoring data cannot be fed back in a timely manner to guide adjustments to control strategies, and the effectiveness of control measures cannot be accurately evaluated and optimized through real-time monitoring. This results in a lack of systematicity and synergy in the entire dioxin emission control system, making it difficult to achieve efficient and precise dioxin emission reduction targets.

[0005] In summary, existing technologies for dioxin monitoring and control suffer from numerous drawbacks, including poor real-time performance, high cost, imprecise regulation, and insufficient system coordination. There is an urgent need for a more advanced, efficient, and accurate technology for dioxin emission monitoring and control to address these issues. Summary of the Invention

[0006] To address the above problems, this invention proposes a dioxin closed-loop control method based on real-time detection and dynamic optimization. The specific steps are as follows:

[0007] Step 1: Deploy an online monitoring system. Configure a tunable laser ionization time-of-flight mass spectrometer at the incineration flue gas emission outlet to detect the concentration of trichlorobenzene, a dioxin indicator, in real time. Detect activated carbon injection rate, furnace temperature, and oxygen content parameters through sensors. Combined with polynomial regression, calculate the dioxin toxicity equivalent concentration in real time to form an integrated online monitoring-toxicity assessment system.

[0008] Step 2: Construct a two-layer random forest-Tianying hybrid optimization model. The first layer is a random forest model trained based on historical operating data to predict dioxin emission trends and potential exceedance risks. The second layer dynamically optimizes the combination of incineration parameters by improving the Tianying optimization algorithm and sends optimization instructions to the incineration system in real time through the industrial Internet of Things.

[0009] Step 3: Deploy a calcium-based inhibitor directional spraying device in the flue gas purification section, using a worm gear drive multi-directional rotary spraying structure to achieve uniform coverage and precise dosage control of the inhibitor in the flue gas channel; adjust the spraying dosage according to the threshold incineration system redundancy correction control to inhibit the low-temperature resynthesis of dioxins.

[0010] Step 4: Calculate the error between the actual dioxin emission concentration and the random forest prediction value. Based on the error data, dynamically update the parameters of the two-layer optimization model and iterate the incremental data to the random forest to improve the model's adaptability to operating condition fluctuations.

[0011] This invention provides a dioxin closed-loop control method based on real-time detection and dynamic optimization, which has beneficial effects. The technical advantages of this invention are as follows:

[0012] 1. This invention constructs a two-layer random forest-Skyhawk hybrid optimization model. The first layer, the random forest model, is trained based on historical operating data and can accurately predict dioxin emission trends and potential exceedance risks, issuing early warnings. The second layer, the improved Skyhawk optimization algorithm, can dynamically optimize the combination of incineration parameters. This combination of accurate prediction and optimized control can adjust incineration parameters in a timely manner according to actual operating conditions, effectively avoiding dioxin exceedance emissions, while improving the stability and energy utilization efficiency of the incineration process, and reducing energy waste and excessive dioxin generation caused by unreasonable parameters.

[0013] 2. This invention deploys a calcium-based inhibitor directional spraying device with a worm gear-driven multi-directional rotary spray structure in the flue gas purification section, achieving uniform coverage of the inhibitor within the flue gas channel. Furthermore, the spraying dosage is precisely controlled based on dioxin emission concentration calculations. When the dioxin emission concentration exceeds a certain threshold, redundant correction control is triggered to adjust the spraying dosage. This not only improves the contact efficiency between the inhibitor and the flue gas but also precisely inhibits the low-temperature resynthesis reaction of dioxins, reducing secondary dioxin formation and further lowering dioxin emissions. Attached Figure Description

[0014] Figure 1 This is a flowchart of the present invention;

[0015] Figure 2 This is a structural diagram of step 1 of the present invention. Detailed Implementation

[0016] The present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments:

[0017] This invention proposes a closed-loop control method for dioxins based on real-time detection and dynamic optimization. Addressing the dioxin emission problem from waste incineration, it constructs a system integrating online monitoring and predictive control. By real-time detection of trichlorobenzene and operating parameters to calculate dioxin concentration, a two-layer model is used to optimize incineration parameters. A directional spray device is deployed to suppress resynthesis, and the model is updated based on errors, effectively reducing dioxin emissions. The invention flowchart is shown below. Figure 1 As shown, the steps of the present invention will be described in detail below.

[0018] Step 1: Deploy an online monitoring system. A tunable laser ionization time-of-flight mass spectrometer is installed at the incineration flue gas emission outlet to monitor the concentration of trichlorobenzene, a dioxin indicator, in real time. Sensors are used to detect parameters such as activated carbon injection rate, furnace temperature, and oxygen content. Combined with polynomial regression, the dioxin toxicity equivalent concentration is calculated in real time, forming an integrated online monitoring and toxicity assessment system. The structure diagram for Step 1 is shown below. Figure 2 As shown.

[0019] Step 1.1: Configure a tunable laser ionization time-of-flight mass spectrometer to detect trichlorobenzene concentration:

[0020] A tunable laser ionization time-of-flight mass spectrometer is installed at the flue gas emission outlet of the incineration plant. The tunable laser is used to ionize the molecules in the flue gas, and the ions are analyzed by time-of-flight mass spectrometry to determine the types and concentrations of molecules.

[0021] When detecting trichlorobenzene, a dioxin indicator, the wavelength of the mass spectrometer laser is adjusted to match the absorption peak of the trichlorobenzene molecule, thereby achieving selective ionization and detection of trichlorobenzene and outputting the trichlorobenzene concentration value in real time.

[0022] Step 1.2, Integrating the operating condition parameter sensor network:

[0023] Install a sensor network to collect operating parameters such as activated carbon injection rate, furnace temperature, and oxygen content.

[0024] A flow meter is used to detect the amount of activated carbon injected, a temperature sensor measures the flue gas temperature, and an oxygen content sensor measures the oxygen content. These sensors transmit real-time measurements to the incineration system via a data acquisition module, ensuring comprehensive monitoring data covering key emission indicators.

[0025] Step 1.3: Calculate the dioxin toxicity equivalent concentration using a multinomial regression model:

[0026] A nonlinear function was established using historical data on dioxin toxicity equivalent concentrations and trichlorobenzene concentrations. By fitting historical datasets, a nonlinear function under a nonlinear mapping relationship was established using polynomial regression.

[0027]

[0028] Where TEQ is the dioxin toxicity equivalent concentration, a k Here are the k-th order coefficients of the polynomial regression model, where m is the order and C is the number of coefficients. TCB This represents the real-time detection concentration of trichlorobenzene.

[0029] Step 1.4: Establish an online monitoring-toxicity assessment system.

[0030] The system collects trichlorobenzene concentration, operating parameters, and dioxin toxicity equivalent concentration information in real time. At the same time, the system will issue alarms for abnormal situations based on the set dioxin toxicity equivalent concentration threshold, forming an integrated online monitoring and toxicity assessment system, which provides basic data and assessment basis for subsequent dioxin emission control.

[0031] Step 2: Construct a two-layer random forest-Tianying hybrid optimization model. The first layer is based on historical operating data to train a random forest model to predict dioxin emission trends and potential exceedance risks. The second layer dynamically optimizes the combination of incineration parameters by improving the Tianying optimization algorithm and sends optimization instructions to the incineration system in real time through the Industrial Internet of Things.

[0032] Step 2.1, First Layer: Train a random forest model based on historical operating condition data

[0033] Historical operating data was collected, including furnace temperature, oxygen content, and activated carbon injection rate. Simultaneously, the corresponding dioxin emission concentrations were recorded, and this historical data was compiled into dataset D.

[0034] D = {(T fuenace1 O 21 M AC1 C Dioxin1 ),(T fuenace2 O 22 M AC2 C Dioxin2 ),...,(T fuenacen O 2n M ACn C Dioxinn )}

[0035] Among them, T fuenacen For the furnace temperature of the nth sample, O 2n M represents the oxygen content of the nth sample.ACn C represents the amount of activated carbon injected for the nth sample. Dioxinn Let be the dioxin emission concentration of the nth sample.

[0036] Constructing a Random Forest Model: A random forest consists of multiple decision trees. For each decision tree, m samples are randomly drawn with replacement from the dataset D to form a new training subset D. i At the same time, two features are randomly selected from all features to perform splits at the nodes of the decision tree.

[0037] At each node of the decision tree, based on two selected features, the Gini coefficient is used to select the optimal feature for splitting, dividing the sample into different child nodes, until the number of samples in a node is less than a threshold or the tree depth reaches its upper limit. This process is repeated to construct multiple decision trees T1, T2, ..., T N , where N is the number of decision trees in the random forest.

[0038] For the new operating condition data (T) fuenace_new O 2_new M AC_new ), and input it into each decision tree T i In the middle, the predicted dioxin emission concentration C was obtained. Dioxin_predi Then, the final predicted value is obtained by integrating the predictions:

[0039]

[0040] Among them, C Dioxin_predi Let C be the dioxin concentration prediction value of the i-th decision tree, and C be the final prediction value of the random forest model. Dioxin_pred According to the established dioxin emission standard threshold C threshold If the following formula is satisfied:

[0041] C Dioxin_predi ≥C threshold

[0042] If a potential risk of exceeding the standard is detected, an over-standard warning will be triggered, and the second layer of optimization will be initiated.

[0043] Step 2.2, Second Layer: Dynamically optimize the combination of incineration parameters by improving the Tianying optimization algorithm.

[0044] A Levy flight perturbation mechanism is introduced into the basic Eagle optimization algorithm to enhance its global search capability. Let the position of the individual eagle be X = [x1, x2, ..., x...]. d ], where d = 3 is the parameter dimension. In each iteration, the position of the individual eagle is subjected to Levy flight perturbation:

[0045]

[0046] Among them, X new Let α be the new position vector after the Levy flight perturbation, and α be a scaling factor. This represents element-wise multiplication. Levy(β) is a vector of random numbers following a Levy distribution, and β is the parameter of the Levy distribution. To balance the global and local search capabilities of the algorithm, an adaptive inertia weight w is introduced. w changes dynamically with the number of iterations t:

[0047]

[0048] Among them, w max and w min These represent the maximum and minimum values ​​of the inertia weight, respectively, where t is the current iteration number, and T is the maximum and minimum values. max This represents the maximum number of iterations.

[0049] The objective function F(X) is set as follows:

[0050] F(X) = w1 × C Dioxin (X)+w2×E(X)

[0051] Where C Dioxin F(X) represents the dioxin emission concentration under parameter combination X, E(X) is the energy consumption, and w1 and w2 are weighting coefficients. The improved Skyhawk optimization algorithm iteratively updates the position of individual Skyhawks to find the parameter combination that minimizes the objective function F(X).

[0052] The constraints are set as follows: furnace temperature control accuracy within 5℃, oxygen content adjustment response time less than 30 seconds, and activated carbon injection quantity matching flue gas load.

[0053] Finally, the Skyhawk population was initialized, 50 sets of incineration parameter combinations were randomly generated, F(X) for each set was calculated, the optimal individual update position was selected, and the iteration was continued until convergence. This yielded a precisely controlled combination of furnace temperature, oxygen content, and activated carbon injection amount to reduce dioxin concentration.

[0054] Step 2.3: Optimization instructions are sent to the incineration system in real time via the Industrial Internet of Things (IIoT):

[0055] The optimal combination of incineration parameters obtained through the two-layer random forest-Skyhawk hybrid optimization model is transmitted to the incineration system in real time via the Industrial Internet of Things. The incineration system adjusts the parameters of furnace temperature, oxygen content, and activated carbon injection in real time to optimize dioxin emissions and the incineration process.

[0056] Step 3: Deploy a calcium-based inhibitor directional spraying device in the flue gas purification section, using a worm gear drive multi-directional rotary spraying structure to achieve uniform coverage and precise dosage control of the inhibitor in the flue gas channel; adjust the spraying dosage according to the threshold incineration system redundancy correction control to suppress the low-temperature resynthesis of dioxins.

[0057] Step 3.1, Deploy the calcium-based inhibitor directional spraying device

[0058] A specially designed calcium-based inhibitor directional spraying device is installed in the flue gas purification section. This device adopts a worm gear driven multi-directional rotary spray structure to ensure uniform coverage of the inhibitor within the flue gas channel, thereby guaranteeing sufficient contact between the inhibitor and the flue gas. The worm gear drives the multi-directional nozzle, and by adjusting the rotation angle from 0° to 360°, it achieves full coverage of the cross-section of the flue gas channel. At the same time, the injection pressure is dynamically adjusted according to the flue gas flow rate to ensure the atomized particle size of the inhibitor and improve the contact efficiency with the flue gas.

[0059] Step 3.2, Precision Dosage Control Model

[0060] If the dioxin emission concentration calculated in step 1 exceeds 1.5 times the threshold, the incineration system will trigger the redundancy correction control for the inhibitor spray dosage, and the inhibitor spray dosage calculation formula will be corrected as follows:

[0061]

[0062] Where D is the dose of the inhibitor spray, k is the dose coefficient, and C Dioxin_pred M represents the predicted dioxin emission concentration from the random forest in step 2. ac T represents the activated carbon injection rate output by the Skyhawk hybrid optimization model in step 2. fuenace T is the current furnace temperature. opt γ is the optimal furnace temperature setpoint output by the Tianying hybrid optimization model in step 2, Q is the oxygen content output by the Tianying hybrid optimization model in step 2, and γ is the temperature correction coefficient.

[0063] Step 4: Calculate the error between the actual dioxin emission concentration and the random forest prediction value. Based on the error data, dynamically update the parameters of the two-layer optimization model and iterate the incremental data to the random forest to improve the model's adaptability to operating condition fluctuations.

[0064] Calculate the closed-loop control error rate E:

[0065]

[0066] Among them, TED actual TED represents the actual measured dioxin toxicity equivalent concentration. predictedThe E-value represents the dioxin toxicity equivalent concentration predicted by the model. This formula measures the deviation between the model's prediction and the actual detection value. When the calculated E-value exceeds 8%, it indicates that the parameters of the random forest model need to be adjusted. New operating data and historical data are merged in a 7:3 ratio to generate an incremental dataset, and the random forest parameters are then adjusted.

[0067] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any other way. Any modifications or equivalent changes made based on the technical essence of the present invention shall still fall within the scope of protection claimed by the present invention.

Claims

1. A dioxin closed-loop control method based on real-time detection and dynamic optimization, the specific steps are as follows, characterized by: Step 1: Deploy an online monitoring system, configure a tunable laser ionization time-of-flight mass spectrometer at the incineration flue gas discharge port to detect the concentration of the dioxin indicator trichlorobenzene in real time; detect the active carbon injection amount, furnace temperature, oxygen content and other working condition parameters through sensors, and convert the dioxin toxicity equivalent concentration in real time by combining polynomial regression to form an online monitoring-toxicity evaluation integrated system; Step 1 in the deployment of the online monitoring system is represented as: Step 1.1, configure a tunable laser ionization time-of-flight mass spectrometer to detect the concentration of trichlorobenzene: Install a tunable laser ionization time-of-flight mass spectrometer at the incineration flue gas discharge port, use a tunable laser to ionize molecules in the flue gas, analyze ions by time-of-flight mass spectrometry to determine the type and concentration of molecules; When detecting the dioxin indicator trichlorobenzene, adjust the wavelength of the mass spectrometer to match the absorption peak of the trichlorobenzene molecule, achieve selective ionization and detection of trichlorobenzene, and output the trichlorobenzene concentration value in real time; Step 1.2, integrate the working condition parameter sensor network: Install a sensor network to collect working condition parameters such as active carbon injection amount, furnace temperature, and oxygen content; Use a flow meter to detect the active carbon injection amount, a temperature sensor to measure the flue gas temperature, and an oxygen content sensor to measure the oxygen content; the sensors transmit real-time measurement data to the incineration system through a data acquisition module to ensure that the monitoring data fully covers the key emission indicators; Step 1.3, convert the dioxin toxicity equivalent concentration by combining a polynomial regression model: Establish a nonlinear function of historical dioxin toxicity equivalent concentration and indicator trichlorobenzene concentration data, fit the historical data set, and establish a nonlinear function under the nonlinear mapping relationship by using polynomial regression; ； wherein, is the dioxin toxicity equivalent concentration, is the kth order coefficient of the polynomial regression model, is the order, is the real-time detection concentration of trichlorobenzene; Step 1.4, form an online monitoring-toxicity evaluation system: The system collects trichlorobenzene concentration, working condition parameters, and dioxin toxicity equivalent concentration information in real time, and simultaneously alarms for abnormal conditions according to the set dioxin toxicity equivalent concentration threshold, forming an online monitoring-toxicity evaluation integrated system to provide basic data and evaluation basis for subsequent dioxin emission control; Step 2: Build a double-layer random forest-hybrid eagle optimization model, the first layer trains a random forest model based on historical working condition data to predict dioxin emission trends and potential over-standard risks; the second layer dynamically optimizes the incineration parameter combination by improving the eagle optimization algorithm, and sends the optimization instructions to the incineration system in real time through the industrial Internet of Things; Step 3: Deploy a calcium-based retardant directional spraying device in the flue gas purification section, use a worm gear transmission multi-directional rotary spraying structure to achieve uniform coverage and precise dose control of the retardant in the flue gas channel; adjust the spraying amount according to the threshold incineration system redundancy correction control to suppress the dioxin low-temperature resynthesis reaction; Step 4: Calculate the error between the actual dioxin emission concentration detection value and the random forest prediction value, dynamically update the double-layer optimization model parameters based on the error data, and iterate the incremental data to the random forest to improve the adaptability of the model to working condition fluctuations.

2. The real-time detection and dynamic optimization based dioxin closed-loop control method according to claim 1, characterized in that: The double-layer random forest-Yingying hybrid optimization model in step 2 is represented as follows: Step 2.1, first layer: training random forest model based on historical operating data Collect historical operating data, including furnace temperature, oxygen content, activated carbon injection amount; at the same time, record the corresponding dioxin emission concentration, and form a data set with the historical data : ； in, For the first The furnace temperature of each sample For the first Oxygen content of each sample For the first The amount of activated carbon sprayed per sample For the first Dioxin emission concentration of each sample; Building a random forest model: A random forest is composed of multiple decision trees. For each decision tree, m samples are randomly drawn with replacement from the dataset D to form a new training subset ; at the same time, 2 features are randomly selected from all features for splitting at the nodes of the decision tree; At each node of the decision tree, based on the selected 2 features, the Gini coefficient is used to select the optimal feature for splitting, and the samples are divided into different sub-nodes until the number of node samples is less than the threshold or the depth of the tree reaches the upper limit; repeat this process to build multiple decision trees where N is the number of decision trees in the random forest; For new operating data which is input into each decision tree to obtain a predicted dioxin emission concentration and the final prediction value is obtained by integrating the predictions ； wherein, Dioxin concentration prediction value of the i-th decision tree, final prediction value of the random forest model , According to the set threshold value of dioxin emission if the following equation is satisfied: ； If it is determined that there is a potential risk of exceeding the standard, an early warning of exceeding the standard is triggered, and the second layer optimization is started. Step 2.2, second layer: dynamically optimizing the incineration parameter combination by improving the Yingying optimization algorithm In the basic algorithm, a Levy flight perturbation mechanism is introduced to enhance the global search ability of the algorithm; the position of the hawk individual is where is the parameter dimension; in each iteration, the position of the hawk individual is perturbed by Levy flight: ； where, is the new position vector after Levy flight perturbation, is a scaling factor, denotes element-wise multiplication, is a random number vector subject to Levy distribution, is a parameter of Levy distribution; to balance the global search and local search ability of the algorithm, an adaptive inertia weight is introduced ; dynamically changes with the increase of the number of iterations t: ； wherein, and are the maximum and minimum values of the inertial weight, respectively, is the current iteration number, is the maximum number of iterations; Objective function is set to: ； wherein is the dioxin emission concentration under the parameter combination X, is the energy consumption, and is the weight coefficient; the position of the hawk individual is updated constantly by the improved hawk optimization algorithm, and the parameter combination that makes the target function minimum is searched. The constraint condition is set as: the furnace temperature control precision is less than 10 ℃ , the oxygen content regulation response time is less than 30 seconds, and the activated carbon injection amount matches the flue gas load. Finally, the initial population of eagles is initialized, 50 groups of incineration parameter combinations are randomly generated, the concentration of dioxin in each group is calculated , the optimal individual is selected to update the position, and iteration is performed until convergence to obtain the precise control of the furnace temperature, oxygen content and activated carbon injection amount combination to reduce the concentration of dioxin; Step 2.3, the optimization instruction is transmitted to the incineration system in real time through the industrial Internet of Things: The optimal incineration parameter combination obtained by the double-layer random forest-Yingying hybrid optimization model is transmitted to the incineration system in real time through the industrial Internet of Things, and the real-time adjustment of the furnace temperature, oxygen content, and activated carbon injection amount parameters of the incineration system is achieved, thereby optimizing the dioxin emission and the incineration process.

3. The real-time detection and dynamic optimization based dioxin closed-loop control method according to claim 1, characterized in that: Step 3 in the redundant correction control adjustment of the spraying amount is represented as follows: Step 3.1, deploy calcium-based retardant directional spraying device Install a specially designed calcium-based retardant directional spraying device in the flue gas purification section. The device uses a worm gear transmission multi-directional rotary spraying structure to ensure uniform coverage of the retardant in the flue gas passage, thereby ensuring sufficient contact between the retardant and the flue gas. The worm gear drives the multi-directional nozzle to achieve full coverage of the flue gas passage cross-section by adjusting the rotation angle from 0° to 360°, while dynamically adjusting the injection pressure according to the flue gas flow to ensure the particle size of the retardant atomized particles and improve the contact efficiency with the flue gas. Step 3.2, precise dose control model If the dioxin emission concentration calculated in step 1 exceeds 1.5 times the threshold value, the incineration system triggers the redundant correction control of the retardant spraying amount, and the retardant spraying amount calculation formula is modified as follows: ； wherein, is the retarder spray dose, is the dose coefficient, is the dioxin emission concentration prediction value of step 2 random forest, is the activated carbon injection amount output by the step 2 vulture hybrid optimization model, is the current furnace temperature, is the optimal furnace temperature set value output by the step 2 vulture hybrid optimization model, is the oxygen content output by the step 2 vulture hybrid optimization model, is the temperature correction coefficient.

4. The real-time detection and dynamic optimization based dioxin closed-loop control method according to claim 1, characterized in that: Step 4 in updating the parameters of the double-layer optimization model is represented as follows: Calculate the closed-loop control error rate E: ； wherein, is the actually detected dioxin toxicity equivalent concentration, is the model predicted dioxin toxicity equivalent concentration, and the formula is used to measure the deviation between the model predicted value and the actually detected value. When the calculated E value exceeds 8%, it indicates that the parameters of the random forest model need to be adjusted. The newly added working condition data and the historical data are combined in a 7:3 ratio to generate an incremental data set, and the random forest parameters are adjusted.

Images